Industrial automation with 5G and beyond

The present disclosure is related to wireless communications networks and describes network architecture, wireless devices, and wireless network nodes suitable for industrial applications, using a fifth-generation (5G) or other wireless communications network.

The fifth generation of mobile technology (5G) will be able to provide wider range of services than the existing 3G/4G technologies. Three main use cases of 5G are: Enhanced Mobile Broadband (eMBB), Massive Machine Type of Communication (mMTC) and Ultra Reliable Low Latency Communication (URLLC). A key objective of the 5G system is to be able to support the stringent system requirements from vertical markets. Those requirements include simultaneously supporting multiple combinations of reliability, latency, throughput, positioning, and availability, as well as, local deployments with local survivability, local data/routing, local managements, security, data integrity and privacy.

An industrial network perspective of 5G system is illustrated in FIG. 1 . The service performance requirements are coming from the automation applications. 5G system is providing the communication service to the automation applications. In order to support automation in vertical domains, 5G systems need to be reliable and flexible to meet service performance requirements to serve specific applications and use cases. They need to come with the system properties of reliability, availability, maintainability, safety, and integrity.

Specifications for 5G are under development by members of the 3 rd -Generation Partnership Project (3GPP). The document “Service Requirements for Cyber-Physical Control Applications in Vertical Domains, Stage 1,” 3GPP TS 22.104, v. 16.0.0 (January 2019), specifies the requirements that provide various sets of performance criteria that need to be met to satisfactorily support different use cases of cyber-physical control applications used by various vertical markets.

In the industrial applications space, requirements include support for mixed services in factory and manufacturing environments, including support for different service levels, such as massive Machine-Type Communications (mMTC), enhanced Mobile Broadband (eMBB), and ultra-reliable low-latency communications (URLLC) traffic in the same deployment. Support for industrial deterministic service is needed. Integration between the 5G System (5 GS) and existing industrial networks is also required. Interoperability, including support for non-public networks and interoperability with the public land mobile network (PLMN) is required.

With respect to system availability and reliability, the 5G system as a communication service provider shall comply with the 3GPP definition of availability and reliability. Communication service availability is defined as the percentage value of the amount of time the end-to-end communication service is delivered according to an agreed quality of service (QoS), divided by the amount of time the system is expected to deliver the end-to-end service according to the specification in a specific area. Required availability is to be determined by business aspects considering the trade-off between monetary loss at times the system is not available vs. complexity to increase the availability, e.g. by increasing redundancy. It will be appreciated that availability beyond 99.95% usually requires an extra power source to prevent the public energy grid (99.9-99.99% availability in Europe) from becoming the weakest component.

Communication service reliability is defined as the ability of the communication service to perform as required for a given time interval, under given conditions. These conditions include aspects that affect reliability, such as: mode of operation, stress levels, and environmental conditions. Reliability may be quantified using appropriate measures such as mean time to failure, or the probability of no failure within a specified period of time.

The use of 5G in industrial applications must meet safety requirements, where safety is defined as the condition of being protected from or unlikely to cause danger, risk, or injury. Safe systems thus should be designed to be functionally safe from the start. Automatic protection functions can be built into the system to ensure safety for the system while in operation. Safety aspects to be considered in system design to ensure automatic protection should, for example, include human errors, hardware and software failures, and operational and environmental stress factors.

Many industries today are in full control of their local network deployments. Thus, local deployment aspect regarding local survivability, local data/routing and local management become requirements for industrial networks. In short, the factory network should run normally even when the connection to the outside world is lost. Furthermore, there may be requirements around data not leaving the premises as well as local IT staff being able to manage and change the network deployment on demand.

Security, data integrity and privacy are important requirements for the industries as well. Business critical information on processes and data from the manufacturing process should not be leaked.

Described herein in detail are various techniques for enhancing performance in Industrial Internet-of-Things (IIoT) scenarios, including techniques for time-sensitive networking (TSN) and 5G wireless network integration. Corresponding devices and nodes are also described in detail.

An example method, performed by a wireless device comprises receiving system information (SI) from a radio base station (RBS) of a radio access network (RAN), the SI being indicative of support for TSN through the RBS, and establishing at least one TSN stream with an external data network, through the RBS. The example method further includes receiving a first timing signal from the wireless communications network, via the RBS, receiving a second timing signal from the external TSN data network to which the wireless device is connected, comparing the first timing signal to the second timing signal to determine an offset, and transmitting the offset to the wireless communications network.

Another example method is performed in one or more nodes of a core network associated with a radio access network (RAN) and is for handling a time-sensitive data stream associated with a user equipment (UE) and an external network. This example method comprises receiving, from the external network, a transmission schedule associated with a time-sensitive data stream and sending, to the RAN, a request to allocate radio resources for communication of the data stream between the RAN and a first UE, wherein the request further comprises information related to the transmission schedule. The method further comprises receiving, from the RAN, a response indicating whether radio resources can be allocated to meet the transmission schedule associated with the data stream. The method still further comprises obtaining configuration information for the data stream, the configuration information indicating respective values for one or more fields within a header of data packets associated with the data stream which are to remain static, initiating transmission of the configuration information to the first UE, receiving a data packet associated with the data stream from the external data network, removing the one or more fields from the data packet to generate a compressed data packet, and initiating transmission of the compressed data packet to the first UE.

Another example method is performed by a wireless device associated with a wireless communications network and is for transport of data packets associated with a data stream in an external data network. This example method comprises receiving SI from an RBS of a RAN, the SI being indicative of support for TSN through the RBS, and establishing at least one TSN stream with the external data network, through the RBS. This method further comprises obtaining configuration information for the TSN stream, the configuration information indicating respective values for one or more fields within a header of data packets associated with the TSN stream which are to remain static, receiving, from the RBS, a data packet associated with the TSN stream, and adding the one or more fields to the data packet to generate a decompressed data packet.

Another example method is performed by a wireless device configured for communication with a RAN and is for scheduling resources in the RAN according to a transmission schedule associated with an external network. This example method comprises receiving SI from an RBS of the RAN, the SI being indicative of support for TSN through the RBS, and establishing at least one TSN stream with the external data network, through the RBS. This example method further comprises receiving, from the external network, a transmission schedule associated with the TSN stream, sending, to a network associated with the RAN, a request to allocate radio resources for communication of the TSN stream between the wireless device and the RAN, wherein the request further comprises information related to the transmission schedule, and receiving, from the network, a response indicating whether radio resources can be allocated to meet the transmission schedule associated with the TSN stream

Still another example method, performed by a wireless device, comprises receiving SI from an RBS of a RAN, the SI being indicative of support for TSN through the RBS, and establishing at least one TSN stream with an external data network, through the RBS. This method further comprises obtaining configuration information for the TSN stream, the configuration information indicating respective values for one or more fields within a header of data packets associated with the TSN stream which are to remain static. The method further comprises receiving, from the RBS, a data packet associated with the TSN stream, and adding the one or more fields to the data packet to generate a decompressed data packet.

Yet another example method is performed by a first device, and is for assisting enrollment of a second device to an Internet of Things (IoT) environment and using the second device. This example method comprises obtaining a representation of an enrollment function associated with the second device, wherein the enrollment function is associated with at least one serialized enrollment application comprising enrollment information associated with the first and second device, deserializing the enrollment application such that enrollment information associated with the first device is separated from enrollment information associated with the second device, and transmitting the enrollment information associated with the second device to the second device for initiating execution by the second device of the enrollment process of the second device by configuring the second device based on the enrollment information associated with the second device. This method further comprises receiving, from the second device, configuration information associated with the second device, and using a first runtime environment executing on the first device to transfer a code module to a second runtime environment executing on the second device, where the code module is configured to execute within the second runtime environment and expose a function of the second device, supported by the second runtime environment, to the first device. The method further comprises executing an application within the first runtime environment, the application remotely invoking the function of the second device via the transferred code module and the second runtime environment.

A corresponding method is carried out by a second device and is for executing an enrollment process to an IoT environment assisted by a first device and providing the first device with access to a function of the second device. This example method comprises receiving, from the first device, enrollment information associated with the second device, executing the enrollment process by configuring the second device based on the enrollment information, and transmitting configuration information associated with the second device to the first device. The method further comprises receiving a code module from a first runtime environment executing on the first device, to a second runtime environment executing on the second device, to expose a function of the second device supported by the second runtime environment to the first device, and using the second runtime environment to control performance of the function of the second device responsive to a remote invocation of the function received via the code module from an application executing within the first runtime environment.

These and other methods are described in detail below and illustrated in the attached figures. Corresponding devices, network nodes, and the like are also described in detail, as are the network arrangements and environments in which these techniques may be advantageously used.

FIG. 1 illustrates a network perspective of the 5G system.

FIG. 2 illustrates the concept of Industry 4.0.

FIG. 3 shows a standalone 5G non-public network integrated into an Operations Technology (OT) system.

FIG. 4 shows a 5G non-public network interworking with a public wide-area network.

FIG. 5 illustrates the concept of network slices.

FIG. 6 shows an example of four different slices throughout the network.

FIG. 7 illustrates features of network slices.

FIG. 8 shows mechanisms for slicing the network.

FIG. 9 illustrates QoS in the 5G system.

FIG. 10 shows resource partitioning between network slices.

FIG. 11 shows an example logical function split in motion control applications.

FIG. 12 shows control functions in a cloud.

FIG. 13 illustrates an architecture for remote robot control over a modelled wireless link.

FIG. 14 illustrates an example of a collaborative manufacturer-agnostic robot assembly.

FIG. 15 shows principles of TDOA geolocation.

FIG. 16 shows cumulative distributions for positioning in the 3GPP Indoor Open Office (IOO) scenario, using different bandwidths.

FIG. 17 illustrates principles of hybrid positioning.

FIG. 18 provides a regulatory view of spectrum leasing.

FIG. 19 shows spectrum allocation possibilities for a frequency band allocated to mobile services.

FIG. 20 illustrates a local network using licensed spectrum.

FIG. 21 illustrates a local network using leasing from a license holder, such as a mobile network operator (MNO).

FIG. 22 shows features of CBRS.

FIG. 23 shows a high-level SAS architecture.

FIG. 24 illustrates PAL protection areas.

FIG. 25 shows an industrial cloud scenario.

FIG. 26 illustrates information management in a simple factory situation.

FIG. 27 illustrates a hierarchical network architecture in a factory.

FIG. 28 shows different packet services and quality-of-service relations.

FIG. 29 introduces concepts of time-sensitive networking (TSN).

FIG. 30 illustrates an example TSN and 5G interworking architecture in an industrial scenario.

FIG. 31 shows the use of virtual endpoints to connect non-TSN devices to a TSN network using 5G.

FIG. 32 illustrates TSN time synchronization across a 5G network.

FIG. 33 shows support of multiple time domains in a 5G system.

FIG. 34 shows multiple time domains in a factory network.

FIG. 35 illustrates time-gated queuing.

FIG. 36 shows frame replication and elimination for reliability.

FIG. 37 shows a fully distributed model for TSN.

FIG. 38 illustrates a centralized network/distributed user model for TSN.

FIG. 39 illustrates a fully centralized configuration model for TSN.

FIG. 40 shows a configuration agent consisting of CUC and CNC.

FIG. 41 shows interaction between CNC and CUC.

FIG. 42 is a signal flow diagram illustrating TSN stream setup in a TSN centralized configuration model.

FIG. 43 shows a potential 5G-TSN integration architecture setup.

FIG. 44 illustrates TSN FRER setup.

FIG. 45 shows interaction between AF, CUC, and CNC to setup FRER.

FIG. 46 shows a 5G network.

FIG. 47 illustrates the chain controller concept.

FIG. 48 shows a high-level functional view of a core network deployment at a factory site.

FIG. 49 illustrates the control plane of the RAN, for multi-connectivity.

FIG. 50 illustrates the user plane architecture of the RAN, for multi-connectivity.

FIG. 51 illustrates different radio bearer types for NR.

FIG. 52 shows latency performance when using mini-slots.

FIG. 53 illustrates long alignment delay due to transmission across slot border restriction.

FIG. 54 shows the use of mini-slot repetitions across a slot border.

FIG. 55 shows the use of a beta-factor to allow omission of UCI on PUSCH.

FIG. 56 illustrates a short PUCCH that occupies 1 OFDM symbol, with a periodicity of 2 symbols.

FIG. 57 shows examples of blocking probability per monitoring occasion as a function of DCI size, number of UEs, and CORESET sizes.

FIG. 58 shows downlink data latency with one retransmission.

FIG. 59 shows uplink data latency with a configured grant and one retransmission.

FIG. 60 illustrates a comparison of downlink data latency.

FIG. 61 illustrates a comparison of grant-based uplink data latency.

FIG. 62 shows a comparison of configured grant uplink data latency.

FIG. 63 shows uplink inter-UE pre-emption.

FIG. 64 shows the performance of MCS14 in a power-controlled multiplexing scheme.

FIG. 65 shows PDSCH BLER after one transmission, for several different modulation coding schemes.

FIG. 66 shows uplink SINR for different multi-antenna techniques, with and without coordinated multipoint and uplink precoding.

FIG. 67 shows an example of scheduling request (SR) and buffer status report (BSR) operation.

FIG. 68 illustrates multiple SR configurations mapped to different traffic.

FIG. 69 shows delayed SR due to ongoing long UL-SCH transmission.

FIG. 70 shows a delay in obtaining a dynamic grant via SR procedures.

FIG. 71 illustrates configured grant Type 1 procedures.

FIG. 72 illustrates configured grant Type 1 procedures.

FIG. 73 shows industrial deterministic streams with different arrivals and payload sizes.

FIG. 74 shows industrial deterministic streams with different patterns, periodicities, latency, and reliability requirements.

FIG. 75 illustrates overlapping configurations.

FIG. 76 shows an example of logical channel prioritization (LCP) procedures.

FIG. 77 shows a problem with sending non-critical traffic over a robust grant.

FIG. 78 illustrates a restriction to avoid the problem of FIG. 77 .

FIG. 79 shows the extra latency arising from sending critical traffic over non-robust short grant.

FIG. 80 illustrates a restriction to avoid the problem of FIG. 79 .

FIG. 81 illustrates a problem with a dynamic grant overriding a configured grant.

FIG. 82 shows the benefit of enabling configured grant to override dynamic grant conditionally.

FIG. 83 shows overlapping grant with different PUSCH durations.

FIG. 84 illustrates the enabling of intra-UE preemption to enhance network efficiency.

FIG. 85 shows packet duplication in dual-carrier (DC) and carrier aggregation (CA) scenarios.

FIG. 86 shows residual errors with and without duplication.

FIG. 87 shows universal time domain and working clock domains.

FIG. 88 illustrates SFN transmissions.

FIG. 89 illustrates an industrial use case with three time domains.

FIG. 90 shows a continuous PTP chain method.

FIG. 91 shows an example of the IEEE 802.3 MAC frame format.

FIG. 92 shows gains from Ethernet header compression.

FIG. 93 shows possible Ethernet header compression anchor points.

FIG. 94 shows radio link failure (RLF) in the case of PDCP duplication.

FIG. 95 illustrates an example mobility procedure.

FIG. 96 shows possible realizations of the Industrial IoT protocol stack mapped to the OSI model.

FIG. 97 shows industrial Ethernet categorization.

FIG. 98 illustrated time-scheduled transmissions as used in Profinet.

FIG. 99 shows a frame structure for Profinet IRT.

FIG. 100 illustrates estimated performance of different wireless technologies with respect to reliability with increasing load and increasing E2E latency requirements.

FIG. 101 shows typical channel access and data exchange in Wi-Fi.

FIG. 102 shows channel access in Wi-Fi.

FIG. 103 illustrates a simulation of the Minstrel algorithm.

FIG. 104 shows possible protocol stacks of OPC-UA.

FIG. 105 illustrates OPC-UA over TSN.

FIG. 106 is a block diagram illustrating a Distributed Time-Sensitive Networking (TSN) configuration model, as specified in IEEE Std. 802.1Qbv-2015.

FIG. 107 is a block diagram illustrating a Centralized TSN configuration model, as specified in IEEE Std. 802.1Qbv-2015.

FIG. 108 is a block diagram illustrating a Fully Centralized TSN configuration model, as specified in IEEE Std. 802.1Qbv-2015.

FIG. 109 shows a sequence diagram of an exemplary TSN stream configuration procedure using the fully centralized configuration model shown in FIG. 108 .

FIG. 110 is a block diagram illustrating a control plane (CP) and a data or user plane (UP) architecture of an exemplary 5G wireless network.

FIG. 111 is a block diagram illustrating an exemplary arrangement for interworking between the 5G network architecture shown in FIG. 110 and an exemplary fully centralized TSN network architecture.

FIG. 112 is a block diagram illustrating transmission selection among traffic queues based on gates, as specified in IEEE Std. 802.1Qbv-2015.

FIG. 113 is a block diagram illustrating an exemplary communication scenario between two TSN talker/listener units via 5G and TSN networks, according to various exemplary embodiments of the present disclosure.

FIG. 114 shows a sequence diagram of an exemplary method and/or procedure for configuring timely delivery of TSN stream packets via the network configuration shown in FIG. 113 , according to various exemplary embodiments of the present disclosure.

FIG. 115 is a block diagram illustrating an exemplary communication scenario between a TSN talker/listener unit and a virtualized controller via a 5G network, according to various exemplary embodiments of the present disclosure.

FIG. 116 shows a sequence diagram of an exemplary method and/or procedure for configuring timely delivery of TSN stream packets via the network configuration shown in FIG. 115 , according to various exemplary embodiments of the present disclosure.

FIG. 117 is a flow diagram illustrating an exemplary method and/or procedure performed by a network node in a core network (e.g., a 5G core network), according to various exemplary embodiments of the present disclosure.

FIG. 118 is a flow diagram illustrating an exemplary method and/or procedure performed by a network node in a radio access network (e.g., NG-RAN), according to various exemplary embodiments of the present disclosure.

FIG. 119 is a flow diagram illustrating an exemplary method and/or procedure performed by user equipment (UE), according to various exemplary embodiments of the present disclosure.

FIG. 120 is a block diagram of an exemplary communications system, according to various exemplary embodiments of the present disclosure.

FIGS. 121 , 122 , and 123 are block diagrams of exemplary radio access nodes configured in various ways according to various exemplary embodiments of the present disclosure.

FIGS. 124 and 125 are block diagrams of exemplary wireless devices or UEs configured in various ways, according to various exemplary embodiments of the present disclosure.

FIG. 126 illustrates 5G Core Network (SGCN) functions and Radio Access Network (RAN).

FIG. 127 shows protocol stacks for Ethernet PDU type data.1

FIG. 128 illustrates the TSN Frame Structure.

FIG. 129 is a signaling diagram for downlink signaling according to embodiments of the disclosure.

FIG. 130 is a signaling diagram for uplink signaling according to embodiments of the disclosure.

FIG. 131 illustrates a method in accordance with some embodiments.

FIG. 132 illustrates another method in accordance with some embodiments.

FIG. 133 illustrates another method in accordance with some embodiments.

FIG. 134 illustrates another method in accordance with some embodiments.

FIG. 135 shows a flowchart for implementing a method of handling Time-Sensitive Networking over a radio access network.

FIG. 136 shows a flowchart for implementing a method of announcing Time-Sensitive Networking over a radio access network.

FIG. 137 shows a flowchart for implementing a method of distributing a configuration message for Time-Sensitive Networking over a radio access network.

FIG. 138 shows a schematic block diagram of a first example of a communication system.

FIG. 139 is a schematic block diagram of a second example of a communication system.

FIG. 140 is a schematic block diagram of a third example of a communication system.

FIG. 141 is a functional block diagram of a fourth example of a communication system.

FIG. 142 shows a first schematic signaling diagram for a communication system.

FIG. 143 is a second schematic signaling diagram for a communication system.

FIG. 144 illustrates the inter-working of 5G and TSN.

FIG. 145 shows multiple TSN gPTP time domains in a factory.

FIG. 146 illustrates how a BS can synchronize a UE to a cellular reference time.

FIG. 147 illustrates a scenario where a device is assumed to be connected over a cellular link to a TSN domain.

FIG. 148 illustrates a shop floor scenario assuming a TSN domain connected to a virtual controller over a cellular link.

FIG. 149 illustrates a scenario where two TSN networks are connected over a cellular link.

FIG. 150 illustrates an example synchronization procedure.

FIG. 151 illustrates another example synchronization procedure.

FIG. 152 is a sequence flow for an example synchronization procedure.

FIG. 153 is a sequence flow for another example synchronization procedure.

FIG. 154 illustrates PTP time transmission using methods disclosed herein.

FIG. 155 illustrates an example method performed by a wireless device.

FIG. 156 is a schematic block diagram of a virtual apparatus in a wireless network.

FIG. 157 illustrates an example method performed by a network node, such as a base station.

FIG. 158 is a schematic block diagram of a virtual apparatus in a wireless network.

FIG. 159 illustrates an example method performed by a wireless device.

FIG. 160 is a schematic block diagram of a virtual apparatus in a wireless network.

FIG. 161 illustrates an example method performed by a network node, such as a base station.

FIG. 162 is a schematic block diagram of a virtual apparatus in a wireless network.

FIG. 163 is a combined flowchart and signaling scheme according to embodiments herein.

FIG. 164 is a block diagram depicting a UE for handling configuration according to embodiments herein.

FIG. 165 is a block diagram depicting a radio network node for handling configuration in a wireless communication network according to embodiments herein.

FIG. 166 is a block diagram of an example wireless device, according to embodiments herein.

FIG. 167 is a block diagram of an example radio network node, according to embodiments herein.

FIG. 168 illustrates a method for assisting enrollment of a device in an Internet of Things (IoT) environment, according to some embodiments.

FIG. 169 illustrates a method for enrolling in an Internet of Things (IoT) environment, according to some embodiments.

FIG. 170 is a schematic drawing illustrating an enrollment process according to some embodiments.

FIG. 171 is a flowchart illustrating example method steps according to some embodiments.

FIG. 172 is a block diagram illustrating an example arrangement according to some embodiments.

FIG. 173 is a block diagram illustrating an example arrangement according to some embodiments.

FIG. 174 is a block diagram illustrating an example network environment according to one or more embodiments.

FIG. 175 is a call flow diagram illustrating example signaling between entities according to one or more embodiments.

FIG. 176 is a flow diagram illustrating an example method implemented by a first device according to one or more embodiments.

FIG. 177 is a flow diagram illustrating an example method implemented by a second device according to one or more embodiments.

FIG. 178 is a block diagram illustrating example hardware according to one or more embodiments.

FIG. 179 is a block diagram illustrating an example first device according to one or more embodiments.

FIG. 180 is a block diagram illustrating an example second device according to one or more embodiments.

FIG. 181 illustrates a flow diagram of one embodiment of a system for querying a federated database in accordance with various aspects as described herein.

FIG. 182 illustrates a flow diagram of another embodiment of a system for querying a federated database in accordance with various aspects as described herein.

FIG. 183 illustrates one embodiment of a network node having a federated database in accordance with various aspects as described herein.

FIG. 184 illustrates another embodiment of a network node having a federated database in accordance with various aspects as described herein.

FIG. 185 and FIG. 186 illustrate one embodiment of a method performed by a network node having a federated database representing one or more autonomous or sub-federated databases that are located in a same or different jurisdiction in accordance with various aspects as described herein.

FIG. 187 illustrates one embodiment of a network node having an autonomous database in accordance with various aspects as described herein.

FIG. 188 illustrates another embodiment of a network node having an autonomous database in accordance with various aspects as described herein.

FIGS. 189 and 190 illustrate embodiments of a method performed by a network node having an autonomous database, in a certain jurisdiction, that is represented by a federated or sub-federated database in accordance with various aspects as described herein.

FIG. 191 illustrates another embodiment of a system for querying a federated database in accordance with various aspects as described herein.

FIG. 192 illustrates another embodiment of a system for querying a federated database in accordance with various aspects as described herein.

FIG. 193 illustrates another embodiment of a system for querying a federated database in accordance with various aspects as described herein.

FIG. 194 illustrates another embodiment of a system for querying a federated database in accordance with various aspects as described herein.

FIG. 195 illustrates one embodiment of a network node in accordance with various aspects as described herein.

FIG. 196 is a schematic block diagram illustrating Ethernet frame handling at UPF from 3GPP TS 29.561.

FIG. 197 is a schematic block diagram illustrating 5G-TSN interworking in an industrial setup.

FIG. 198 is a schematic block diagram illustrating TSN control and data plane with virtual endpoint.

FIG. 199 is a schematic block diagram illustrating VEP deployments as part of the UPF for different PDU session types.

FIG. 200 is a schematic block diagram illustrating VEP(s) as seen by the external TSN network configuration.

FIG. 201 is a flowchart illustrating example method steps according to some embodiments.

FIG. 202 is a flowchart illustrating example method steps according to some embodiments.

FIG. 203 is a combined flowchart and signaling diagram illustrating example method steps and signaling according to some embodiments.

FIG. 204 is a combined flowchart and signaling diagram illustrating example method steps and signaling according to some embodiments.

FIG. 205 is a schematic block diagram illustrating an example apparatus according to some embodiments.

FIG. 206 shows transmission of TSN data streams using redundant paths.

FIG. 207 shows a communication system according to embodiments of the disclosure.

FIG. 208 is a signaling diagram according to embodiments of the disclosure.

FIG. 209 is a schematic diagram showing redundant paths in a wireless network according to embodiments of the disclosure.

FIG. 210 is a schematic diagram showing redundant paths in a wireless network according to further embodiments of the disclosure.

FIG. 211 is a schematic diagram showing redundant paths in a wireless network according to further embodiments of the disclosure.

FIG. 212 is a flow chart of a method in a core network node according to embodiments of the disclosure.

FIG. 213 is a flow chart of a method in a configuring node according to embodiments of the disclosure.

FIG. 214 is a table illustrating a PTP header format.

FIG. 215 is a schematic block diagram illustrating embodiments of a wireless communications network.

FIG. 216 is a flowchart depicting a method performed by a transmitting device according to embodiments herein.

FIG. 217 is a flowchart depicting a method performed by a receiving device according to embodiments herein.

FIG. 218 is a schematic block diagram illustrating embodiments of a multiple time domain support in the 5GS using broadcast according to some embodiments herein.

FIG. 219 is a schematic block diagram illustrating embodiments of a multiple time domain support in the 5GS where only relevant gPTP frames according to some embodiments herein.

FIG. 220 is a schematic block diagram illustrating embodiments of a multiple time domain support in the 5GS according to some embodiments herein.

FIG. 221 is a flowchart depicting a method performed by a transmitting device according to embodiments herein.

FIG. 222 is a flowchart depicting a method performed by a receiving device according to embodiments herein.

FIG. 223 schematically illustrates a telecommunication network connected via an intermediate network to a host computer, according to some embodiments.

FIG. 224 is a generalized block diagram of a host computer communicating via a base station with a user equipment over a partially wireless connection, according to some embodiments.

FIG. 225 , FIG. 226 , FIG. 227 , and FIG. 228 are flowcharts illustrating example methods implemented in a communication system including a host computer, a base station and a user equipment.

Following are detailed descriptions of concepts, system/network architectures, and detailed designs for many aspects of a wireless communications network targeted to address the requirements and use cases for 5G. The terms “requirement,” “need,” or similar language are to be understood as describing a desirable feature or functionality of the system in the sense of an advantageous design of certain embodiments, and not as indicating a necessary or essential element of all embodiments. As such, in the following each requirement and each capability described as required, important, needed, or described with similar language, is to be understood as optional.

Operation Technology Communication System and 5G

A variety of technologies are today used in industrial communication systems. For manufacturing systems in factories, a hierarchical communication structure is used (often referred to as the automation pyramid), as depicted on the left side of FIG. 2 . This design is based on the ISA95/99 model. Industrial equipment is connected in small sub-systems which cover, for example, a production cell. These subsystems are separated by gateways and may use different communication technologies; each subsystem is closely managed to be able to guarantee critical communication performance. On the next higher levels these subsystems are interconnected e.g. for coordination among production cells and supervisory control of the production system. This part related to manufacturing operations is called the operations technology (OT) domain containing the critical communication, where the requirements become typically more demanding on the lower levels. Critical communication is today predominantly based on wired communication technologies like fieldbus or industrial Ethernet. The OT part of the network is securely separated from the IT part of the network containing the enterprise applications and services.

A broader digitalization of the manufacturing system is foreseen to provide increased flexibility and efficiency, by transforming manufacturing to a cyber-physical production system. Such a transition is also referred to as the fourth industrial revolution, or Industry 4.0. It is envisioned that the entire production system can be modeled, monitored, evaluated and steered with a digital twin. To that end, a full connectivity throughout the factory is desired, avoiding isolated connectivity islands on the shop floor level, as shown on the right side of FIG. 2 . The separation of different domains of the network is thereby moved from physical separation (via gateways) to a logical separation. In this transition, IEEE 802.1 Time-Sensitive Networking (TSN) plays a central role, as it allows to provide guaranteed high-performance connectivity services for certain traffic flows on a common Ethernet infrastructure which is shared between critical and non-critical communication. As a fully standardized solution, it allows also convergence of the plurality of proprietary fieldbus technologies existing today to a global standard.

Wireless connectivity can bring great value to a manufacturing system. It can provide cost savings by avoiding extensive cabling, it can support new use cases that cannot be realized with wires (e.g. connecting mobile components). But in particular, it provides significant flexibility in redesigning the shop floor—which is a major trend towards Industry 4.0. Today the use of wireless technology on the shop floor is very limited and focused on non-critical communication provided via various different technologies. For critical communication services there is today no wireless technology that can provide reliable and deterministic low latency.

5G promises to provide reliable deterministic low latency services, while at the same time supporting eMBB and mMTC. (Note that 5G mMTC is based on LTE-M and NB-IoT, which can be embedded into an NR carrier. Eventually an NR-based mMTC mode is expected.) To this end it may play a similar role on the wireless side to what TSN does for wired connectivity. It provides a universal, globally standardized technology that converges all service types and can spread wireless connectivity into much larger fields of the shop floor communication.

TSN has an additional role to play for 5G. Industrial networks are long-living installations and the large majority of factories are already deployed. Introduction of new technology is slow and cumbersome into existing brownfield installations. TSN is expected to trigger a redesign of building practices, which is expected to enter even industrial brownfield networks when feasible. By linking 5G to TSN as the wireless equivalent, TSN provides an opening market opportunity to help in transforming the brownfield market. This motivates a need for the 5G architecture solution to be largely aligned with TSN.

The integration of 5G has to address a number of requirements:

Local content: Production related data may not leave the industry/factory premises i.e. all such data needs to be kept locally for e.g. security and trust reasons. Full control of critical connectivity: critical communication has to be in control of the industrial end user and linked to the operation system where interruption-less operation is managed. Local management: The management solution needs to be easy to integrate with the industry's business and operational processes and include network observability. Local survivability: The connectivity solution is not to be dependent on any external failures, i.e., it shall be self-contained when it comes to survivability. Life Cycle Management (LCM): Several industries require LCM in the range of tens of years. This means that long-term availability of industrial devices and network infrastructure is needed, including ways for device configuration, firmware updates, application software updates, provisioning of identity credentials, installation, provisioning and maintenance in field. Security: The connectivity network shall assure that only authorized traffic is allowed, and with the required level of confidentiality protection (e.g., encryption and/or integrity protection) applied. Functionalities like protection against intrusion (hackers) from the Internet, malware reaching devices and servers, tampering of data etc. shall also be supported. The support of different security zones should be enabled. And the network infrastructure itself needs to be secure and protected from external attacks. Integration with existing solutions: The connectivity solution needs to be integrated into existing wired OT system as well as to other wireless connected devices. One example is transport of Industrial Ethernet frames.

System Architecture

A 5G network integrated into an industrial system, as shown in FIG. 3 , needs the following functions:

5G radio access and core network for 5G connectivity, including radio connectivity, mobility support, service management and QoS including all service categories URLLC, eMBB and mMTC with deterministic performance High availability and redundancy Network identities that enable private network services (i.e. restricting network access and network services to a defined group of devices) Security solutions based on secure credentials Support for positioning and time synchronization Network monitoring and QoS assurance mechanisms A lightweight network management solution A cloud computing infrastructure with deterministic performance and high availability for industrial applications Capability to integrate with existing industrials systems (i.e. connectivity, cloud computing infrastructure, management system) Capability to interwork with external public networks for cases, where service continuity in and out of the factory is required

A 5G system can be deployed in different variants. In cases where local access to dedicated spectrum can be obtained by an industrial user, a standalone local 5G system can be deployed, as depicted in FIG. 3 . Such a standalone 5G network may allow interworking with a public network, e.g. via roaming. Alternatively, federated network slicing can be applied, by creating a logical network slice, which is based on the physical infrastructure of two (or more) networks.

A local 5G system can also be realized as a non-public network service, that is provided by public mobile network operator at the industrial location, as depicted in FIG. 4 . An on-site local deployment of at least parts of the network infrastructure is typically needed. On-site data breakout ensures low latency and allows data privacy policies for information not leaving the site. The core control functionality can be provided from the outside MNO sites, or it can be fully or partially on-site, to support, for example, local survivability. While critical communication services are kept on-site via local breakout, some other functions may also use outside termination of data sessions.

A combination of a standalone local network and a public MNO network can also be used as basis for providing a non-public network service across the two network domains. An industrial user might deploy a local network on-site, which together with the public network infrastructure provides the non-public network service via federated network slicing. For example, the local deployment may be deployed to “harden” the public network, in terms of local coverage, availability, capacity and computing resources.

A local network can also provide neutral-host capabilities, by extending a public network on site in addition to providing a local standalone network. For this purpose, network sharing solutions such as multi-operator core network (MOON) or multi-operator radio access network (MORAN) can be applied. In shared network approaches, a resource management solution is needed that can provide guaranteed resources and performance for the different supported networks (or network slices). A network sharing solution may be well motivated for both local and public network providers. The local provider can provide a free local site for the MNO, while the MNO may provide its spectrum resources for the network. Since the same base stations can support public and private services, some improved coexistence between the local and the public network should be possible. Further, a shared solution may be motivated by different services. For example, a public MNO may provide conventional enterprise services on the industrial site, e.g., telephony, mobile broadband and IT connectivity, while the private standalone local network is used for local industrial OT connectivity.

Network Slicing for Industrial Internet-of-Things (IoT)

Network slicing is considered as one approach to enable or realize Industrial IoT network solutions. Network slicing can provide separate and isolated logical networks on a common shared infrastructure. It can e.g. be used, to

Separate different security zones in a factory, Separate different service categories, e.g., to isolate critical communication from non-critical communication, Provide a non-public IIoT network on a public network infrastructure that is also used for public mobile communication.

Network slicing is a conceptual way of viewing and realizing the provider network. Instead of the prevailing notion of a single and monolithic network serving multiple purposes, technology advancements such as Virtualization and SDN allows us to build logical networks on top of a common and shared infrastructure layer.

These “logical networks”, which may be called “network slices” are established for a particular business purpose or even a particular customer (of the provider). They are end-to-end and complete in the context of the intended business purpose. They are and behave like a network of its own, including all the required capabilities and resources. This extends all the way from the share of the infrastructure resources, through configured network functions to network management or even OSS/BSS capabilities. It encompasses both mobile and fixed network components. One expectation is that different slices are independent and isolated, even if they share common physical resources, and thus provide a separation of concerns. Network slices may be defined to span across multiple physical network infrastructures, which is sometimes referred to a federated network slicing. This can provide even enable an alternative network realization to roaming.

Just as existing networks are built to realize services, so are network slices. They are not services in themselves, but they are built to realize one or several services. As a special case, a service (or instance thereof) maps one-to-one with a network slice, allowing, for example, wholesale type of services. Resources (physical or logical) can be dedicated to a slice, i.e. separate instances, or they could be shared across multiple slices. These resources are not necessarily all produced within the provider, some may in fact be services consumed from other providers, facilitating e.g. aggregation, roaming etc.

Network slices may be defined as comprising a set of resources, as shown in FIG. 5 . This could be physical resources, either a share or profile allocated to a slice or even dedicated physical resources if motivated. Slices may also be defined as comprising logical entities such as configured network functions, management functions, VPNs etc. Resources (physical or logical) can be dedicated to a slice, i.e. separate instances, or they could be shared across multiple slices. These resources are not necessarily all produced within the provider, some may in fact be services consumed from other providers, facilitating e.g. aggregation, roaming etc. Network slicing allows leasing of network capacity with associated service-level agreements (SLAs), for example.

As slices can be created to address a new business requirement or customer and may need to adapt to changes, they require a new type of life cycle management functions, which has the role of creating, changing (e.g., upgrading) or removing them. Network slicing allows for different network architectures which are optimized for the specific use case that the slice is being used for. This optimization for different network slices may include both optimizations in the functional domain and in the geographical deployment of different functionality in the network. This can be seen in FIG. 6 , which illustrates an example of four different slices through the network. They also expect the service provider to support them with inclusion of industry specific services and or applications from other third parties or service providers in a cost efficient and timely manner.

The definition for network slicing is twofold. For the general definition the one from GSMA is used: “From a mobile operator's point of view, a network slice is an independent end-to-end logical network that runs on a shared physical infrastructure, capable of providing a negotiated service quality”. Besides this general definition, several implementations realizing the above exist and are often meant when “network slicing” is mentioned. The most prominent one comes from the 5G Core specification, “System Architecture for the 5G System (5GS), Stage 2,” 3GPP TS 23.501, v. 15.4.0 (December 2108): “Network Slice: A logical network that provides specific network capabilities and network characteristics [ . . . ] A Network Slice is defined within a PLMN and shall include: the Core Network Control Plane and User Plane Network Functions [ . . . ]”. Methods at least partly realizing above definition are not limited to 5G but also available in 4G networks.

With these definitions the basic network slice can be explained according to FIG. 7 :

There is a shared physical infrastructure (see (1) in FIG. 7 ). One or more independent end-to-end logical networks (see (2) in FIG. 7 ) are defined, which comprise:

a. A core network control plane, b. User plane network functions,

These logical networks can support negotiated service quality or specified service capabilities, or, in other words, a service level agreement (SLA) about the network slice capabilities (see (3) in FIG. 7 ).

Once network slices have been defined, a first question is how a data traffic flow is assigned to or routed through the corresponding network slice. In many cases, a single device is making only use of a single slice, so the allocation can be made by assigning each UE to a specific network slice. However, in some cases a device may serve traffic for multiple slices.

A baseline in mobile networks for service treatments to provide specific service performance and QoS are dedicated bearers; they are often a solution to fulfill the requirements of specific use cases or service. In the radio access network (RAN), the dedicated bearers map to radio bearers that can be used by the scheduler to deliver bearer-specific QoS. Specific resources may be reserved for certain dedicated bearers. At the network edges, bearers can be identified and treated individually based on filter on the packet headers, like the 5-tuple source IP address, destination IP address, source port number, destination port number, and protocol (UDP or TCP).

FIG. 8 shows four available 4G methods to slice a network. For the first one, RAN sharing is applied, allowing eNBs to announce multiple PLMN IDs. To utilize this approach, the RAN and Core need to support that feature, to assure the PLMN IDs are announced and traffic is appropriately routed to/from the right Core network. The UE selects the PLMN based on usual procedures of network selection, including having preferred (home) networks. A UE can only be served by one PLMN (except for the case of multi-SIM UEs). Currently, this solution is supported by every UE and by at least some network-side systems.

The second solution relies on Access Point Names (APNs) configured in the UE. In this case, one PLMN ID is announced by the RAN but user plane traffic is routed to the right Core network based on the APN. A UE can even have multiple APNs configured resulting in multiple IP addresses (multi-homing) when PDN sessions are established. Assuring the right source IP address is used when transmitting in the uplink is not straight forward. Setting more than one APN in the same UE for internet applications might not be supported by every device. This solution requires no changes to RAN but must be supported in the Core networks.

3GPP had a study item named DECOR, now described in standards document as dedicated core network (DCN), which allows for the selection of a slice based on configuration in the network, rather than a preferred PLMN ID or APN settings, as done for the previous solutions. The feature must be supported in RAN and Core, and information from the home subscriber server (HSS) is used to determine the “UE usage type” and, by this, attaching it to the right slice. There is no UE impact in this solution.

A concept known as eDECOR further enhances this by allowing the UE to submit a DCN-ID to select the slice. To utilize this approach, the RAN, Core, and UE need to support the feature.

Both DECOR and eDECOR, only allow one slice per UE but assure that UEs of different types are served by different slices. Within each dedicated core, multiple dedicated bearers and APNs can be used.

For Release 15 and beyond, 5G Slicing extends this feature to a theoretically unlimited number of slices, but implementation and resource dependent constraints in UE, RAN and Core will likely apply. As for 4G, several sub-options to realize slicing in 5G exist, but they will not be further distinguished in this document.

Once traffic has been assigned to the corresponding slice, the next question is on how service performance can be provided. In many Industrial IoT use cases, guaranteed service performance for prioritized traffic is required. In a normal (i.e., unsliced) 5G network, or within a single network slice, different traffic flows can be separated according to traffic flow separation, as shown in FIG. 9 , which shows how QoS is applied in the 5G system. Dedicated resource allocations can be provided for critical traffic. Admission control is used to ensure that the number of admitted prioritized traffic flows with guaranteed transmission resources (i.e., guaranteed bitrate) do not exceed the available resources, with sufficient margin for resource variations.

For resource partitioning between slices, the reservation of resources in the physical infrastructure is not per individual traffic flow, but instead based on the sum of critical traffic flows within a slice. This total requirement needs to be defined in the network slice SLA. The resource partitioning does not need to be static. Better efficiency can be achieved if unused resources of one slice can be used by another slice. This can be seen in FIG. 10 , which illustrates resource partitioning between example slices A and B. What is required is that each network slice can get access to the guaranteed service flows at any point in time (or at least to the availability level defined in the SLA).

Industrial Applications

Following is a discussion of several applications and activities that connect to industry technologies. This discussion includes a discussion of cloud robotics, which is a new technology that provides many additional benefits, compared to previous technologies.

In Chapter 5.3.2 of “Study on Communication for Automation in Vertical Domains,” 3GPP TR 22.804, v. 16.2.0 (December 2018), motion control is introduced as a use case for factories of the future. Motion control is essential for any automation application and, for example, is also fundamental for industrial robots. A robot's motion or a printing machine's functionality is basically just a coordinated motion control of multiple actuators.

Motion control refers to the task of driving an actuator (or a group of actuators) in a way (and ensure that it is doing so) an application requires. Electronic motors are the most common actuators in industries. There are diverse ways to classify electrical motors (e.g. AC-DC (brushed/brush-less), stepper-servo-hybrid stepper). Nevertheless, the motion control principles are similar for each motor class. Communication technologies are used to coordinate and synchronize multiple actuators and for higher layer control. Motion control applications with requirements on accuracy or precision are always implemented as a closed-loop control.

There is a common logical split in motion control systems:

Physical actuator (aka motor) & encoder (i.e. one or more sensors for speed, position etc.) Driver (also called inverter) Motion controller Programmable Logic Controller (PLC)

This logical split of motion control functions is illustrated in FIG. 11 .

Typical communication patterns in the motion control architecture (numbers as in FIG. 11 ):

1) A PLC communicates higher level commands to the motion controller—this imposes less stringent communication requirements 2) The motion controller generates so called set points (might be speed, torque etc.) for the driver, by using, e.g.:

a. Pulse-width modulation, which is no communication technology b. Protocols like EtherCat or Profinet IRT or similar, with support of very low cycle times, usually below 1 ms.

3) Currents fed from the driver to the motor—energy supply to the motors based on the set points, no communication technology. 4) Encoder (sensor) feedback to driver and/or the motion controller; feedback depends on type of motor and encoder. Feedback can be analogue or based on for example EtherCat or Profinet IRT as in 2) with same requirements (cyclic closed loop set point transmission and feedback).

A single motion controller might control multiple actuators, if for example several motors are used in the same machine. In the technical report mentioned above, the requirements for motion control applications (there the closed-loop is addressed: motion controller-driver-encoder) are listed—these requirements are reproduced in Table 1, below.

In 3GPP TR 22.804, it is further mentioned that two consecutive packet losses are not acceptable and a very high synchronicity between all devices involved (with a jitter below 1 usec) is required. The latter is mandatory to be able to take samples from distributed encoders and also apply new set points from the motion controller to drivers at common sampling points. This is referred to as isochronous communication, which means that applications (so the motion control program as well an all actuators and encoders) are in sync to the communication cycle timings given by the communication technology (for example, of Profinet). This also ensures minimal and deterministic latencies using timed channel access.

Several vendors of motion control equipment, such as motion control manufacturer Lenze, also combine functionalities into single physical entities. On an upper “control level,” they use a combined PLC+motion controller (Logic & Motion), next to a human-machine interface. This controller takes input from IO-devices (3) and feeds its set points to e.g. the servo-inverter (2) over EtherCat (“Field Level”).

Another trend is to integrate encoder and/or driver and/or motion controller into the motor. This is also sometimes referred to as an integrated motor or a smart motor.

Furthermore, it is possible that multiple motion controllers are used for the same application; each motion controller controls a subset of drivers. Coordinated motor movement requires communication between separated motion controllers. In 3GPP TR 22.804, this is referred to as ‘Controller-to-Controller’ (C2C) communication. Cycle times between 4 to 10 ms are assumed. The requirements for synchronization are equally strict, with a jitter below 1 usec also on the C2C level. Payload sizes may be up to 1 kB.

For safety reasons there may be an additional functional safety control deployed in wireless motion control applications. Functional safety is implemented as an additional closed-loop, next to the one used for motion control itself. This is done through additional hardware or integrated safety functions in the motion control components. Communication protocols like ProfiSafe are used. One safety restriction is, for example Safe Torque Off (STO) (from IEC 61800). STO defines, that in case any error/safety issue is detected by the PLC or an additional safety PLC, power delivery to the motor need to be stopped. An STO can for example be triggered by pressing an emergency button. In 3GPP TR 22.804, it is explained that for functional safety, a strictly cyclic data communication service is required between both ends. If connectivity is disturbed, an emergency stop is triggered, even if no real safety event has occurred. There are different requirements (4 ms to 12 ms cycle time, packet size 40-250 byte, tolerable jitter in transmission according to 3GPP TR 22.804) for different use cases. Safety functions might be implemented in different components of the motion control architecture.

To sum up, there are four different types of communication for motion control:

1) Lowest level closed-loop motion control (motion controller-driver-encoder) 2) Controller-to-controller communication 3) Functional safety communication 4) PLC to motion controller communication

The requirements for the communication system regarding latency are decreasing from 1 to 4. Whether it makes sense to establish links 1.-4. over a wireless communication technology is application dependent. In most cases it might be most relevant to establish wireless links for 2), 3) and 4), but perhaps not for 1).

Cloud Robotics

In industrial robotics research and robotics in general, cloud robotics is a major topic. It describes how different cloud technologies can be used to provide additional benefits for various robotics tasks and thereby improving the flexibility and the capabilities of the whole system. Several studies have already shown the benefits of connecting robots to a cloud:

Usage of more powerful computing resources in the cloud (e.g., for artificial intelligence, AI, tasks). Use of almost unlimited data for analytics, decision making and learning (including digital shadows and real-time simulations). New types of use-cases are enabled (e.g., cooperative control in cloud). Lower cost per robot, as functionalities are offloaded to a central cloud. A possibility to perform a failover in case one robot physically breaks from an up-to-date backup in the cloud. Reliability of functions can be improved by running multiple instances as hot standby in the cloud and the operation can immediately be taken over from faulty primary function without interruption. Makes the operation and maintenance easier (software updates, configuration change, monitoring, etc.). Saving energy, particularly for mobile, battery driven, robots, by offloading CPU energy consumption to the cloud.

High flexibility is indeed a key requirement for Industry 4.0. It is needed to realize cost effective and customized production by supporting fast reconfiguration of production lines, as well as, easy application development. Typical industrial applications are time-sensitive and require highly reliable communication end-to-end. Therefore, 5G URLLC and edge cloud are necessary technologies to address the requirements. Although some cloud robotics applications don't require real-time communication, still some do heavily, especially if the cloud's processing is relevant for the immediate motion of the robot. In the following, some of the major challenges with industrial applications that can be addressed with cloud robotics are listed:

Fast closed loop control (1-10 ms) between controller and device. Wireless link between controller and device. Real-time industrial application in cloud execution environment (e.g., servo controller). Industry-grade reliability (e.g., the same as with cable based ProfiNet). Flexible production lines (easy rearrangement and reprogramming, low delay software updates, reconfiguration, i.e., FaaS capability). Cooperative control and modular architecture. Adaptive algorithms (e.g., for human-robot collaboration the control has to be adaptive to changing dynamic environments and need learning and cognitive capabilities). Shared data for different control applications.

In the following, briefly described are some cloud robotics scenarios that involve (real or emulated) 4G/5G connections and cloud technologies for industrial robotics applications.

One application of cloud robotics is to replace the hardware Programmable Logic Controller (PLC) in a robot with a software version (soft-PLC) and run that in a virtualized environment/cloud on commodity HW components. A concept study of this involved a real robot cell with two large robot arms, a conveyor belt and some other industry devices. For the communication, ProfiNet was used.

One issue is what level of robot control can be shifted to the cloud over LTE. This is illustrated in FIG. 12 . The high-level control that is typically done by the PLC is not very delay critical, i.e., it has a latency requirement of several tens of milliseconds (e.g., −30 ms), depending on the configuration. However, the whole communication is very sensitive to delay variations (jitter) and packet losses. For instance, in the case of periodic traffic with 8 ms frequency, three consecutive packet losses or 3*8 (24) ms jitter can make the whole robot cell stop. Those requirements are straightforward to fulfill when using dedicated hardware components over a cable-based solution but can be challenging using virtualized execution over wireless technologies.

From the cloud platform perspective, one of the main challenges that virtualized control brings in is the execution of real-time applications. An application might use a soft-PLC that uses Windows 7 as a base OS, next to a real-time OS that is responsible for executing the PLC code. Both run in parallel and communicate via inter-process communication (IPC). The control logic implementation is always executed by the RTOS and Windows is often used as a user interface. The RTOS typically has some specific requirements to ensure the necessary performance such as precise timers and specific network interface cards. A virtualized environment that can host the software PLC platform and execute the same control logic as the one that was running on the hardware PLC can be created.

In a real factory environment, placing the PLC-level of control logic from dedicated HW into an edge-cloud platform is feasible, and works adequately even over LTE. However, if we investigate applications such as trajectory planning, inverse kinematics and control loops that accurately steer the speeds, accelerations or positions of the actuators, significantly lower latency in the range of 1-5 ms is required. To support those applications, the ultra-reliable and low-latency service of 5G is essential, as shown in FIG. 12 .

One motivation behind moving motion control of robots to the cloud is again to increase flexibility. For instance, much easier to rearrange production lines with cloud-controlled devices, since only the devices need to be moved (no controller boxes), easier to manage, reprogram, do failover, or software updates in such an environment. However, some functionalities should remain inside/close to the robot (using cables), for instance, some safety mechanisms in case of connectivity problems. The requirements on the network are lowered if the robot could also perform its task without connectivity. In case of temporarily connectivity loss or reduced performance (due to for example extended robot mobility on the shop floor) the robot could reduce its working speed or activate other mechanisms to ensure safety or process targets while being independent from the network. The robot controller as an additional entity next to the robot itself should be removed from the shop-floor. An architecture for this type of deployment is shown in FIG. 13 .

Another approach could be that the motion control is done inside the robot autonomously and the connectivity is used only to enable new use-cases such as cooperative robot control. For cooperative control, one control entity may still need quick access to the actual state of the other control processes. This option is valid in some scenarios when, for example, the motion control has 5 ms control loop inside the robot, but it needs to coordinate with another instance only in every ˜100 ms.

A robot controller including trajectory planning and execution, has been implemented, with the performance of a robot arm control application from a local cloud over a modelled wireless channel being evaluated. The application under evaluation included the closed-loop control of a industrial robot arm, where control was connected to the robot arm through a modelled 5G link.

The effect of the link delay on the performance of the robot arm movement quality can be measured by specific key performance indicators (KPIs). The industrial robot arm has an externally accessible velocity control interface that accepts velocity commands for each joint (servo) and publishes joint state information with 8 ms update time. KPIs may be response time and precision of trajectory execution, i.e., spatial and temporal deviations from the planed trajectory. Measurements have shown that network delays below 4 ms have no significant performance impact in this application. This is because (1) the internal operation of the robot ends in about 2 ms standard deviation in response time due to the internal sampling used in the robot, and (2) the ticks of the robot and the controller are unsynchronized. The impact of network delays below 4 ms is masked by the background “noise” of the measurement setup.

Several other conclusions can be reached:

Reaction on external events: low network delay is desired, because the network delay between robot and controller directly increases the reaction time. Realtime trajectory refinement (i.e., accurate positioning of the end of the robot arm): Deadline on trajectory execution time leads to requirement on maximum tolerable network delay. In general, higher network delay makes the refinement time longer and, in this way, increases the total trajectory execution time. Trajectory Accuracy: Some tasks require accurate movement along the path such as welding, and not only at the final position. Another example is the collaboration of more robot arms where the precise and synchronized movements are crucial. For these tasks, low network delay is desired if external information shall be respected in the trajectory planning.

The internal mechanisms of a robot arm can also put requirements on the network delay. In general, a system with low update time requires lower network delay. For instance, the control of a robot arm with 20 ms update time, tolerates higher network delay than a more precise and faster robot arm with 1 ms update time. In addition to this, providing ultra-low latency connection for a system with relatively high update time has limited performance advantage.

Performance requirements of trajectory execution can also put requirements on the network delay. Faster robot movements require lower network delay for accurate movement. On the other hand, if only a higher latency connection is available then using lower robot speed can compensate increased network delay to some extent. Performance optimization can also give guidelines for the required network delay. Choosing a proper required accuracy can improve the execution time. For example, if less accurate movement is enough, then relaxed accuracy can shorten the refinement time.

New robotic concepts and applications include massively collaborative robot control, as well as the use of digital twins in cyber-physical production systems. These are briefly discussed in the below sections.

Hexapod

When introducing higher collaboration and adaptation capabilities into industrial applications such as robot arms and robot cell control, collaboration of a massive number of servos may be required, making the use case even more challenging. The hexapod robot is a useful application for evaluating a wide spectrum of challenges arising in an Industry 4.0 robot cell, e.g., servo control, collaboration, etc. FIG. 14 illustrates a hexapod robot, which may be viewed as a collaborative, robot-vendor-agnostic system, coupled with a 5G slice for cloud-based control.

The hexapod can be considered as six 3-degree of freedom robotic arms connected via a base link. For evaluating 5G requirements, the servos at the 18 joints may be controlled separately from a computer residing a wireless network hop away from the hexapod. This way the hexapod proves to be an appropriate choice for visualizing the effect of synchronized collaboration. Well-synchronized collaboration should result in a stable center position, while any glitch in the system results in jiggling of the platform. Results of an evaluation of wireless control of the hexapod have been reported in Geza Szabo, Sandor Racz, Norbert Reider, Jozsef Peto, “QoC-aware Remote Control of a Hexapod Platform,” ACM Sigcomm, Budapest, 2018.

Digital Twin

The Digital Twin (DT) concept is useful for analyzing the effects of network on the control of a real robot, where its DT runs in a complex robot cell executing agile robot tasks. A realizable DT may be implemented in the Gazebo simulation environment and evaluated against a fully simulated scenario solving the Agile Robotics for Industrial Automation Competition (ARIAC). This evaluation deals with issues of the different command frequencies, control loops and handling of dynamics of the real and simulated robot. An evaluation of the architecture in a hardware agnostic Gazebo plugin shows that the simulation of the network controlling a simulated robot can be used in low-delay scenarios. In high-delay scenarios the simulated latency provides approximately ˜10% more room regarding the delay size till the complete failure occurs in the robot cell. These results are reported in Ben Kehoe, Sachin Patil, Pieter Abbeel, Ken Goldberg, “Survey of Research on Cloud Robotics and Automation,” IEEE Transactions on Automation Science and Engineering (T-ASE): Special Issue on Cloud Robotics and Automation. Vol. 12, no. 2. April 2015.

Positioning

Positioning is recognized as an important functionality in industry and manufacturing scenarios, with use cases such as personnel tracking (e.g., in mines), safety (e.g., when working close to forklifts), locating tools in manufacturing/assembly floors, supply chain optimization, operation of automatic guided vehicles, etc. Most use-cases require only relative positioning, e.g., where all positions are defined relative to a common reference point in a factory hall.

The required positioning accuracy, as well as the environment and radio conditions where positioning is to be performed, vary significantly between different use cases. However, most manufacturing use cases are indoor, like for example a factory hall or the tunnels in a mine. This implies that global navigation satellite system (GNSS) based solutions are difficult to use because of the very low signal strength levels received indoors from satellite transmissions, resulting in no or bad coverage.

The limitations of GNSS systems indoors have opened for cellular based positioning solutions. Commonly used positioning solutions in industries and factory floors today are based on Wi-Fi, radio-frequency identification (RFID), Bluetooth low energy (BLE), ultra-wide band (UWB) and LTE. Narrowband (NB)-IoT and CAT-M are 3GPP LTE technologies to address low complexity, low power, low cost devices, and are therefore the only realistic 3GPP positioning solution for use cases where the asset to be positioned doesn't already contain a 3GPP modem for communication needs. Radio solutions, such as RADAR, and non-radio solutions, like LIDAR and computer vision systems, are also important especially when positioning with high (sub-meter) accuracy is required.

Multipath propagation is often a critical error source for positioning. In industry halls, the delay spread of the paths is typically relatively short, but these are still critical given the requirements for accurate positioning in such environments. Most positioning algorithms work under the assumption of availability of line-of-sight measurements and there are no straightforward ways to distinguish between line-of-sight (LoS) and non-line-of-sight (nLoS). If an nLoS path is mistakenly used instead of a LoS path for positioning, both the time-of-flight and the angle of arrival might be misleading. The time-of-flight of the nLoS path will be an upper bound of the time-of-flight of the LoS path, while the angle of arrival can be completely wrong. Therefore, nLoS paths may greatly degrade the performance of positioning algorithms. Future industrial positioning schemes need to tackle this issue satisfactorily.

Another obstacle to precise positioning is network synchronization errors. Practical network synchronization algorithms can imply network synchronization errors of up to 360 ns, which corresponds to ±110 m of positioning error. A promising alternative to improve the positioning accuracy is radio-interface-based-monitoring (RIBM). This solution is based on base station timing measurements of positioning reference signals from neighboring base stations and estimates the synchronization offset between base stations so that “virtual synchronization” with much better accuracy can be provided. Alternatively, positioning techniques that don't require network synchronization, e.g. techniques based on round-trip time and/or angle of arrival measurements, can be considered. Note that any estimates of positioning accuracy stated herein assume that good network synchronization has been achieved, e.g. using RIBM.

Positioning accuracy, especially between time instants where measurements are performed, can be significantly improved by considering the trajectory of movement. Furthermore, inertial measurement units (IMUs) are becoming increasingly widely adopted in terminals as a means of updating position estimates. They use accelerometers and gyroscopes (and sometimes also magnetometers) to track movement of the terminal.

Deployment Aspects

To reduce cost and simplify deployment, solutions where one system is used both for communication and positioning are preferred. This is especially important in environments where deployment of a separate positioning system is difficult and costly, for example in mines where the installation cost of each node is often high. However, if the communication deployment, which may comprise one or a few micro base stations, for example, doesn't provide sufficiently good positioning accuracy, a separate or complimentary positioning system added on top of the communication system may be the best solution, since high-precision positioning typically requires a much denser deployment than communication.

The positioning accuracy that can be achieved depends to great extent on how dense the deployment is and the characteristics of the radio environment. Densification of the communication network may thus be a means to achieve improved positioning accuracy. A dense deployment is especially important in environments with severe multipath propagation, especially if the multipath propagation is dynamically changing, since there may otherwise not be sufficiently many LoS paths available to estimate the position. A dense deployment may also be necessary to ensure that hidden objects with high signal attenuation can be localized.

The density of the network is a key aspect for providing good enough positioning accuracy in manufacturing scenarios. Another deployment aspect to consider is how simple it is to install the anchor nodes. Installation may for example include manually providing the exact position of the anchor, which may be difficult, time consuming and error prone. To avoid this, a simultaneous localization and mapping (SLAM) algorithm may be used to estimate the position of each anchor in the initialization phase.

To make dense deployments cost effective, the cost of each anchor must be kept low. However, the cost of each anchor/base station for technologies providing both communication and positioning is naturally higher than for technologies only providing positioning, e.g. RFID and UWB. One way to reduce the cost involved in densification of the communication network to achieve high-precision positioning may be to develop simple anchors, using the same technology as the costlier base stations, with reduced capability that only provides positioning. For example, only one or a few highly capable NR base stations mounted in the ceiling of a factory hall may be sufficient to provide communication coverage. Less capable NR positioning nodes/anchors can then be used for densification to achieve positioning with high accuracy.

Another way to reduce the need for a very dense deployment may be to combine the advanced beamforming capability of NR with reflectors. In this way, every pair of transmit beam and reflector may act as a virtual anchor, thereby achieving the benefits of a very dense deployment with only a few NR base stations. One challenge with such a solution is stability, since the reflectors should be stationary or at least only slowly changing to ensure a stable positioning accuracy.

Spectrum Aspects

It is well-known that positioning accuracy improves with increased bandwidth. Furthermore, higher signal bandwidth enables greater resolution of the LoS leading edge from the nLoS dominant received signal and it is therefore easier to accurately detect the LoS paths. On the other hand, using higher carrier frequency may reduce the detectability of the LoS path due to increased signal attenuation.

Frequency bands for local usage are currently being defined, e.g., the 3.7-3.8 GHz band in Germany and Sweden. Nation-wide spectrum and/or unlicensed spectrum may be used for industries as well. 100 MHz bandwidth should be sufficient to achieve sub-meter positioning accuracy. To improve performance further, different spectrum chunks may be combined.

Accuracy Requirements

Positioning accuracy requirements range from millimeter to several tenths-of-meters level. For example, drilling and blasting in mines as well as automated manufacturing (alignment, assembly) may benefit from millimeter-to-centimeter accuracy. Other examples where centimeter-to-decimeter accuracy is desired include locating tools in manufacturing/assembly floors and tracking of automated guided vehicles. Decimeter-to-meter accuracy is required for some safety solutions, for example tracking of personnel and real-time warnings for personnel working close to a forklift, but also when considering for example supply chain optimization and asset tracking (e.g. tools, machines).

3GPP have documented positioning requirements for 5G positioning services in TS22.104, Section 5.7, “Positioning performance requirements.” Table 2 excerpts some of these requirements. According to 3GPP, depending on the use case, the 5G system should support the use of 3GPP and non-3GPP technologies to achieve higher-accuracy positioning.

Overview of Positioning Technologies

Below, an overview of positioning technologies that may be useful for manufacturing is given, with some focus on how they can be applied in a manufacturing scenario. The focus lies on 3GPP techniques, but a number of other techniques are considered here as well. Positioning using RFID and BLE beacons has not been included in this overview, but it will be appreciated that many of the same principles apply there, and that the various technologies described herein may be combined with these and other positioning technologies.

Since Release 9, LTE supports observed time-difference of arrival (OTDOA) positioning, which is based on reference-signal time difference (RSTD) measurements that are described in 3GPP TS 36.305. The UE receives positioning reference signals (PRSs) from neighboring cells, estimates time of arrival (TOA) for each cell using RSTD measurements, and reports back the TOA with respect to a reference cell. Afterwards, the evolved serving mobile location centre (E-SMLC) estimates the position of the UE based on known eNB positions. The time difference of arrival (TDOA) is used with respect to a reference cell instead of TOA, because this removes requirement that the UE be time-synchronized, although the network needs to be synchronized. In principle, a minimum of 3 cells are needed for 2D positioning and a minimum of 4 cells are needed for 3D positioning.

FIG. 15 illustrates how the UE position can be estimated from 3 eNBs, in accordance with the principles of OTDOA, and is a conceptual plot of 2D TDOA-based positioning, assuming perfect TDOA measurements. Each TDOA (TOA of reference eNB minus TOA of eNB) translates into a difference in distance (e.g., in meters) when multiplied by the speed of light. Each TDOA returns a hyperbola on the 2D plane of possible UE positions. The intersection of such hyperbolas is then the UE position. In practice, the position is estimated by the E-SMLC using Gauss-Newton search or similar numerical algorithms.

In LTE, RSTD can be estimated based on cell-specific signals or based on optionally defined PRSs. However, the TDOA estimation procedure typically uses PRSs because other cell-specific reference signals cannot guarantee high enough probability of detection of neighbouring cells at low (sub −6 dB) signal to interference and noise ratio (SINR). The PRSs are defined from Gold sequences initialized by time variables (slot number within a frame and OFDM symbol number within a slot) and PRS ID, and allocated in a diagonal pattern that is shifted in subcarrier. Essentially, three main factors contribute to high PRS detectability:

The Gold sequences guarantee low cross-correlation properties. There are 2 PRS resource elements per resource block and OFDM symbol with a distance (reuse factor) of 6 subcarriers. The specific location of each PRS diagonal is determined by PRS ID mod 6. The subcarriers not used for PRS are empty in order to create LIS (low interference subframes). The PRS can be muted on some transmission occasions to increase the SINR when receiving PRS from distant cells.

The RSTD is drawn from the power delay profile (PDP) generated by cross-correlating the received downlink baseband signal with the PRS. The challenge here is to detect the earliest peak in the PDP which is not a noise peak and then take the peak delays in terms of multiples of samples. A major source of TOA error is nLoS conditions where the LoS path is not detected due to blocking or shadowing.

The positioning accuracy that can be achieved with LTE OTDOA in practical deployments is in the order of 50-100 m for Release 9. In LTE Release 14, the report resolution to the E-SMLC was changed from Ts to Ts/2, where Ts is the basic time unit in LTE (32.55 ns), to improve the relative distance resolution from 9.8 m to 4.9 m. It is however still unclear what accuracy can be achieved in practice. In addition, OTDOA requires network synchronization and any synchronization error reduces the positioning accuracy that can be achieved. For LTE, there are mobile broad band (MBB) UE chipsets available that cover most of the positioning methods standardized up to Release 14.

FIG. 16 shows OTDOA positioning results for the 3GPP Indoor Open Office (IOO) scenario, using different bandwidths of 100 MHz (30 kHz SCS, 275 PRBs), 50 MHz (15 kHz SCS, 275 PRBs), 10 MHz (15 kHz SCS, 50 PRBs), 5 MHz (15 kHz, 25 PRBs). The plot is based on the use of the already existing tracking reference signal (TRS) for positioning as baseline in NR.

The scenario assumes 6 gNBs (a total of 12 gNBs) separated by 20 meters. (“gNB” is the 3GPP terminology for an NR base station.) The results show that the positioning accuracy is improved significantly when increasing the bandwidth from 5 MHz to 10 MHz and as much when further increasing the bandwidth to 50 MHz. However, it can also be seen that the 100 MHz and 50 MHz results do not differ much, with around 8 meters accuracy at the 80% percentile. The 100 MHz case can be further improved by using a more advanced peak search algorithm for time-of-arrival (TOA) estimation. In the current simulations, the earliest peak in the PDP that is at least half as high as the highest peak is taken as the LOS peak. If the signal-to-noise ratio (SNR) is increased, the probability of detecting a peak above the noise floor can be improved. Furthermore, errors larger than the inter-site distance (ISD) of 20 m can in practice be compensated by combining with a simple cell-ID (CID) estimate if the OTDOA becomes unreasonable, which is not done here.

Narrowband (NB)-IoT and CAT-M are 3GPP LTE technologies to address low complexity, low power and low-cost devices. The availability of such low-cost devices makes this the only realistic 3GPP solution for use cases where the asset to be positioned doesn't already contain a 3GPP modem for communication needs. However, the positioning accuracy is significantly worse when using IoT devices, mainly because of the narrow bandwidth used. A simulation study demonstrated 100 m positioning error at the 70% percentile for NB-IoT in an indoor deployment, while LTE using 50 PRBs for positioning gave ˜23 m at the 70% percentile in the same scenario.

The narrow bandwidth of NB-IoT devices is compensated partly by enabling longer PRS occasions in time. However, the correlation properties are poor since the PRS is repeated every frame (10 ms). NB-IoT devices also have lower sampling rates to reduce power consumption, which reduces the accuracy of RSTD measurements.

Chipsets for LTE IoT-devices are not as readily available as for LTE MBB. However, development is ongoing, and the availability is slowly improving.

IoT positioning for NR is not defined as of December 2018, but one enabler for better IoT positioning accuracy is to improve the time-correlation properties of the PRS, e.g. by increasing the PRS repetition interval. Carrier-aggregation for NB-IoT has also been discussed and the increased bandwidth may in this case be another enabler for improved IoT positioning. Another alternative may be to modify the phase of the eNB PRS, thereby ensuring that the NB-IoT devices can sample at low rates and still detect the phases of the PRSs.

LTE Enhanced Cell-ID Positioning

Enhanced Cell ID, or E-CID, was introduced in LTE Release 9. The UE reports to the network the serving cell ID, the timing advance and the IDs, estimated timing and power of the detected neighbor cells. The eNB may report extra information to the positioning server, like the angle of arrival, cell portion, round-trip time, etc. The positioning server estimates the UE position based on this information and its knowledge of the cell's locations.

The accuracy of E-CID depends mainly on the deployment density. For outdoor deployments, the accuracy of E-CID may be in the order of 100 m, for urban environments with ISD of less than a few hundred meters, or in the order of 3000 m, for rural environments with ISD up to several kilometers. The accuracy of E-CID for manufacturing-like environments has not been studied, but the accuracy is expected to be in the order of the ISD, since the environment contains many multipaths and for example angle-of-arrival data may be misleading, due to reflections. On the other hand, even in such a challenging scenario, RF fingerprinting should be able to give an accuracy of a few meters if the radio propagation is stable and a calibration/training phase is feasible.

NR Positioning Features

As of December 2018, there is no defined concept for NR positioning. One can envision NR features which enable improved positioning accuracy over LTE OTDOA. Some of these features come with new challenges as well:

Better ranging and angle-of-arrival/departure (AoA/AoD) estimates in beam-based systems. Higher carrier frequencies are supported in NR, meaning that signals are more susceptible to blocking/shadowing. This may in part be handled by beamforming. Furthermore, higher carrier frequencies typically come with wider bandwidths, enabling better RSTD resolution. Denser deployments in terms of smaller inter-site distance and cell radius. This, combined with beamforming using beam IDs for specific beams, require more sophisticated Gold sequence initializations to preserve code orthogonality for all possible beam/cell-ID combinations. Better time alignments are expected in NR, thereby reducing the time synchronization error. The basic time unit is reduced in NR compared to LTE. The maximum number of PRBs is 275 (compared to 110 in LTE), requiring an FFT length of 4096 which is double to that of LTE. Furthermore, the sub-carrier spacing ranges from 15 kHz to 240 kHz. Essentially, this implies shorter sampling intervals than in LTE, improving the TOA positioning resolution.

Expectations are that the solutions standardized for NR Rel-16 will provide the tools needed to achieve sub-meter accuracy. Link-level simulations showing the technology potential of NR indicate that sub-decimeter accuracy could be possible in theory. The Release 16 NR positioning 3GPP study item started in October 2018.

Positioning Using the Radio Dot System

Ericsson's Radio Dot System (RDS) is well suited for communication in indoor industry and manufacturing scenarios. However, with RDS products available as of December 2018, only cell-ID based positioning is available, since it is not possible to distinguish DOTs connected to the same IRU from each other. Furthermore, an RDS is often deployed with large cells since up to 8 DOTs can be connected to the same IRU. With the digital 5G DOT, up to 16 DOTs can be connected to the same IRU.

Improvement of the positioning accuracy, making per DOT positioning possible, has been proposed. UE position is calculated using an uplink time difference of arrival (UTDOA) algorithm combined with DOT level power. Simulations have shown that positioning errors of less than 1 meter can be achieved with good SNR and good DOT geometry layout. However, positioning errors in the order of 1-5 meters are likely, when taking various error sources, like the accuracy of DOT positions and DOT cable length delays, into account.

For typical manufacturing scenarios with severe multipaths, it appears that the radio dot system (RDS) is the most suitable solution for providing combined communication and positioning, due to dense deployment and low cost of the nodes.

Wi-Fi Positioning

Wi-Fi is already commonly deployed in industries and is therefore often used for positioning as well. One commonly deployed Wi-Fi solution is the ARUBA solution, which can achieve, with access point received signal strength indicator (RSSI) alone, around 5-10 m accuracy, depending on shadowing and antenna patterns. To achieve better positioning accuracy, the ARUBA solution can be combined with Bluetooth low energy (BLE) battery powered ARUMBA beacons. With this specialized positioning solution, very good accuracy of <3 m is likely to be achieved, and even accuracy of <1 m can be possible when the device to be positioned is located close to a beacon.

The leading industrial Wi-Fi positioning solution achieves 1 m to 3 m average accuracy in office environments. This positioning solution includes an additional WiFi radio, with a specialized antenna array, that is included in the same unit as the WiFi radio used for communication. The positions are estimated using a combination of RSSI and angle-of-arrival (AoA) measurements.

A difference between Wi-Fi positioning and 3GPP based OTDOA positioning is that Wi-Fi positioning (IEEE 802.11mc) may be based on round-trip time (RTT). In contrast to the OTDOA algorithm described above, the advantage of using RTT is that there is no need for network time synchronization.

Ultra-wide-band (UWB) techniques have become increasingly popular in positioning solutions, since the inherent high time resolution in UWB signals enable precise positioning. There are several UWB-based positioning products available. Many of them are based on DecaWave UWB technology, but there are also proprietary solutions (e.g., Zebra).

UWB can be used in multiple algorithms. It can support downlink or uplink TDOA, angle of arrival using multiple antennas, as well as direct range measurements, where network time synchronization is not needed at all.

Due to the nature of the very short transmission pulses used in UWB techniques, UWB can detect and eliminate problems due to multipath propagation, since reflections are individually detected and can be filtered. This is a clear advantage compared to narrow band systems, where such discrimination is not possible. The precision of time of flight is in the range of 2-5 cm. When applied in a real environment, the positioning accuracy with UWB is on the scale of 10 cm.

One advantage of UWB is the potential for cheap devices, compared to 3GPP modules. Commercial UWB transceivers are available for approximately 3-4 USD. This enables increased installation density, flexible choice of algorithms to support various use-cases and cloud platform to support a global ecosystem that can serve various segments globally.

Lidar

Some positioning techniques estimate the distance by measuring the round-trip delay of an ultrasonic or electromagnetic wave to the object. Ultrasonic waves suffer large losses in air and cannot reach distances beyond a few meters. Radars and lidars use electromagnetic waves in radio and optical spectra, respectively. The shorter wavelengths of the optical waves compared to the radio frequency waves translates into better resolution and lidar solutions are therefore a favorable choice for high accuracy positioning. As in radar solutions, the main components of a typical lidar include a transmitter and a receiver and the distance is measured based on the round-trip delay of light to the target. This is achieved by modulating the intensity, phase, and/or frequency of the waveform of the transmitted light and measuring the time required for that modulation pattern to appear back at the receiver.

Popular lidar architectures include pulsed and frequency-modulated continuous-wave (FMCW) schemes. Pulsed lidars rely on the particle properties of light and can provide moderate precision over a wide window of ranges, while FMCW lidars rely on the wave properties of the light. In these lidars, the modulation is applied to the frequency of the light field and the large frequency bandwidth in the optical domain becomes accessible and can be exploited to achieve very high precision localization with accuracy in the nano-meter range.

Summary of Positioning Techniques

Important properties of the positioning techniques discussed in this section are summarized in Table 3. Note that the accuracy numbers stated are only for indicative purpose. The actual positioning accuracy depends on various factors, including but not limited to the network deployment, cell planning, radio environment, etc.

Hybrid Positioning

Many devices in the market today are equipped with sensors such as an inertial measurement unit (IMU). The IMU may contain a 3-axis gyroscope and a 3-axis accelerometer, for example. Data provided by the IMU can enable the location server to estimate the UE trajectory between, after, or during an OTDOA/E-CID positioning session, and can reduce the need for frequent OTDOA/E-CID measurements. A hybrid positioning solution using IMU may also be beneficial in scenarios where the device may move out of positioning coverage part of the time, thereby increasing the positioning reliability. An example use of IMU data together with position estimates is illustrated in FIG. 17 . The same method may be applied even if IMU measurements are not available, by estimating the speed and direction from old position estimates and predicting the UE trajectory.

Note that a positioning system solely based on IMU is a relative positioning system, i.e., it can estimate the position of a UE relative to a known coordinate. For example, the pressure difference over a period translates to an altitude change, and an acceleration during a period indicates a change of speed.

In order to fuse the radio measurements with IMU data, it is required that the data reported from the IMU equipped UE is aligned with a standardized earth bounded coordinate system. Or, that the UE reported IMU measurements can enable the location server to translate the measurements into an earth-bounded coordinate system. To get the UE position in earth-coordinates, the orientation of the device is needed. A common method to determine the orientation is to use gyroscope, magnetometer and accelerometer. After the orientation is estimated, one can use the orientation and accelerometer to estimate the acceleration relative the coordinate system (accelerometer minus gravity). By having the relative acceleration, it is possible to estimate the relative displacement of the device by, for example, double integration.

LTE Rel-15 includes support for IMU positioning and specification of the signaling and procedure to support IMU positioning over the Location Positioning Protocol (LPP), as well as hybrid positioning that includes IMU related estimates.

Network Synchronization Accuracy

For OTDOA as well as uplink time-difference of arrival (UTDOA), which is the positioning method supported in LTE, network synchronization errors leading to errors in the TDOA estimates may dominate the overall positioning error. It is therefore important to understand what network synchronization accuracy that can be achieved. The synchronization errors can in principle be directly translated to positioning errors by considering the distance light travels during the timing error caused by the synchronization error, that is, a synchronization error of 1 ns corresponds to a positioning error of 0.3 m.

The synchronization error mainly consists of four additive parts:

1) Error in external synchronization reference delivered to the anchor (baseband unit for macro base station and DOT). 2) Synchronization error between anchor (baseband unit) and external synchronization reference. 3) Synchronization error within the radio base station (RBS). 4) Synchronization error between the RBS antenna and the UE.

When considering manufacturing scenarios, we have the following situation for each of the four parts:

1) The external synchronization reference typically comes from a GNSS receiver. A GNSS receiver might have an accuracy of <50 ns when it has LoS to a large portion of the sky, but when there is multi-path, the accuracy decreases rapidly and the assumed accuracy from an indoor GNSS receiver is <200 ns. The external synchronization can be improved by using a more expensive GNSS receiver with better multipath filtering and better internal accuracy. 2) The baseband unit can synchronize to a GNSS receiver with an accuracy of around 150 ns. This number can only be improved by using better hardware, that is, a new baseband unit. 3) The budget for internal distribution is 130 ns. What it will be in practice depends a lot of the hardware configuration of the RBS. The simplest is a single baseband unit connected directly with one hop to an antenna integrated radio (AIR) unit. 4) How large the synchronization error between the RBS antenna and the UE will be unclear. However, this error will not impact the OTDOA positioning error since OTDOA builds on the phase of the PRS observed by the UE in the position it is, and synchronization between network and UE is not required.

The discussion above assumes the RBS is time locked to the reference. When the RBS goes to time holdover, the accuracy may decrease.

The numbers above can be significantly improved by radio-interface based monitoring (RIBM), where the phase difference between an RBS antenna and a reference is measured and reported. The reported phase difference can then be accounted for when calculating the position of an object. The synchronization error remains, but the part of the error which is accounted for will not affect the OTDOA positioning accuracy. This is a promising method, since RIBM may achieve a virtual synchronization accuracy of about 20 ns between RBSs. Thereby, the external synchronization reference error 1) does not affect the OTDOA positioning and can be ignored.

For RDS, the synchronization between the DOTs connected to the same indoor radio unit (IRU) is in the order of 6 ns, under the assumption that the IRU contains common DOT hardware. This holds for both the legacy DOT available today and the digital 5G DOT that will be available in 2019. RIBM may be a solution to achieve synchronization between DOTs connected to different IRUs but may require a specialized feature to operate, since the standard RIBM algorithm require the node to be synchronized to receive when not transmitting, which DOTs do not.

In summary, practical network synchronization algorithms can imply network synchronization errors of up to 360 ns, when locked to the GNSS reference, which corresponds to a positioning error of up to ±110 m (3 sigma). If the RBS is in time holdover, the accuracy can be even worse. In the future, RIBM may provide virtual synchronization accuracy of about 20 ns, which corresponds to a positioning error of 6 m. For RDS, positioning using DOTs connected to the same IRU is affected by synchronization errors of about 6 ns, which corresponds to a positioning error of around 2 m. If DOTs connected to different IRUs are utilized to estimate the position, RIBM may in the future be applied to provide virtual synchronization accuracy of about 20 ns.

Improving Accuracy in Severe Multipath Scenarios

LTE OTDOA, as well as other positioning algorithms, suffer severe penalties in terms of positioning accuracy if there is no way to handle the problem of multipath. In essence, OTDOA assumes that the RSTD represents the LoS path, but it is in general hard to determine whether or not the LoS-path is blocked or very damped. At the very least, one can say that an nLoS path represents an upper bound on the distance between transmitter and receiver. The multipath problem is significant in typical industry environments, especially for high frequencies. Some approaches which address the multipath problem include:

Design the network to have good LoS conditions by placing nodes at favorable locations, for example by using a planning tool which can estimate the positioning accuracy of a number of reference locations. This may imply higher installation complexity and may not be feasible for industry environments with many moving objects. Estimate nLoS using hypothesis testing, e.g., feasibility tests using positions from subsets of TOA/TDOA estimates. If the estimate is incoherent, categorize it as nLoS. This method may have high computational complexity for dense networks. Another alternative is to consider position updates based on IMU measurements and dead-reckoning as a reference and categorize paths as nLoS if the measured TDOA doesn't match the reference position well enough. Use an environment model for ray tracking and associate distance estimates with rays. In this way, nLoS estimates may contribute in a better way to the positioning estimation. Such environmental models are often not available but may be used if a digital twin is developed. Estimation of nLoS using polarization. The polarization changes at the bounce events, so nLoS is detected if a reference polarization is erroneous. Compare individual distance estimations with a communication-independent velocity/position measurement (gyro/accelerometer) to determine which estimations are feasibly LoS. Of course, not all UEs are equipped with such location sensors. Rel-14 introduced an important enhancement to LTE OTDOA called multipath reference-signal time difference (RSTD). The main idea was to include several possible peak candidates from the power delay profile (PDP) and estimate the position using maximum likelihood. In this way, positioning accuracy was improved significantly, since the algorithm made no hard decisions on the LoS paths.

In summary, we can conclude that selection of the most suitable positioning system for manufacturing must take several different aspects into account, including:

Required accuracy Required latency Deployment aspects Devices

When it comes to accuracy, there are industry use cases with positioning requirements ranging from mm-level accuracy to several tenths of meters. In an environment with a high probability of LoS and few blocking obstacles, it may be possible to estimate position within ten meters to a few tenths of meters accuracy in deployments with only one or a few micro base stations mounted, for example, in the ceiling of a factory hall. However, in many manufacturing scenarios, the reality is an environment with multipaths, reflections and many blocking obstacles. In such scenarios, it seems like the radio dot system (RDS) is the most suitable solution for providing combined communication and positioning, due to dense deployment and low cost of the nodes.

The accuracy numbers stated for different positioning techniques often assume that network synchronization is sufficiently good. However, in many cases network synchronization errors in the order of 10-20 ns, corresponding to 3-6 m positioning error, is very difficult to achieve. For positioning needs, virtual network synchronization achieved through RIBM is the most promising solution.

UWB solutions are becoming increasingly popular in industry and manufacturing use-cases due to the high accuracy and relatively cheap devices. However, one drawback is that the UWB solutions are not integrated with a communication system, which is the case for the 3GPP-based solutions, for example. In the future, NR may replace UWB, since NR will also use very wide bandwidth.

Positioning latency requirements in manufacturing are relaxed for most use cases. For example, keeping track of tools and assets don't require very frequent positioning updates. The most demanding manufacturing use cases in terms of positioning latency may be safety related. One example is a real-time alarm to warn workers when a forklift is close. The trade-offs between high accuracy and positioning latency for such a use-case have not been thoroughly studied yet.

The relation between device and anchor is also important to consider. For LTE and NR, both devices and anchors are complex and costly, and solutions built on these techniques are therefore not suitable for use cases where many small objects should be tracked, since each object must have their own LTE or NR device. In this case, UWB or RFID may be more suitable since most of the complexity lies in the anchor node and the device/tag is therefore cheap. 3GPP-based technology with cheap devices, such as NB-IoT or CAT-M, can be also be considered for this type of use cases, but since these techniques are narrowband, the positioning accuracy that can be achieved is low.

Spectrum

For industrial applications, some particular issues relating to spectrum include:

Spectrum suitable for local usage, Regulatory means to enable local spectrum usage, e.g. leasing and local licensing, Technical means to enable local spectrum usage, e.g., evolved Licensed Shared Access (eLSA) and Citizens Broadband Radio Service (CBRS). 5G NR will be used as an access technology over a variety of frequency bands, including many that are currently used for LTE. A few frequency ranges are likely to be more globally harmonized, such as 3400-3800 MHz and the upper part of the range 24-29 GHz, although variations in band plans will likely exist. The study process for WRC-19 (World Radio Conference for 2019) leading up to the identification of bands for IMT2020 may yield additional millimeter-wave (mmW) spectrum bands where a good potential for harmonization exists, e.g. 37-43 GHz.

Recent regulation has designated bands for “local or regional use” (e.g., 3700-3800 MHz in Germany and Sweden, 3300 MHz in China). Such actions are not a global radio solution for Industrial IoT. Still, the introduction of these bands is a good first step for private use licenses of spectrum. It is not anticipated that these regulatory actions will spread across all markets immediately. Therefore, it is clear that additional access to globally harmonized spectrum will require opportunities derived from spectrum licensed to mobile network operators (MNOs), essentially through business arrangements that allow leasing of spectrum, or of capacity under well-defined Service Level Agreements (SLA).

The availability of mmW spectrum does pose challenges for Industrial IoT, mainly due to the time needed to establish products in the market, and the complexity of semiconductor manufacturing at such high frequencies. Building practices for equipment are not established, with significant challenges remaining on cost effective solutions for devices. Millimeter wave equipment may also offer some advantages: the propagation characteristics of narrow-beam transmitters can enable better reuse with transmit power control and beamforming, and coexistence can likewise be easier. The frequency bands are also suitable for wider bandwidth signals, although uncertainties remain in the amount of spectrum that may be made available for industrial wireless.

The fourth Industrial Revolution, known in literature as Industry 4.0, is an opportunity for 5G wireless technologies in manufacturing, exploration and process control situations within factories, mines, process industries etc. The exploitation of these opportunities will require access to spectrum, either unlicensed, shared or exclusively licensed. Indeed, there are clear indications from industry that lack of access to high-quality licensed spectrum is a key roadblock that has to be overcome. Access to licensed spectrum for Industrial IoT can be provided in one of three ways.

1. Service Level Agreements (SLA): Agreements with an MNOs can fulfill these requirements, with MNO-provided or provisioned service; e.g.,

On-premise MNO deployments on a turn-key basis or MNO permitting end-user deployment of approved equipment that is optionally connected to public networks. This case is not dealt with further in this discussion. An alternative is to establish a private virtual network that guarantees capacity on an MNO network through the use of network slicing.

2. Spectrum Leasing: MNO acting as a Lessor towards verticals. 3. Local Licensing: The regulator license spectrum directly to verticals over a limited geographical deployment, typically associated with property rights for the covered area.

The regulatory situation for spectrum leasing is summarized below. The regulation for spectrum leasing is of interest for possible business models for 5G use-cases

The US is most mature regarding spectrum leasing; regulations for leasing date back to 2003-2004. Spectrum leasing is used commercially. The public database Universal Licensing System (ULS) records all leasing agreements In South America, a few countries allow leasing between MNOs In EU, leasing of major mobile/cellular bands is allowed in regulation since 2012. It is not necessarily implemented in regulation in member states. No commercial leasing examples for MNO-owned frequencies to non-MNOs found so far In Asia, beauty contests generally prevent spectrum leasing in several important countries, and is thus not part of regulation In Africa spectrum leasing is generally not allowed.

Regarding local licenses, such licenses are presently non-existing for private/public use. Some 5G use cases, especially when associated with addressing industrial automation, would benefit from local licensing; 5G introduction offers an opportunity in this regard. Planned auctions of the first 5G band in Europe (3.4-3.8 GHz) have triggered regulatory activity in two countries (Germany and Sweden) by defining a particular realization of local licensing. In China, industry has shown interest in dedicated spectrum.

In order to give a complete view of possible solutions, unlicensed bands suitable for industrial applications are also mentioned. The unlicensed bands are generally unsuitable for URLLC due to the possibility of interference due to contention-based operation; the variation in access performance creates uncertainty in throughput and delay performance.

Evolved LSA (eLSA) is a solution, currently being specified, to support leasing and local licensing within regulations by the means of a database/controller architecture. The eLSA is supposed to support any band and be technology neutral. Similarly, the Citizens Broadband Radio Service (CBRS) in US, to be first used in 3550 to 3700 MHz band, will use a Spectrum Access System (SAS) to handle the regulatory requirements for that band. This is also a database/controller architecture that provides leasing opportunities for local area use while the actual licensing is covering larger areas as per FCC regulations. The SAS can be used for other bands as well given appropriate regulatory requirements. The eLSA and the CBRS would cater for co-existence between different deployments according to the required regulatory requirements in a country or region. However, the way in which coexistence is ensured differs between eLSA and CBRS.

Many different spectrum bands are identified for 5G. Here, only bands that are likely to use NR technology are discussed. For example, the 700 MHz band is identified as a 5G band within the European Conference on Postal and Telecommunications Administrations (CEPT), but will likely implement 4G. The same applies for APAC regarding the 700 MHz band. Also, the 2.3 GHz band has been discussed, for example in Sweden, but presently mainly in the context of 4G.

There are currently no 5G 3GPP harmonized bands valid and allocated in all countries of the world, but harmonized spectrum ranges, like 3400-3800 MHz, 24.25-29.5 GHz, do exist. Several 3GPP bands will be defined within each range. Many mmW bands pending allocation are dependent of the outcome from WRC-19.

Europe

The 3400-3800 MHz band is identified as a “pioneer band” for 5G in CEPT. The plans for different countries vary a lot, depending on incumbents with very different license expire dates. There are countries that plan to auction the full band and then usually 100 MHz blocks are proposed, like in Sweden. Others only have the upper or lower e.g. 200 MHz available presently, due to incumbent usage. This results in more narrow band licenses like in the UK. When the remaining spectrum becomes available new auctions will take place. This will result in non-consecutive spectrum holdings for the operators, if nothing is done, such as a re-allocation of the band. Re-allocation might not happen, since “carrier aggregation exists”.

Most countries promote national licenses, except Germany and Sweden who propose to set aside 100 MHz (3700-3800 MHz) for local services according to existing plans. The block is generally available in Sweden 2023.

In the “5G action plan” from EC (European Commission) it is defined that all countries shall have:

A 5G network in service in at least 1 city in each country during 2020. Full build out ready 2025.

This will probably mean that most countries will focus on mid-band (3-8 GHz), since various national coverage requirements will exist, in order to fulfill the EC ambitions.

The 26 GHz band (24.25-27.5 GHz) is also identified for 5G. The exact definition is to large extent depending on the outcome of WRC-19. In most countries, the range 26.5-27.5 GHz is empty and can be auctioned now. In some countries auctions have already started.

United States

In the United States, there are several bands targeting 5G on mmW (24/28/37/39 GHz), and it is only recently that the Federal Communications Commission (FCC) has started to consider mid-band spectrum (e.g., the 3.7-4.2 band). Operators have identified portions of the existing bands for deploying NR, e.g., T-Mobile on 600 MHz and Sprint on 2600 MHz.

There are upcoming FCC auctions of mmW spectrum at 28/39 GHz which are not owned by existing licensees.

On mid band, 5G will be allowed in the CBRS band (3550-3700 MHz). The band has licensed (PAL) and general authorized (GAA) blocks, based upon 10 MHz blocks. On 37-37.6 GHz it is proposed that licenses for local use is defined.

Asia-Pacific

Several of the major countries in the Asia-Pacific region are planning auctions during 2018/2019.

Korea auctioned 3.5 GHz (3420-3700 MHz) and 28 GHz (26.5-28.9 GHz) in June 2018 to operators.

China are planning to allocate additional frequencies on 2.6 GHz (160 MHz in total) and 4.9 GHz (100 MHz) to CMCC during 2018. Auctions for 3.5 GHz (3300-3600 MHz) is planned for 2019, where 3300-3400 MHz is planned for indoor usage. Presently 2300 MHz is mainly 4G indoor, but so far, no indication of allowing 5G.

Japan is planning a contest for band 3.6-4.1 GHz, 4.5-4.8 GHz (200 MHz for private operation) and 27-29.5 GHz (900 MHz for private operation) during 2019. Note that parts of 3400-3600 MHz are already allocated to LTE and will eventually be converted to 5G, but in Japan the band allocation is defined by law and can take long time to change.

Australia is planning the 5G auction on 3400-3700 MHz during late 2018.

Other countries that are in the process to plan 5G auctions are India, Indonesia, Pakistan, Thailand and Vietnam.

Middle East

Countries like UAE, Saudi Arabia and Qatar have concrete auction plans 2019-2021 for 3.5 GHz and 26 GHz. Other countries have also indicated upcoming auctions, but no detailed are known yet.

Summary of Spectrum

A summary of 4G/5G spectrum bands in different countries is shown in Table 4. The items that are shaded can be used can be used for local service and for industrial automation.

Regulatory Methods to Control Access to Local Spectrum

There are two other regulatory methods to get access to local spectrum:

Spectrum lease, Local licensing.

These methods are applicable if the operator accedes to customer control over spectrum, or if regulators establish local licensing as a viable policy.

Under the spectrum lease approach, a Licensee/Lessor lease parts of his license to a Lessee, with or without a fee. The Lessee can lease parts of the frequency band, a portion of the spectrum to a particular geographical area, or both. A sublease is when the Lessee leases out spectrum to a secondary lessee. A regulatory view of spectrum leasing is shown in FIG. 18 .

Regulations for leasing of spectrum differ from country to country. Numerous aspects can be regulated:

The terminology Differing or not differing between de jure (legal) and de facto (in principle the radio network owner) control over the spectrum The process for application, for example the stipulated time to approval Which bands are available for leasing, considering for example competitive implications The term of the lease, while not exceeding the term of the license authorization The possibility of subleasing The area defined for the lease And more . . . .

Also, the regulators can choose to make more or less of the leasing agreement public.

The following is an overview of the regulatory situation regarding spectrum leasing:

The US is most mature regarding spectrum leasing; regulations exist since 2003-2004. Spectrum leasing is implemented commercially. There are examples of spectrum leasing, spectrum aggregators, spectrum brokers. There is a public searchable database (ULS), which contains all leasing agreements In South America, a few countries allow leasing between MNOs. In EU, leasing of major mobile/cellular bands is allowed in regulation since 2012. It is not necessarily implemented in regulation in member states. It does not appear that there any commercial leasing examples for MNO-owned frequencies in Sweden, Finland, UK, Ireland, Germany, France or Italy. In the UK, leasing is not allowed by Ofcom in the major cellular bands, due to competition implications. In Ireland, leasing is allowed in the major cellular bands, after ComReg review of competition implications. In Sweden, spectrum lease is permitted. The regulator has so far only allowed short-term leasing due to their auction planning (lack of stable long-term plans). Operators have so far not allowed long-term leasing with protection guaranteed, due to uncertainties in their network planning/build-out. In Finland, spectrum lease is permitted, but leasing has never been addressed in relation to MNO licenses and thus this case has never been implemented by the regulator. For Germany, spectrum leasing was for the first time addressed in a consultation on 3.7-3.8 GHz autumn 2018. The regulator defined property owners and users (tenants) as licensees. In Italy, spectrum lease is permitted, and used commercially. In Asia, contests generally prevent spectrum leasing in several important countries, and thus spectrum leasing is not part of regulation, e.g., not allowed in China, India, Japan, but there is growing interest in spectrum trading. Spectrum leasing is allowed but not used in Korea. In Africa, spectrum leasing does not generally appear to be allowed. However, Nigeria recently published Spectrum Trading Guidelines including spectrum leasing. The commercial leasing of MNO spectrum in the US concerns several cases. Nationwide operators lease spectrum among themselves to address markets where capacity/coverage/growth is needed. Nationwide operators lease spectrum to non-nationwide operators, for example Verizon's LTE in Rural America (LRA) program. Verizon has signed up 21 rural and smaller carriers to the program, and 19 have launched LTE networks via the program. The program allows Verizon to quickly build out rural areas. In The MNO interest in leasing out spectrum for 5G verticals, for example in a factory automation use case, is yet to be seen.

Spectrum leasing of mobile bands from operators has primarily been done towards other operators in order to fulfill coverage and other requirements from the regulator. Volume-wise, this is almost exclusively in the US.

In higher bands (>10 GHz), fixed services are a use-case which involves leasing from operators by service providers. This is established in both US and Europe.

With the arrival of 5G, verticals provide use-cases in need of dedicated (mobile) spectrum. One question is which actor is going to be Lessee.

The operators' reactions long term on the possibility of leasing out mobile spectrum to verticals are not known. There are opportunities and issues for both the Lessor and Lessee for such a leasing agreement to take place, for example:

Interesting for MNOs for spectrum that is not fully exploited MNO may hesitate to lease out spectrum in areas with heavy demand, or where demand is anticipated within 5-10 years The needed lease time of 30+ years, due to investments in processes and buildings, is far longer than the MNO's license duration

The introduction of 5G will cause a widespread change in the ability of operators to provide SLAs through network slicing. While network slicing is supported to various extents with all 3GPP based networks, the 5G CN will provide operators with a framework that enables programming network slices to effect separation between use cases, QoS classes as well as service providers. It would then be possible to have a deployment case where slicing can enable the leasing of network capacity. This would allow the local user to be in control of end-to-end SLAs and even control the behavior of the RAN, including, for example, QoS, within limits. The MNO would deploy and integrate the RAN with the CN according to the SLA of the leasing without losing overall control over planning and administration

Bands possible for spectrum leasing must comply with certain requirements. Leasing of the band must be allowed in the regulation in the specific country/region. Removing China, Africa and Japan from Table 4 above results in the following table:

Local Licensing

The majority of licenses use pre-defined administrative boundaries for defining the area for a license, such as:

National borders Regional borders, or other larger administrative structures Communities/Municipals

The next level of granularity could be property. With this approach, property and land usage rights can be used the administrative definition to be used for local licenses. If local spectrum is needed from a larger area spectrum license, one solution to increase the granularity would be to use leasing of sub-areas. This solution can define areas bigger than property, if needed.

Presently when a regulator defines a local license for an area that is smaller than a region/municipal, the definition has been a coordinate and a radius, an event name, an address, coordinates defining an area, etc. This has not been a problem since the number of such licenses has been low. However, with the arrival of 5G use-cases, this will change. It is work in progress for regulators.

If the number of local licenses grows with the 5G use-cases, the coordination needs also grow:

Geographical data bases are needed to show the licensed areas, for example for new applicants. Interference between the implementations needs coordination through regulative requirements.

While national and regional licenses for commercial services exist, local licenses exist for non-commercial purposes, such as test labs and test plants. Possibly licenses for some program-making and special events (PMSE) services could be seen as local. The arrival of 5G, with new types of use-cases, will require local licenses for factories, for example, and puts new requirements on the regulation.

The primary band for 5G services in Europe is 3.4-3.8 GHz, and the auctions of this band triggers regulative activity regarding local licenses. Higher bands, for example 24.25-27.5 GHz (pioneer band for early implementation in Europe), are suitable for local use in the sense that their propagation characteristics are less likely to cause coexistence problems, especially when being used indoor. Presently, regulatory discussions regarding local licenses for these bands are largely out of scope or may just be entering the realm of possibility, but this is expected to change with industry interest.

Certain indoor environments are amenable to reuse of spectrum across multiple uses, especially if the networks are separated by floors in modern buildings. It is well known that the loss across multiple stories of a building can be many tens of dB, even at mid-band frequencies like 3.5 GHz.

Industry in China has shown interest in local licenses in standards fora, pointing to the proposal for 3.7-3.8 GHz in Germany.

An example of a license assigned to industry is the allocation from 1800-1830 MHz that was provided to the Canadian hydroelectric power industry in the 1990s.

Europe: 3.7-3.8 GHz (Part of 3.4-3.8 GHz, Primary Band for 5G Services)

Several countries have auctioned the spectrum and more to follow. Two countries have had consultations including local licenses and are following each other.

Germany promulgated auction rules in the third quarter of 2018 with an auction scheduled for 2019Q1-Q2. The rules distinguish between indoor and outdoor usage for “local property-related uses”, implying property as a definition of local. Note that there is property which is not private property, for example streets, parks, etc. Sweden, auction latest Q1 2020. Recent consultations define “local block allocations.” Local here is defined as referring to “small geographical areas,” e.g. mines, indoor facilities, and hot spots. It should be noted that also regional licenses, usually corresponding to a municipality, are mentioned in the consultations.

USA: 3.4-3.55 GHz (Possible Extension to CBRS)

In the US, the National Telecommunications and Information Administration (NTIA) is evaluating spectrum sharing between military radar and mobile broadband in this band. The incumbent systems differ here in comparison to the existing CBRS range. If CBRS rules apply, licensed operation would be across counties, and the third tier would be comprised by General Authorized Access. The large area sizes in the CBRS band make it unsuitable for industrial use as verticals are unlikely to participate in the auction market.

Bands for Local Licensing

Bands possible for local licensing must comply with certain requirements. Local licensing must be allowed by the regulator

Keeping the local initiatives in Table 4 above results in the following table:

Technological Support for Leasing and Local Licensing

In Europe, eLSA is a continuation of the ETSI specified system Licensed Shared Access that manages access to spectrum in IMT bands where the incumbents cannot be evacuated within a reasonable foreseeable time. The access can be managed in time and geographical area. The system creates geographical protection and exclusion zones which incumbents does not allow others to use. In eLSA, allowance zones are introduced to enable also local licensee handling where the process of granting and managing the many local access licenses can be automated. It would also include the handling of leasing of frequencies to local area users from established licenses such as MNOs. FIG. 19 shows the assumed spectrum allocation possibilities for a frequency band allocated to mobile services such as IMT.

The specification work on the eLSA system has started in ETSI (Europe) and is based on the ETSI Technical Report “Feasibility study on temporary spectrum access for local high-quality wireless networks”. The technical specification on System Requirements is assumed to be ready end of 2018 and to be followed by specifications on Architecture & Procedural Flows, and then followed by a protocol specification. In Asia-Pacific information related to “local area” services have been shared and starting work on a technical report has been approved.

The eLSA system is based on a Database/Controller concept. It supports licensing and leasing but not unlicensed/license-exempt operation with e.g. granted access such as white space or licensed-by-rule access, as this will not provide the necessary interference protection requirements.

The database is named eLSA Repository and assumed to be in the Regulatory domain. The controller is named eLSA controller and ensures that the eLSA licensee's system would have the needed configurations to operate according to the licensing conditions, thereby readily supporting high quality needs to support the URLLC use cases. The controller will get the required regulatory sharing and co-existence requirements from the eLSA repository.

FIG. 20 shows a possible architecture sketch for local licenses, while FIG. 21 shows the possible architecture for leasing. In the latter case, the eLSA controller box also contains some of the eLSA Repository functionality because the MNO is the one leasing out frequencies.

In the US, the Federal Communications Commission (FCC) has defined the Citizens Broadband Radio Service (CBRS) in the 3550-3700 MHz band, in regulations codified in the FCC rules. FIG. 22 illustrates aspects of the CBRS.

The CBRS band is in use by naval radar and by the Fixed Satellite System (FSS) service, both services constituting Tier 1 incumbent primary use. Grandfathered Wireless Broadband Service Users, such as Wireless Internet Service Providers (WISP), operating under the rules of 47 CFR Part 90, Subpart Z, are also protected from interference from the CBRS until April 2020. The two remaining tiers respectively allow the issue of Priority Access Licenses (PAL) and General Authorized Access (GAA) in the band for wireless broadband use. PAL users benefit from licenses to spectrum based on the acquired licensed area and bandwidth. GAA users are allowed access any spectrum not utilized by higher tiers based on authorized access.

Radio devices are registered as Citizens Broadband Radio Service Devices (CBSD) based on their location and their operating parameters. Any eligible radio device may request access to Priority Access License (PAL) and GAA spectrum. Since the FCC does not confer any regulatory protection for GAA spectrum users, it is left to industry agreements to create solutions for GAA coexistence. While the Wireless Innovation Forum (WInnForum) is specifying technology agnostic protocols that are mostly geared towards regulatory compliance, the CBRS Alliance is seeking to improve the performance of LTE networks operating in the CBRS.

The CBRS Alliance was chartered as an industry trade organization seeking to promote and improve the operation of LTE in the band, for a variety of use cases, including operator-deployed small cell networks associated with public service, fixed wireless service for last mile replacement and Industrial Wireless. The Alliance is specifying changes to network architecture to allow both the traditional operator deployed operation and private network operation including neutral hosts and has provided a platform to establish the impetus for contributions in 3GPP for defining Bands 48 and 49 for LTE-TDD and LTE-eLAA operation in the band. The CBRS Alliance will also introduce 5G NR into the band in 2019. The focus of 5G on Industrial wireless applications fits with the mission of the CBRS Alliance.

The Spectrum Access System (SAS), a geolocation database and policy manager, authorizes access to CBRS spectrum by CBSDs. The SAS primarily protects higher tier users from lower tier operation in accordance with the FCC regulations. The logical relationships in the CBRS are described by the SAS-CBSD and the SAS-SAS interface, as shown in FIG. 23 , which illustrates the high-level SAS architecture, including coexistence manager (CxM) functionality for the GAA spectrum. Federal radar systems are protected by the implementation of a network of sensors forming the Environmental Sensing Component (ESC) that informs the SAS about coastal radar activity. PAL users are awarded regional licenses over large geographical areas over 10 MHz blocks.

Each PAL is 10 MHz and is limited to a maximum of seven licenses confined within the first 100 MHz of the CBRS band, i.e., 3550-3650 MHz. New rules have based license areas on counties, which number 3142 in the United States. There are seven PALs in each license area, the license terms are ten years with a guarantee of renewal, and licenses can be partitioned and disaggregated. Single operators are capped at a maximum of four PAL licenses. The ability to lease spectrum under geographical constraints and the ability to disaggregate licenses will support a secondary market in spectrum use for industries. In this way, PAL licenses can likely support URLLC without significant encumbrance.

The WInnForum is defining technology-neutral mechanisms for administering the band, including protection of incumbents, and PALs. Additional requirements for coexistence between GAA users is being developed, with much debate about whether coexistence should be engineered by the central authority of the SAS, or by local action by CBSDs arising from knowledge of the radio environment.

FIG. 24 is an illustration of PAL spectrum management. PAL users are protected only within a coverage area with a contour drawn around an actual deployment of one or more CBSDs. These coverage areas are known as PAL protection areas (PPA) and are bounded by a signal level of −96 dBm from the transmitting station. PAL Protection Areas (PPA) represent deployed clusters of CBSDs with overlapping coverage areas that may be fused to register a polygonal region qualifying for interference protection from other unassociated use of PAL or GAA. The figure shows several license tracts, each of which corresponds to a county. PAL users that span across multiple license tracts, i.e., having licenses in more than one tract, can combine their licenses to create a common channel assignment.

GAA users may use PAL spectrum so long as actual PAL deployments in PPAs are protected from aggregate interference that exceeds −80 dBm within the PPA. This is illustrated for PPA C in the figure, where two GAA CBSDs are allowed to operate on the same channel as the PAL user so long as the aggregate interference from their transmissions does not exceed −80 dBm over most of the PPA boundary. PPAs from different operators that overlap or are in close enough proximity will obviously use exclusive spectrum allocations. Thus, GAA users are guaranteed access to all of the 150 MHz of spectrum if the band is unencumbered by higher tiers.

While GAA users are not protected from mutual interference or interference from higher tiers, the WInnForum, and the CBRS Alliance have been engaged in attempting to specify methods of creating higher quality of experience for GAA users. The CBRS Alliance procedure reallocates the spectrum assigned by the SAS to the CBRS Alliance Coexistence group, creates local interference graphs based on environmental modelling, and optimizes spectrum allocation from a Coexistence Manager (CxM) that advises networks of CBSDs. In addition, the CxM manages Uplink-Downlink coordination for the TDD signal. LTE-TDD networks are all expected to be cell-phase synchronized and the CBRS Alliance coexistence specification details how this is to be achieved, independent of the SAS or CxM. The WInnForum and CBRS Alliance will try to guarantee at least 10 MHz of spectrum per CBSD. This is likely to be inconsiderate of eMBB service in congested areas.

Both PAL and GAA spectrum can address URLLC requirements, but URLLC quality cannot be guaranteed in all GAA situations. An operator would therefore not be able to enter into SLAs that would promise a customer that capacity or latency performance would not degrade, unless the facility being covered were physically isolated from other interferers.

The CBRS has several disadvantages:

The three-tier nature of the licensing regime, and especially the rules allowing GAA, create much uncertainty in the utility of the band. Some of this uncertainty arises out of a lack of understanding on whether GAA spectrum is like unlicensed spectrum or is akin to white space: it is not. Indeed, a strict interpretation of the FCC rules around GAA use would lead one to wrongly believe the white space analogy. The WINNF specifications have created an impression among some operators that they might solely use GAA spectrum in a manner that assures interference protection. Such an impression would perhaps require division of spectrum among users to the extent where bandwidth is traded off for quality. This is not suitable for eMBB use. The WINNF and CBRS Alliance Coexistence specifications are prone to devalue outdoor deployments. High power base stations are counted towards incumbent protection to a greater extent than low power ones. Indoor deployments may be favored when assigning spectrum, and large networks of indoor nodes could grab more than their fair share of spectrum unless the SAS introduces fairness measures. We expect there to be pragmatic approaches based on business guarantees. The interesting use cases for the CBRS are in urban small cells and micros for coverage and offloading. The reason for this is mainly the large license areas, and operators are more likely to bid for licenses in lucrative markets. For the industrial automation use of the CBRS, operators must acquire licenses with an intention to either disaggregate or lease their spectrum. The CBRS has been defined in a band that is of prime interest for 5G. The terms under which the spectrum is being offered, especially PAL, makes the band questionable for eMBB service and has mixed utility for industrial purposes, including URLLC modes of operation.

On the other hand, there are things to like about the CBRS:

The use-it-or-lose-it approach used by the CBRS is a well thought out exercise in improving spectrum utility. The rules motivate operators to use their license with real deployments, and operators have an incentive to either deploy their own radios or to lease out PPAs to realize revenue out of their spectrum. If most users in the CBRS are indoor small cell users, the success of the band would be guaranteed. A large number of industrial use cases qualify. The FCC cannot skew auction procedures to favor industry and enterprise use of spectrum. Industrial users are moreover not really interested in competing in the license market. Indeed, Ericsson's local license concept depends on the licenses being assigned by rule to real-estate owners, possibly for a small registration fee. By allowing disaggregation of licenses, the FCC has provided industrial users with choices—leasing, buying, or operation with GAA rules. The CBRS is due to go into commercial deployment and will be proven in the field. The establishment of industry organizations developing the standards, including those for LTE use in the band assure actual deployment and success of the band in some form. The same cannot be said about LSA in 2.3 GHz. Indeed, there is noticeable interest around the CBRS among other regulators, e.g., Ofcom. The experiences in deploying the CBRS may encourage the telecom industry to accept the CBRS as a tolerable way of implementing spectrum sharing.

Coexistence

In general, coexistence problems will exist when industrial networks using cellular or RLAN technologies share spectrum with other services, e.g., satellite. It is possible for industries to gain access to spectrum that is globally designated for use by radio navigation, satellite services, or fixed services, provided there is sufficient isolation in geography or through path loss between such services. For example, indoor factory use of spectrum can easily occur in satellite bands. It is desirable that such bands be close to bands allocated for RLAN use or IMT so that there is an incentive for manufacturers to include such bands within radio equipment. The CBRS band is one such band having close association with bands already designated for mobile use in most markets around the world.

Another aspect of industrial use of spectrum is the problem of spectrum utility. While cellular technologies have the advantage of high spectral efficiency, it is also necessary that regulation enable a high degree of reuse of spectrum. In many cases, this will involve understanding the extent to which spectrum can be reused in license areas that are in close proximity.

There are cases when co-existence in reality is not problematic. Indoor spectrum for use-cases for industry can benefit from unusable spectrum (including not mobile allocation) for others, for example satellites, FS, FFS, radar (not indoor only). However, use-cases for outdoor spectrum for industry must also be considered

While it is true that indoor industrial use cases can benefit from secondary use spectrum that may be designated for other services, e.g., satellites, FS, FSS, radar etc., it is just as important to designate local spectrum for outdoor use by industries.

Shared Spectrum—Market Considerations

The philosophy behind shared spectrum regulations differs between the USA and Europe. The FCC has been willing to define the CBRS in a manner that places high value on spectrum utility, while the EU has tended towards spectrum quality and stability.

In Europe, the support of industrial IoT is focused on licensed bands since the level of interference from others can be contained to a certain level. It is expected that there will be many licenses and leasing contracts in a country.

An evolved LSA system is being designed in ETSI to handle deployment and coexistence issues in an efficient manner. Depending on the country the co-existence scenarios with co-channel possibilities can be local indoor to local indoor, local indoor and overlapping regional coverage and local indoor to local outdoor. The regulatory sharing conditions needed to handle this in, for example, the frequency domain (and possible need of reuse pattern), guard distance (if needed), wall loss assumptions, and by setting a permissible maximum signal strength level at the border that would secure a predictability of interference to neighbors. This would facilitate the deployment of the network with known expected interference levels to secure the wanted network quality levels.

Unlicensed & licensed exempt bands which has unpredictable interference behavior can be used for services that does not require high QoS and can be simultaneously be used together with the licensed network.

In the US, it is most likely that leasing and private use of spectrum will happen within the CBRS. The PAL regulations are oriented towards protecting actual deployments. Areas within a license definition that are not covered by a licensee's radios are available to GAA users, provided established PPAs are not interfered with. The licensee is however free to monetize their license by allowing private PPAs to lease the license. This is likely to improve utility of the spectrum.

GAA use is open to private deployments and will have mixed quality of experience depending on a variety of factors: urbanization, population density, commercial interests, indoor vs outdoor deployments, outdoor deployments of small cells vs. large cells (low power vs. high power).

The WInnForum and the CBRS Alliance are engaged in defining coexistence principles for GAA that can reduce the interference impact to GAA users from cochannel use of allocated spectrum. The development of the procedures for coexistence are contentious and generally involve orthogonalizing spectrum allocations between neighboring CBSDs that are deemed to interfere with one another. This has the disadvantage of reducing the individual spectrum allocations to CBSDs under some circumstances. This is particularly worrisome for NR use in the band, especially in cases where eMBB coverage is anticipated.

Unlicensed Spectrum

Industrial use of wireless can overlap either cellular or radio local-area network (RLAN) technologies. Indeed, it is not necessary to classify all industrial use of spectrum as derivative of highly reliable or pertaining to critical communication. Certain characteristics of the industrial automation environment are a high level of importance to spectrum availability, ease of deployment and low regulation, and opportunities for developing trustworthy and secure networking. However, many use cases, and perhaps the majority of use cases for industrial use may also avail of license-exempt spectrum. The development of Multefire and LAA as technologies for license-exempt operation provide an avenue for the cellular industry to enter into the RLAN domain.

The key disadvantages of unlicensed spectrum include the necessity to operate in the presence of interference and the low reliability caused by shared use of spectrum based on distributed intelligence. This typically means that radio nodes use collision avoidance and listen-before-talk etiquette to access the channel instantaneously. This is not amenable for KPI guarantees. Therefore, unlicensed spectrum is usually not suitable for mission critical applications.

Unlicensed spectrum bands of interest for industrial applications span a wide variety of spectrum bands. The FCC in the United States has been the most aggressive in recent years in expanding the availability of unlicensed spectrum.

Table 7 lists unlicensed bands in various countries, where underlined text refers to bands that are under consideration. The bands in the table are those listed for broadband use and do not include several bands designated to be short range device communication bands. The unlicensed bands are generally unsuitable for URLLC due to the possibility of interference.

Spectrum leasing in Europe and US is not a regulatory problem, but the interest and business for the operators is to be seen. In Asia and Africa, spectrum leasing discussions has just begun.

Indoor spectrum for use-cases for industry can benefit from unusable spectrum (including not mobile allocation) for others, for example satellites, FS, FFS, radar (not indoor only). However, use-cases for outdoor spectrum for industry must also be considered

While it is true that indoor industrial use cases can benefit from secondary use spectrum that may be designated for other services, e.g., satellites, FS, FSS, radar etc., it is just as important to designate local spectrum for outdoor use by industries.

The eLSA is agnostic to bands and technology for spectrum leasing and local licensing. CBRS is going to be the best opportunity for industrial use of spectrum in the United States. CBRS is being specified in a spectrum range that is globally accessible by IMT, making economy of scale possible, and allows leasing as well as disaggregation of licenses. Allowing disaggregation does not mean that CBRS will enable local licensing.

Global harmonization of IMT and MBB spectrum has been a desire that has never been adequately realized. The consequence of diverse regulatory action on spectrum for mobile services over the years will make it difficult to achieve harmonization for industrial use cases. However, there is an interest in the telecommunications industry towards assigning portions of the spectrum ranges 3400-4200 MHz and 24.25-29.5 GHz for industrial use.

The 3400-4200 MHz will likely be the first frequency range, outside China and US, where buildout of 5G will start. The limited regulatory support for local licensing in this range will create delays for non-MNO dependent spectrum usage. For example, in Sweden, local licensing can at the earliest be available after 2023 for use by 5G, and in most other European countries regulatory action is not yet considered. Therefore, leasing will likely be the sole option for access to spectrum, except for an MNO-provided service, in that time frame. The mmW spectrum can be of interest to enable availability of local licensing in a timelier manner, but that will to some degree depend of the allocation of mobile bands in WRC-19.

Security

It is often said that security of a system is only as strong as the weakest link. However, depending on which part of the system that is, breaking (or neglecting) it can have very different consequences. When talking about a system involving more than one entity, the secure identities used, and the handling and protection of them, are building blocks that a big part of other security functionality relies on. The identities are used for authenticating the entities, for granting access and authorizing actions, and for establishing secure sessions between the entities. This means that a device needs to have a secure identity and to provide hardware (HW) and software (SW) based mechanisms that protect and isolate the identity and the credentials of the device. It is also not only the identities that need to be protected, but also the devices themselves should be secured, e.g. through proper control of what SW is being run on the device. All the aforementioned things are enabled by having a HW root of trust (RoT) in the device, basically a trust anchor to base security on.

Identity

An identity of a device is used to identify the device to a communicating party. The identity typically consists of an identifier and credentials such as a key, key pair, or password that is used for the authentication of the device. An authenticated identity enables the communicating party (such as network, service or peer device) to make well-founded security policy decisions for network/resource access control, service use, charging, quality of service settings, etc.

Identities based on a shared secret rely on the fact that all communicating endpoints, and only them, know a secret value. The randomness of the shared secret is one key characteristic. It is typically quite weak for a username-password pair, the most basic form of a shared-secret-based identity. In addition to randomness, the length of the secret and that the secret is handled securely, both at the device and the server side, are important.

With asymmetric keys, the identifier of an entity is the public key of the asymmetric key pair and the corresponding private key acts as the authentication credential. A signature generated using the private key can be verified by anyone having access to the corresponding public key. This is perhaps the main strength of asymmetric compared to symmetric (shared secret) keys.

To give additional value to the asymmetric key-based identity, it is possible to get the identity certified by a Certificate Authority (CA). The CA verifies the identity of the entity owning the key pair and issues a certificate certifying the link between owner and public key. The drawbacks with certificates include the size of the certificate (or certificate chain), which could be an issue in constrained environments, and the added cost of getting and maintaining (renewing) the certificate. To reduce costs, an enterprise can also set up its own CA.

Raw Public Key (RPK) mechanisms make a compromise between the simplicity of pre-shared keys and benefits of asymmetric cryptographic solutions. An RPK is a minimalistic certificate, significantly smaller than a typical certificate, containing only the public key in a specific format. An RPK is like self-signed certificates: there is no trusted entity that vouches for the provided identity, i.e., the peer receiving this identity needs to use an out-of-band mechanism to trust that it is the identity of the entity it wants to communicate with.

For all Public Key Infrastructures (PKI), it is recommended to have a way of revoking compromised keys. A Certificate Revocation List (CRL) that can be fetched from a Certificate Authority (CA) or checking the certificate status online using the Online Certificate Status Protocol (OCSP) are common ways.

3GPP cellular systems are a prime example of where shared-secret-based identities are used. A 3GPP identity consists of the IMSI, a 15-digit identifier, and its associated credential, a 128-bit shared secret. This information is stored in the subscriber database (e.g., HLR or HSS) in the 3GPP core network and on the UICC or SIM card installed in the User Equipment (UE). The UICC acts both as a secure storage and a TEE for the 3GPP credentials. For IoT devices, permanently integrated embedded UICCs (eUICC) can be used instead. eUICC has a smaller footprint and allows remote updates of the subscription data.

For 5G, 3GPP is also considering alternatives to traditional SIM credentials, namely so called “alternative credentials”. TR 33.899 looks at different identity solutions, including certificates. In specifications, support for certificates is described e.g. in 33.501 where EAP-TLS is defined as an alternative to AKA. EAP-TLS implies certificates are being used for authentication. For identifying the network, the use of certificates is partly available through the definition of the concealed identifier (SUCI), which is the private identifier (SUPI) of the UE encrypted with the public key of its home network, i.e. the network already has an asymmetric key pair for which the public key is part of the subscriber profile. The SUPI is defined in 3GPP TS 23.501; there, network address identifier (NAI) is given as one possible format of the identifier, which would support the use of certificates as well.

End-to-End (E2E) Security

In most cases, protecting the communication of a device is important. Both to protect against the information leaking to some unauthorized third party and against having the third party modify the data on the path. This can be achieved by applying confidentiality (encryption) and integrity (signing of data) protection. The exact security need for the data is very much use case dependent and relates to the data, its use, sensitivity, value, and risk associated with misuse of the data. However, as a rule of thumb, integrity protection should always be applied, while the need for confidentiality protection should be evaluated case by case. In general, a single key should be used for only one purpose (encryption, authentication, etc.).

For protecting the data, there are many standardized protocols available, including “regular” Internet security solutions such as TLS, IPsec and SSH. IoT optimized solutions include DTLS as a TLS variant for IoT and the ongoing work to profile IoT friendly IPsec. Also, application layer security solutions, such as OSCORE defined in IETF, are available and especially useful for constrained devices. The benefits compared to TLS include that end-to-end security can be provided even through transport layer proxies, e.g. for store-and-forward type of communication used with sleepy devices. The protocol overhead is also optimized.

3GPP also provides tools for protecting traffic end-to-end to a service, even outside the 3GPP network. The Generic Bootstrapping Architecture (GBA), 3GPP TS 33.220, uses the SIM credentials for authenticating the UE/subscription to a network service, called Network Application Function (NAF) in GBA lingo. GBA requires that there is a trust relationship between the service/NAF and the operator. Using that trust, the NAF can request session keys from the network, which are based on the SIM credentials of the UE. Those session credentials can be used for authentication and secure session establishment between the UE and the service.

Hardware Root of Trust

The concept of a hardware root of trust (HW RoT) includes the following aspects:

Secure storage Secure/measured boot HW-enforced Trusted Execution Environment (TEE) HW-protected crypto and key management (crypto acceleration, HW-based random number generator, generating/storing/accessing keys securely)

HW security is also extended to the environment where devices are manufactured such as protection of interfaces and mechanisms used during manufacturing and development, the use of secure key provisioning, key generation, secure configurations of devices, code signing, etc.

The base for securing that a device behaves as intended is to be able to ensure that only authorized firmware/software is running on the device. This requires a secure boot mechanism that originates from a hardware Root of Trust. The secure boot mechanism verifies during device boot-up that all loaded software is authorized to run. A HW RoT is an entity that is inherently trusted, meaning that its data, code and execution cannot be altered from the outside of its trust boundaries. It consists of functions that must operate as expected (according to its design), no matter what software is executing on the device.

A device must also have a secure storage mechanism to protect device sensitive data such as cryptographic keys while stored in (off-chip) non-volatile memory. Such a mechanism also relies on a HW RoT, e.g., a chip individual key stored in on-chip non-volatile memory or OTP memory.

In order to be able to recover from malware infections, and to minimize the risk of loss of sensitive data or changed behavior of the device, the security related parts of the firmware and software should be separated (and run in isolation) from other software. This is achieved using a Trusted Execution Environment (TEE) created using HW isolation mechanisms.

Device Hardening

A device typically contains interfaces and mechanisms for debugging and HW analysis with the goal to find issues of a given device discovered during ASIC production, device production, or in field. Joint Test Action Group (JTAG), IEEE standard 1149.1, is a common interface for debugging and various HW analyses. These mechanisms and interfaces must be protected such that they cannot be used by unauthorized persons for retrieving or modifying FW/SW and/or device data. This can be achieved by permanently disabling the interface, only allowing authorized entities to use the interface, or limiting what can be accessed with the interface. Also, for authorized access it must be guaranteed that sensitive data belonging to the owner/user of the device such as keys cannot be accessed by the person performing the debugging/fault analysis.

SW security is one of the most important building blocks of device security. Both HW and SW security complement each other. While it is not possible to build a secure device without HW security as a foundation, same also applies to SW security.

While IoT GWs with application processors commonly run Linux based OS, MCU based IoT devices mainly run light weight OSes such as mbed OS and Zephyr OS. There are also other highly security certified OSes being used on devices that have to meet high availability and security requirements. Choosing the right OS is important and security hardening of that OS is also equally important. Hardening entropy, user space components and network functionality can also be considered as a part of the OS security hardening process. Other aspects to consider relating to device hardening include

Using SW being developed according to the best secure software development practices. Sandboxing and isolation—running SW in a sandboxed environment. Least privilege concept—processes only get required privileges. Crypto hardening—using secure crypto libraries with traceable and reviewed code. Use of cryptographically secure PRNG. Certification (when applicable). Secure SW update—signed updates applied in a timely manner.

Security for Safety

However, these security mechanism/tools (maybe excluding non-repudiation) should be implemented in any secure system, regardless of whether the aim is making a safe system or not. The safety requirement is maybe more of an indication of the level of security required and that the configuration of security needs to be double checked as any error might have larger consequences than in a system without the safety requirement. The security configuration is also about choosing the right level of security/algorithms/keys used in the system. In addition to security, an (at least) equally relevant part for safety is e.g. the availability/reliability of the system, relating to both communication channels and services making up the system, or the correct operation of the components (values that are reported are accurate, time synchronization etc.).

Jamming

Like safety, jamming is also a topic that has one foot in the security domain. Jamming is a form of Denial of Service (DoS) attack. Some DoS countermeasures apply also for jamming, e.g. load balancing and rerouting traffic, which on the air interface would mean load balancing and rate limiting, backup base stations, and additional frequencies.

Industrial Devices

Industrial devices range from small simple one purpose sensors to large complex sets of devices such as robot cells and paper mills. Thus, one very relevant question we address here is what is a device? IoT devices are often categorized in two main ways: sensing devices and actuating devices. Sensing devices are equipped with some sort of sensor that measures a specific aspect such as temperature, light level, humidity, switch, etc. Actuating devices are such that receive a command and change state accordingly, e.g., a light bulb that can be on or off or air conditioning fan speed. More complex devices have a set of sensors and actuators combined but still only one communication interface. Even more complex machines may consist of several smaller devices made up of several sensors and actuators. Typically, even for a small device, a microcontroller or small computer is in place to host the communication stack as well as processing capabilities, memory and so on. In essence, a very complex device is in fact a small network in itself comprising of several parts that may or may not need to interact with each other and that may or may not communicate through the same communication module.

The range of requirements put on the communication itself varies depending on the purpose and criticality of the task the device in question is aimed for. These requirements can include throughput, latency, reliability, battery life and extended coverage. For instance, a simple sensor reporting temperature changes can be seen having relaxed communication requirements, whereas controlling robots wirelessly from the cloud requires an URLLC service. Networks need to be able to support a mix of devices and services in the same deployment. In case the device in question is in fact a complex thing with different sets of sensors and actuators that communicate through the same interface, networks may also need to support a mix of services, i.e., different types of traffic, from the same device. This could be for instance, a robot with a video camera for monitoring purposes (mobile broadband traffic stream) and manipulation arm (URLLC traffic), or a harbour straddle crane with remote control functionality.

In order to enter vertical industrial markets, it is necessary to address different use cases described above and answer a few crucial research questions regarding devices. How to combine devices with different URLLC requirements, how to combine different URLLC streams within a device and how to combine non-URLLC streams with URLLC streams within a device. How to monitor QoS metrics within the device and timely send this information to the BS or to the network controller? How to ensure redundancy within the device (UEs, carriers, etc)?

Finally, the device is not an isolated part of the network, especially if it has high processing capabilities. Rather, the device is part of the system and may host system functions, e.g., part of the edge cloud, or application of federated machine learning algorithms, what may be beneficial both from computational and privacy points of view.

Distributed Cloud

The following discussion introduces the concept of a distributed cloud designed specifically to meet the requirements of industrial scenarios—the industrial cloud. Moreover, an information management system is described that is able to collect, store, and manage large amounts of data from the manufacturing site. Access to stored information is handled through a well-defined API that allows developers to focus entirely on what to do with the data rather than trying to figure out how to get hold of specific data of interest.

For traditional (IT) centralized computing in the cloud offers many benefits over local hosting. The technical merits include ubiquitous on-demand access to compute resources (CPU, storage, network, applications, services), elasticity (scaling in and out of resources), and metering (monitor and pay for actual use). The service provider's resources are pooled to service multiple consumers simultaneously. By utilizing remote hardware that is deployed, managed, and maintained by a service provider much work can be offloaded from the local IT department. All these properties translate into a lower overall cost for the individual consumers.

A centralized cloud model has many advantages but does not solve all industrial requirements. There are two major problems to consider. First, signaling over large distances add to the overall latency. For (hard) real time processes with strict timing constraints, the round-trip delay to the cloud may be detrimental to the performance or even make certain use cases impossible to implement. Delay jitter may also become a big problem as the communication to and from the cloud can involve many external links over which little control is possible. Second, the computational tasks related to industrial production tend to put quite strict requirements on availability, robustness, and security. Even though cloud-native applications and services can and should be designed and set up in redundant and failsafe way, the communication is not easily guaranteed. For instance, fiber cables may break due to construction work, routing tables can become corrupted, and power outages happen. Regardless of the reason, any interruption of the network connectivity might become catastrophic for the production. In particular, anything relying on a closed-loop control algorithm that is executed in a central cloud must be made such that communication losses are handled with great care. Whether that means on-site replication of the control algorithm, a graceful degradation, or something else has to be decided case by case.

To mitigate the problems described above while still retaining the benefits of cloud computing, a distributed approach is proposed. The principle is depicted in FIG. 25 . Basically, a central cloud (also known as a data center) is connected to several other compute instances at physically different locations. These peripheral instances might have quite different capabilities with respect to processing power, memory, storage, and bandwidth available for communication. Typically, the applications are also distributed to run different parts of them on separate hardware. Used in connection with manufacturing, this system is called the industrial cloud. Another notion that is often used synonymously for distributed cloud is edge cloud. However, the term “edge cloud” may also be used to refer specifically to cloud resources located in base stations. Clearly, as shown in FIG. 25 , an industrial cloud scenario is more general and also spans over locations other than base stations.

The functional requirements (i.e., the specified behaviors and what to do) and non-functional requirements (quality attributes relating to the system's operation) determines where to deploy certain tasks. Keeping data close to where it is used is advantageous for time constrained tasks. In other use cases bandwidth restrictions might necessitate temporary storage where the data is produced. Consequently, there is need for local (on-site) compute and storage resources. However, there are also plentiful of less time critical tasks that are better handled in the central cloud. For instance, predictive maintenance and anomaly detection often depend on long and complete time series of log and sensor data. Storing this information in the data center simplifies a posteriori analysis and training of deep learning algorithms.

Real-Time Manufacturing Software Platform

On-site edge cloud deployments are seen as enablers for new and improved applications that reduce cost of deployment and management, including the possibility of parts of equipment to be replaced by software-only solutions. A typical example is the robot controller, which in existing legacy deployments is a hardware box, essentially an industry-grade PC, installed next to each manufacturing robot. This equipment is responsible for real-time control of the robot, like motion control, requiring millisecond-scale control loops. The first step in cloudification of such brownfield technology is the movement of the software from the controller to the on-site cloud, thus simplifying the installation by removing the extra hardware element.

The next step towards a fully software-defined factory is decomposing the functionality of today's software controllers into more fine-grained functions to take advantage of per-function reliability, scaling, state data externalization, and ease of management like updating and version control that are the benefits of executing programs in the cloud. Each such function encapsulates a specific part of the overall domain-specific program that makes up the actual business logic controlling each manufacturing process, and ideally, they are reusable across different such programs. In the 5G manufacturing context, the programs are envisioned to be developed in and run on top of the Manufacturing Software Platform (MSP) which provides commonly used functionality, like object recognition, motion control, or real-time analytics in a function-as-a-service (FaaS) way, reusing the concepts, tool set, and experience of the web-scale IT industry. The provider of the MSP enables high flexibility and programmability of the physical devices via components that stack on top of each other and provide an increasing level of abstraction of reality. Such abstractions are both used on detection/sensing/input and when commanding/actuating/output.

Such high-level concepts are for example observations synthetized from low level sensory input, often combining information from several sources. For example, “unit #32 has reached its destination” is a trigger that can be calculated from indoor localization triangulation, a database of destinations and maybe camera verification. Each piece of raw input is likely to be processed first in an input device specific component, such as localization system or image recognition system. Using the result of these higher-level components may correlate the AGV location to end up with more precise coordinates. Finally, even higher-level of components may correlate it with the target database and the overall goal of the system. Processing the input is thus done by the stack of components each raising the level of abstraction by a little and adding more context.

Similarly, high-level commands are like ones that would be given to a human worker, such as “hand over this object to that robot,” “paint it white,” or “drill 2 holes there.” The exact procedure to carry out such commands is then calculated by the stack of components going down, from task scheduling, trajectory planning, motor control all the way to raw commands to servos.

This approach eventually allows programming of manufacturing processes using high-level concepts humans easily understand, simplifying or hiding the complexity the distributed nature of cloudified applications altogether. It also supports re-use and saves on development time, as the low-level components are likely application agnostic and can be used in many contexts, whereas high-level components are easier to develop them working with high-level concepts.

Both the execution environment and the MSP platform can be added value provided by components in and connected to the 5G network, especially if they are bundled with connectivity solutions, both wired and wireless, to provide a strong concise industry control vision. For this to happen, an ecosystem of robotics vendors and manufacturing companies has to be on-board and use such components. Collaboration at an early stage is essential.

Data/Information Management

To take care of all data produced within an industrial plant, an information management system is needed. Important characteristics for such a system is that it is distributed (for robustness and accessing data where it is needed), scalable (doing this for one or one hundred machines should have the same degree of complexity), and reusable (adding data management of yet another manufacturing site to an existing instance should be simple), and secure (honor confidentiality and privacy, ensure data integrity, provide means for data ownership and access control). The task of this system is to collect, manage, store, filter, extract and find data of interest. Clearly, the system must cater for different types of data (e.g., time series, streaming data, events of interest, alarms, log files, et cetera) with quite different requirements on time to live, latency, storage and availability, bandwidth and so forth. Moreover, it must handle a mixture of both sensitive and open data. Storage requirements for data varies, but a solution based on the concept of a distributed cloud with “safe” storage is needed to cope with the wide range of different requirements that is anticipated. The safety aspect includes privacy concerns and implementation of access rights, both in-flight (i.e., while data being in transfer) and in storage.

A rich set of production data is the basis for all further processing and analysis. Collecting more data facilitates new use cases with respect to planning, flow control within the production, efficient logistics, predictive maintenance, information sharing, control and actuation of individual machines, anomaly detection, quick response to alarms, distribution of work orders, remote monitoring, daily operations, and much more. The more data being collected, the more challenging become the task of managing it. For a large industrial site, the total number of sensors and actuator that can be read, monitored and controlled can easily exceed 10 000. The sampling rate varies a lot, but over time the aggregated amount of collected data becomes substantial. Even finding the data of interest tend to become problematic.

Production is often less static than it appears to be. Clearly, changes in settings or a slightly different set of work stages might be needed for product variations concerning shape, material, size, surface polish, placement of drill holes, et cetera. Moreover, the same set of tools and machines can be used for entirely different products in different production batches. When a new product is to be manufactured it could even require a completely new production line to be set up. Variations in production will have an impact on what data to look at with respect to operation and analysis. As new sensors and actuators are utilized, the data management must be able to adapt to changed conditions.

Often the same data can be useful for multiple purposes (e.g., both for the monitoring of productions and for quality assurance after the product is finished), and, as discussed above, entirely new parameters become of interest when production changes. When sensor data is collected it is advantageous to annotate it with additional information (a.k.a. metadata) for future use. A simple example of this is adding a timestamp to every sensor reading, something that not always exist from start. Other useful metadata are information about location, product id, specifics about used tools, and/or batch number. In general, this kind of metadata simplifies searching and improves on traceability. Specifically, it can be used for filtering and extracting particular information that is needed for analytics and machine learning purposes.

Some sensor data that is collected may be used for other things than the industrial process that is being run in the factory. For instance, it might be readings that relate to monitoring the condition or status of certain equipment that is used in production but is owned by someone else. The owner is interested in monitoring the equipment to plan maintenance and service, but also to collect statistics for improving future generations of the equipment. This data can be sensitive and should not be visible to the factory owner. On the other hand, the factory owner may not want to reveal data that relates to the quality or quantity of the products leaving the production line. Consequently, there is a need to define ownership of and provide means to restrict access of data only to authorized parties. The information management system should cater for this while still handling all data in the same way regardless of its purpose or two whom it belongs.

FIG. 26 illustrates a typical manufacturing scenario. On the left hand side, the factory is depicted, while the right hand side represents the data center (i.e., the central cloud). Connected tools, machines, and sensors produces data that are annotated and forwarded for processing and storage. A “global” device registry keeps track of all available producers (sensors) and consumers (actuators). Applications obtain information on where to find needed data by asking the device registry. Storage is being taken care of both on-site and in the data center, as is provenance (more on that later). This design allows for both on-site (low-latency) and off-site based control applications. Clearly, this set-up can be replicated in case multiple production sites are to be included.

The set-up is an example of a distributed cloud where data is handled both in the factory and in the central cloud. With the obvious exception of available resource capacity, the local set-up and its functionality can be made very similar to the corresponding set-up and functionality of the data center. Doing so will greatly simplify deployment, operation, and life cycle management of the applications running at both locations.

In addition to annotating, storing, and processing data, an information management system must also handle data provenance. In short, this is the process of keeping track of data origin, where it moves over time, who uses it, for what purpose, and when it is used. Keeping records of these parameters facilitates auditing, forensic analysis, traceback off and recovery from use of erroneous data series. Provenance gives the administrator of the information management system a way to obtain a detailed view of data dependencies and derived results. For instance, a faulty or uncalibrated sensor might go unnoticed for some time if it does not cause immediate havoc in production. Then, if its sensor data is used for training purpose in a machine learning algorithm, the resulting model might become flawed which will negatively impact its usage. With proper provenance in place, it is possible to find out where and when the potentially flawed model has been used and take appropriate actions to mitigate problems caused by this.

In order to simplify for developers, it is important that the information management platform provides a well-defined API to find and access all data. This is true both for “raw” sensor data that is collected in real time, and for historical records of older data. In particular, it can be noted that the distributed cloud model implies that data of interest can be stored at geographically different locations and its placement can vary over time. This fact follows from different needs (e.g., tolerance for variations in latency), overall robustness (e.g., handling link failure to the data center), and requirements on long term availability. Applications that use the data should not need to keep track of the storage location themselves; the underlying information management platform does this allowing developers to focus on more important things.

A prototype of an information-management system is now being deployed at one of SKF's roller bearing factories in Gothenburg. This work is part of the SGEM II research project that is running June 2018-September 2019. The software is based on both open source projects (e.g., Calvin for handling data flow from factory to data center, and Apache Pulsar and VerneMQ for pub-sub messaging handling) as well as internal proprietary code. The information management platform is for data what Kubernetes is for containers. Clearly, not all functionality is in place yet, but we iterate and update frequently. The work is done using a modern continuous integration/continuous deployment development methodology. This means changes to the code will be automatically tested, and deployment to the distributed system can be made with a single command. The overall design is deliberately made such that most system updates can be done without interrupting the running applications. Thus, production does not have to be stopped for deploying software updates to the platform. This property is particularly important at the factory site as updates can then be made also outside of the scheduled maintenance windows for the production site. Usually a production stop is very expensive for the manufacturer which means planned maintenance windows are very few and separated with in time as much as possible.

A distributed cloud retains all properties of a central cloud such as elasticity, on-demand compute, resource pooling, measured services, and network access. In addition, the ability to place processing closer to where results are used facilitates more robust solutions, decentralization, and implementation of low-latency use cases.

With an adequate information management system in place, developers can build new applications and access data produced at the factory without physical access to the manufacturing site and without detailed knowledge of how data is collected or where it is stored. Different types of data are handled and stored both on-site and within the data centre. A well-defined API exposes services and allows for efficient searching and filtering based on any parameters and metadata that is available. Access rights to data can be defined based on user and/or role of said user. Advanced logging features facilitate audits and traceability of the usage of the collected data.

Operations and Management

The term operations and management (O&M) refers to the act of operating and managing the network and the devices in a factory deployment. Operations Support System (OSS) refers to the software used to accomplish this task.

The factory floor consists of machinery used to produce and manufacture goods. The machines are organized often into an assembly line through which the goods flow with or without human intervention and depending on the level of automation. The different tools and machines used for the production may or may not be connected. If connected, typically some kind of data is gathered from the machinery to either for predictive maintenance of the tools and machinery itself or to aid in the quality assurance process of the goods being manufactured. This is called the operational technology (OT) part of the factory floor.

Most enterprises, factories included, also have communication infrastructure in place for the work force comprising of wired and wireless communications (typically Ethernet and Wi-Fi), computers, mobile phones, etc. This equipment is used to access Intranet and internet, email, and other typical office applications. This is called the information technology (IT) part of the factory floor.

The merger of OT and IT has been identified as an emerging trend. In practice, this means a single interface to operate and manage both the devices, connectivity, the data generated by these devices, and the network infrastructure in a factory. Research questions related to the OT/IT merge in factories include:

What kind of device management protocols are used for the OT and can those interface to the IT system? What kind of platform is needed to handle all the different aspects?

Digital twin concept is very popular in the industry setting. Here the idea is to bring the data gathered to a digital data model of the physical asset or the whole factory and then apply analytics on the data to predict, describe and prescribe the past, current and future behavior of the asset or process. Research questions around the digital twin concept include:

How to model the physical assets? What data is relevant to capture and for how long? What kind of latency is required for real time interaction and how to provide that? What kind of models are needed to predict possible futures? Where to compute and what kind of compute capabilities are needed to perform meaningful prediction?

All of this should be achieved with an easy to use system that can bring increased reliability and availability, reduce risks, lower maintenance costs and improve production. Solutions where an operator exposes/delegates only a small portion of its O&M to its customer may be desirable. The customer should get a simple interface. Solutions should be possible to scale down to just a handful of devices, such that even households can use it.

Finally, augmented reality and virtual reality in conjunction with the digital twin ideas may have a large impact on future network management in the merge IT and OT space. Equipment management may be done remotely with the feel of being present in the same space. Also, technical documentation and guidance on equipment use or repair can be provided remotely to a person on site through smart glasses, tablets, etc.

Time-Sensitive Networking

Following a general idea and initial overview of Time-Sensitive Networking (TSN), where the material presented will help to get a good starting point in TSN. Also provided are certain details of 5G-TSN integration.

TSN is envisioned to improve wired IEEE 802.3 Ethernet communication, to enable support for the very demanding domain of industrial applications (and others). TSN stands for Time-Sensitive Networks (or Networking). It is an ongoing IEEE standardization initiative by the TSN task group. They define TSN as a set of individual features. Most TSN features are extensions to the IEEE 802.1Q standard. A TSN network comprises Ethernet end stations (sometimes also called end points), Ethernet cables and bridges (also called switches). An Ethernet bridge becomes a TSN bridge if it supports a certain (not-defined) set of TSN features.

The different features in the TSN standard in general aim at:

zero packet loss due to buffer congestion (usual Ethernet bridges indeed drop packets if buffers are filled) extremely low packet loss due to failures (equipment, bit errors, control plane etc.) guaranteed upper bounds on end-to-end latency low packet delay variation (jitter)

Communication in TSN happens in TSN streams. One specific feature in TSN, to give an example, is that streams are subject to an agreement, as arranged between the transmitters (end stations called Talkers) and the network till the receivers (end stations called Listeners), that ensures low latency transmissions without unforeseen queueing.

In the following, TSN is introduced from a high-level perspective. Afterwards are technical details of what a TSN and 5G interworking will look like and how certain TSN features can be supported in 5G.

The TSN standardization arose from a standardization initiative that was found to define an Ethernet-based communication standard for Audio and Video communication called Audio-Video Bridging (AVB). TSN is based on AVB and is enhanced with features to make it suitable for industrial usage. Up to now, the TSN community focuses on the following industrial use cases:

Industrial communication for factory automation (major use case #1)

Shopfloor TSN links (horizontal) Shopfloor to cloud TSN links (vertical) Intra-machine communication TSN for the factory backbone

Intra-vehicle communication (major use case #2) Electrical power generation and distribution (Smart Grid use cases) Building automation (no practical examples on this found so far) Fronthauling (according to IEEE P802.1CM)

In this document, the use in industrial communication for factory automation is addressed, although some of the detailed techniques and concepts may be applicable to other use cases.

FIG. 27 illustrates a hierarchical network architecture in a factory. Shop floor TSN links (horizontal) appear within production cells, connecting devices or machines and controllers. The production line areas enable the connection between the Operational Technology (OT) domain and the Information Technology (IT) domain, but are as well used to connect production cells on the shop floor, if necessary. In the TSN categorization we introduced above, the first one (OT-IT) is obviously based on shop floor to cloud TSN links (vertical) and the latter again on shop floor TSN links (horizontal). TSN used for intra-machine communication is in so far different from horizontal shop floor TSN link, as this is probably a TSN network deployed by a single machine vendor inside for example a printing machine or any other machine tool—from a 5G perspective it is less likely that these horizontal links need to be addressed. TSN for the factory backbone is used in the factory/building/office network (light-orange area). If deterministic communication from virtualized controllers is desired, for example, TSN is necessary end-to-end down to the shop floor.

TSN communication is another kind of packet service that is based on a best effort Ethernet packet network but enhanced though TSN features. An agreement is used between devices involved in communication, to achieve determinism. The agreement limits the transmitter of a TSN stream to a certain bandwidth and the network, in return, reserves the needed bandwidth, reserving buffering mechanisms and scheduling resources. The resources can be exclusively used by the specific stream. Compared to other packet services such as CBR (Constant Bit Rate) and best effort type of packet services, certain observations can be made.

A best effort packet service is perhaps the most known one, where packets are being forwarded and delivered as soon as possible. There are no guarantees, however, on the timely delivery of the packets. The end-to-end latency and the variation of the latency is rather large, and thus a statistical language is preferred to express the overall performance (loss, end-end-latency and jitter). The top part of FIG. 28 shows the typical performance of a best effort packet service network. The typical tail in end-to-end latency causes a problem for most industrial use cases.

On the contrary, there is also the CBR packet service that offers fixed latencies and jitter (latency variation) close to zero as seen in the application layer. CBR is typically offered by multiplexing in the time domain with typical examples being SDH (synchronous digital hierarchy networks) or OTN (optical transport networks). Typical performance of CBR can be seen in the middle part of FIG. 28 . A drawback of CBR is that it is very inflexible in the way network resources are shared. So, it is hard to adapt to different application needs, for example in terms of latency or bandwidth—but of course in industrial context the requirements are manifold, and a single network to server all is desired.

TSN aims at supporting all type of traffic classes (Quality of Service (QoS) and non-QoS) over the same infrastructure. Therefore, the TSN network sits somewhere between a CBR and a best effort type of packet service, where the latency is typically larger compared to a CBR network, but the latency variation and the jitter are bounded—no tails. In other words, TSN offers a guarantee that the network will not perform worse than a specific agreed end-to-end latency and jitter, as seen in the bottom part of FIG. 28 . These guarantees can be flexibly adapted. This behavior is required by most industrial applications.

A core feature of TSN is the “stream concept,” where a stream comprises dedicated resources and API. A TSN stream can be seen as the unicast or multicast from one end station (talker) to another end station or multiple end stations (listener(s)) in a TSN capable network. Each stream has a unique StreamID. The Stream ID is created of talker source MAC address and a unique stream identifier. Bridges will use the StreamID plus the priority code (PCP) field and VLAN ID (VID) that is included inside an 802.1Q VLAN tag in the Ethernet header for internal frame handling. In that sense, TSN streams are standard 802.1Q Ethernet frames that are given more privileges than regular Ethernet non-TSN frames. Before a talker starts sending any packet in a TSN stream, the specific stream has to be registered in the network and certain TSN features to be configured. Next to TSN streams with guaranteed QoS, also best-effort traffic can be sent in a TSN network by peers—but of course without or just limited guarantees on QoS. TSN streams are sent in TSN domains. A TSN domain can be seen as a continuous domain, where all devices are in sync and continuously connected through TSN capable ports. a TSN domain is defined as a quantity of commonly managed devices; the grouping is an administrative decision.

Stream management is defined in IEEE 802.1Qcc, Qat, Qcp and CS. It defines the network discovery and the management of network resources and TSN features in a network as for example the creation of the required protected channels for TSN streams. Moreover, stream management offers users and network administrators functions to monitor, report and configure the network conditions. In TSN, there are three configuration models: a distributed, a centralized and a fully centralized one. In the latter two models a Central Network Controller (CNC) is used similar to a Software Defined Networking (SDN) controller to manage TSN switches. In the fully centralized model, a Central User Controller (CUC) is used in advance as a central interface for end stations and users. In the distributed model, there is no central control, so bridges and end stations need to negotiate on TSN requirements; in this model some TSN features that require a central instance for coordination are not applicable. A lot of TSN features also aim at a common protocol and language standard for interactions between CNC/CUC, end stations and bridges (i.e. YANG, Netconf, Restconf, LLDP, SNMP etc.).

Time synchronization is used to establish a common time reference that is shared by all TSN enabled network entities. The time synchronization is based on the exchange of packets containing time information as defined in IEEE 802.1AS-rev; it defines some amendments to the Precision Time Protocol PTP, widely used in industrial context, that is then called gPTP (generalized PTP). gPTP is an advanced version of PTP in a sense that it supports also redundant grandmaster deployments as well as also the establishment of multiple time domains in a single PTP network and some other enhancements and also restrictions to the broader PTP. The ambition of gPTP is to achieve a sub-microsecond accuracy in synchronization. The precise time synchronization is used for some TSN features (for example IEEE 802.1Qbv), as well as provided to applications relying on a common notion of time (like distributed motion control).

Stream control, which provides for bounded low latency, specifies how frames belonging to a prescribed TSN stream are handled within TSN enabled bridges. It enforces rules to efficiently forward and appropriately queue frames according to their associated traffic classes. All existing stream controls follow similar principles, namely, certain privileges are associated with TSN streams while frames not from prioritized TSN streams might be queued and delayed. Relevant features for industrial networking are IEEE 802.1Qbv (introduces “time-gated queueing,” i.e. time-coordinated handling of frames) and IEEE 802.1Qbu plus IEEE 802.3br for frame pre-emption. 802.1Qbv relies on precise time synchronization and is only applicable if a CNC is used to schedule frame forwarding in bridges similar to a time-division multiplexing manner. Using Qbv, a CNC tells each bridge alongside a path in the network exactly when to forward frames. An alternative to Qbv is Credit-Based Shaping (802.1Qav) originating from AVB, likely not used for strict industrial use cases as it is not as deterministic. An additional feature called Asynchronous Traffic Shaping (802.1Qcr) is in an early stage of development. An argument against Qbv, which is maybe the best suited to achieve guaranteed latency bounds, is the complexity it requires in terms of scheduling and time synchronization. Qbv and frame preemption (Qbu and br) can might be used separately or also combined.

Stream integrity is important for providing ultra reliability. Apart from delivering packets with ultra-low latency and jitter, TSN streams need to deliver their frames regardless of the dynamic conditions of the network, including transmission errors, physical breakage and link failures. Stream integrity provides path redundancy, multipath selection, as well as queue filtering and policing. One main feature therefore is IEEE 802.1CB, including Frame Replication and Elimination Redundancy (FRER).

A visual summary of the TSN features described above is given in FIG. 29 .

Interworking between 5G and TSN is discussed here. Since both systems provide diverse ways for ensuing QoS and for network management, new solutions are required. The basic idea according to some of the techniques described here is that the 5G-System (5GS) adapts to the network settings of the TSN network. It should be noted that ongoing TSN standardization defines a set of features. and not all features need to be supported for every use case. Announcements about which set of TSN features are relevant for which use cases have not been done yet. An ongoing initiative to address this issue is the joint project IEC/IEEE 60802: “TSN Profile for Industrial Automation”. It is under development and updated frequently. Publication is planned for 2021.

Real-time Ethernet is one of the established wireline communication technologies for vertical applications. For wireless communication technologies, 3GPP TS 22.104 specifies 5G system requirements to support real-time Ethernet. When some sensors, actuators and motion controller are connected using a 5G system and others are connected using industrial (i.e. real-time) Ethernet, the interconnection between real-time Ethernet and 5G is realized using gateway UEs connected to Ethernet switches or a device is connected directly to a data network using an Ethernet adapter.

Potential baseline system requirements are:

The 5G system shall support the basic Ethernet Layer-2 bridge functions as bridge learning and broadcast handling The 5G system shall support and be aware of VLANs (IEEE 802.1Q) The 5G system shall support clock synchronization defined by IEEE 802.1AS across 5G-based Ethernet links with PDU-session type Ethernet. The 5G system shall support for TSN as defined by IEEE 802.1Q, e.g. IEEE 802.1Qbv (time-aware scheduling) The 5G system shall support coexistence of critical real-time traffic following a time-aware schedule and non-TSN lower priority traffic.

A TSN network consists of four types of components: bridges, end stations, network controller(s) and cables (minor notice: it is common in industrial context that end stations are switches as well to enable daisy-chaining and ring topologies for example). A 5G network will in most cases need to act like one or more TSN bridges if a seamless integration into a TSN network is envisioned. Therefore, in many cases the 5G network will take part in the TSN network configuration as a usual TSN bridge.

FIG. 30 illustrates a baseline architecture in a factory network, where TSN components are used on the shop floor, as well as in the factory backbone TSN. 5G is used to replace the shop floor to cloud (vertical) connection (5G for vertical TSN links). In general, a shop floor TSN as illustrated in FIG. 30 might be at least a single TSN-capable end station without any TSN switch. Talker and listener(s) might appear on both sides (UE and UPF) of the 5G network. The 5G network is used to connect or merge both TSN domains. Wireless Access Points or 5G Base Stations may be used to connect TSN domains. A CUC and CNC in FIG. 30 are deployed on the factory backbone-side, although they might well be implemented on the shop floor, for example as part of an intra-machine TSN network.

Connecting two TSN domains on the same shop floor (5G for horizontal TSN links) is one possible scenario. In this case, the 5GS replaces a single hop on the shopfloor. Because NR does not presently support a device-to-device (D2D) capability, this would be a two-hop (UE A-gNB/Core-UE B) connection in 5G.

For TSN used inside machines (intra-machine communication), an interworking with 5G is obviously of less relevance as introduced above. Two nodes inside a (possibly metallic) machine will likely not rely on a central connection to a 5G base station to communicate wirelessly. A typical example, is a printing machine where different motors must be controlled very precisely to achieve an accurate result.

A further option is that a legacy 5G device (i.e., a device without TSN feature support, or maybe not even an Ethernet device) is connected to a 5GS that is connected to a factory backbone TSN network. As the 5G device is not aware of any TSN features or capable to support them itself, the 5GS might act as a virtual endpoint that configures TSN features on behalf of the 5G device to be able to communicate to a TSN endpoint with seamless QoS end-to-end. A virtual endpoint function could be part of a UPF in the 5GS. From a TSN network point of view it looks like the virtual endpoint is the actual endpoint—the 5G endpoint is covert. FIG. 31 illustrates the conceptual way of working, showing how virtual endpoints may be used to connect non-TSN devices to a TSN network using 5G. In the figure, “UP” refers to “user plane,” while “CP” refers to “control plane.” This concept may be referred to as “Application Gateways”.

Some TSN features introduce challenges to the 5GS. In the following, it is highlighted how some key TSN features might be supported by the 5GS, to enable a seamless 5G TSN interworking.

Network-Wide Reference Time (IEEE 802.1AS-Rev)

In TSN, reference time is provided by the IEEE 802.1AS-rev synchronization protocol that allows local clocks in the end stations and switches to synchronize to each other. More specifically, the so-called Generalized Precision Time Protocol (gPTP) described therein employs a hop-by-hop time transfer between the different TSN capable devices of the network. The protocol supports the establishment of multiple time domains in a TSN network and a redundant grandmaster setup as well as other features. A 5GS should be able to take part in the gPTP processes, allowing the same clock accuracy and synching capabilities as in TSN. The gPTP processes must always run periodically to compensate clock drifts. The clock information received by the 5GS over cable from a grandmaster in the TSN network need to be carried over the air from a basestation (BS) to a UE or maybe as well the other way around. Different options are currently discussed how that could be done and it is an ongoing topic in standardization. In the following and in general, a grandmaster is a device that carries a source clock used for gPTP.

A simple example of TSN time synchronization across a 5G network is illustrated in FIG. 32 . A grandmaster's time signal is received in the 5GS at the side of the UPF and send over the air by a BS. The UE forwards the time signal it receives from the BS to Device 1 (“Dev 1,” in the figure). Device 1 might need the time signal of the grandmaster to be able to communicate with Device 2 (“Dev 2, in the figure).

Internally, the 5GS might use any signaling not related to gPTP to carry the grandmasters time signal. In that case the ingress points in the 5GS (at UE and User Plane Function (UPF)) need to act as gPTP slaves. They get synched themselves to the grandmaster from the gPTP signals arriving and forward that notion of time on the RAN. Of course, the requirements on time synchronization accuracy are defined by the application and need to be satisfied. In LTE Release 15, a signaling mechanism for accurate time synchronization with sub-microsecond accuracy has been introduced and might been reused for NR.

For industrial use cases the support of multiple time domains might be relevant, as depicted in FIG. 33 and FIG. 34 . One time domain might be a global one, such as the Coordinated Universal Time (UTC). This time domain might be used by applications to log certain events on a global time base. Furthermore, additional time domains might be used based on local clocks, i.e., clocks that are based on an arbitrary timescale and don't have a certain defined start-point epoch (e.g., a clock at a grandmaster that started at the boot up of the device instead of referring to a global clock timescale). This local clock might have a much higher precision than the global clock. It is distributed from a grandmaster to a few other devices and used on the application layer to coordinate very accurately synchronized actions or for example for timed communication as defined in 802.1Qbv. To support multiple time domains in the 5GS, one possible way of implementation is to establish a common reference time between all gNBs and UEs, for example using the UTC timescale, and then based on that, transfer individual time domain signals in the 5GS only to end-stations requiring that specific time domain. For transmission of individual local time signals it is possible to use timestamping from the common reference time or transmit offsets periodically that are referenced to the common reference time. In addition, it might also be possible that a forwarding of gPTP frames is done transparently through the RAN by using a similar timestamping mechanism.

The concept to use a common reference time to support multiple other time domains in general is illustrated in FIG. 33 . In this figure, the clock in the 5G time domain depicts the common reference time, while the clocks in the TSN work domains are local clocks that need to be forwarded to some UEs over the 5GS. Based on the timestamps done using the common reference time at UE and UPF, it would be possible to correct the times inside gPTP packets (belonging to a TSN work domain clock) to account for varying transmit times in the 5GS. Only a subset of all gPTP frames arriving at the ingress might need to be transported across the 5GS, like for example Announce (config) frames and Follow-Up (carry timestamps) frames. Other frames could be consumed at the 5GS ingress and not forwarded. At the egress the 5GS need to act like a gPTP master in any case. To detect and differentiate time domains, the domain Number field in the gPTP header of each frame can be used. There are some efforts necessary to identify which UE needs to be synched to which time domain. A recent research activity has addressed this issue.

In FIG. 33 the Application Function (AF) in the 5GS is used as an interface towards the CNC in the TSN network—in one possible way the CNC might provide information to the 5GS about how time domains need to be established, i.e., which UE needs which time domain signal.

Timed Transmission Gates (IEEE 802.1Qbv)

The TSN feature IEEE 802.1Qbv provides scheduled transmission of traffic controlled by transmission gates. Each egress port in an Ethernet bridge is equipped with up to eight queues and each queue with a separate gate. This is illustrated in FIG. 35 .

Ingress traffic is forwarded to the queue at the egress port it is destined to; the egress queue is for example identified by the priority code point (PCP) in a frame's VLAN-header field. A regular cycle (“periodic window”) is established for each port, and at any particular time in that window, only certain gates are open and thus only certain traffic classes can be transmitted. The queue coordination is done by the CNC. The CNC gathers information about the topology, streams and also individual delay information from all switches and creates a Gate Control List (GCL). The GCL controls the timing of the opening and closing of queues at each switch, but not the order of frames in the queue. If the order of frames in the queue, i.e. the queue state, is not deterministic, the timely behavior of the two streams may oscillate and lead to jitter for the overall end-to-end transmission. By opening and closing gates in a time-coordinated manner it is possible to achieve a deterministic latency across a TSN network, even if indeterministic best-effort type of traffic is present on the same infrastructure. Best effort traffic is simply held back by closing its queue and let priority traffic pass from another queue. It is important to mention that a timely delivery does not just mean to not sent a frame to late from one bridge to the next but also prohibits to send it too early as this might lead to a buffer congestion at the consecutive hop.

The 5GS should be able to transmit frames in a way that the 802.1Qbv standard expects, i.e., according to a GCL created by a CNC in case the 5GS acts as one or multiple TSN switches from a TSN network perspective. This means keeping specific time windows for ingress and egress TSN traffic at UE and UPF respectively. So, data transfer in the 5GS has to happen within a specific time budget, to make sure that the packets are forwarded at the configured point in time (not earlier or later) to the next TSN node in both uplink and downlink. As the biggest part of the latency in the 5GS is probably added in the RAN it seems reasonable to use the timing information from a CNC at gNBs to improve the scheduling of radio resources. It might be possible to exploit the information about transmission timings according the Qbv scheduling for an efficient scheduling of radio resources at a BS using mechanisms like configured grants and semi-persistent scheduling. As a BS anyhow needs to be time aware to be able to forward the time signal to UE(s) it just might need to get aware about the transmission schedules in advance. The Qbv mechanism ensures frames arrive at the 5GS from the TSN network with minimum jitter.

The Application Function (AF) in the 5GS might be an option to interface the CNC. There, a topology could be announced, as well as latency figures could be provided to the CNC as if the 5GS would be a regular TSN switch or any TSN switch topology. The AF then could also accept a time schedule from the CNC and translate it into meaningful parameters for the 5GS to support the time gated queuing happening in the external TSN network. It is important to understand that in the current way the CNC is specified it will accept only fixed numbers to define the delay that is added through a typical TSN switch. Therefore, some new methods are required to allow also the 5GS to be a more “flexible” TSN switch regarding the latency numbers that need to be reported to the CNC.

One way of achieving a timely delivery of packets might involve the use of playout-buffers at the egress points of the 5G network (i.e. at a UE and the UPF for downlink or uplink). Those playout buffers need to be time aware and also aware of the time schedule used for Qbv and specified by the TSN network's CNC. The use of playout buffers is a common way to reduce jitter. In principle for downlink for example the UE or any function following the UE will hold back packets until a certain defined point in time has come to forward it (“play it out”). Same would be possible in uplink, probably in the UPF or after the UPF as an additional function for TSN traffic.

Frame Preemption (IEEE802.1Qbu)

The IEEE 802.1Qbu amendment, “Frame Pre-emption”, and its companion IEEE 802.3br, “Specification and Management parameters for Interspersing Express Traffic,” add the capability of interrupting a frame transmission to transmit a frame of higher priority. Because they do not have to wait for the lower-priority transmission to fully finish, any express frames have shorter latency. The eight priority levels are split into two groups: express and preemptable. The queues assigned to priority levels belonging to the express group are referred to as express queues. The transmission of the pre-empted frame resumes after the express traffic is finished, and the receiver is able to reassemble the pre-empted frame from the fragments.

The 5G network already supports pre-emption techniques with the existing mechanisms. Whether there is additional effort needed to fully support frame pre-emption is not clear yet. It should be noted that there is an important difference between IEEE frame pre-emption and 5G pre-emption techniques. IEEE frame pre-emption is just interrupting transmission and after forwarding express frame(s) the pre-empted frame transmission is continued. There is no retransmission.

Frame Replication and Elimination for Reliability—FRER (IEEE 802.1CB)

The IEEE 802.1CB standard introduces procedures, managed objects and protocols for bridges and end systems that provide identification and replication of packets for redundant transmission. One of these procedures is Frame Replication and Elimination for Reliability (FRER), which is provided to increase the probability that a given packet will be delivered—in case one Ethernet plug is removed for any reason, or a cable is cut accidentally the communication should continue.

FIG. 36 illustrates some of the basic features of FRER. Some of the important features of FRER are:

Appending sequence numbers, to packets originating from a source, or from a particular stream. Based on the exact needs/configuration the packets are replicated. These creates two (or more) identical packet streams that will traverse the network At specific points in the network (typically close to or at the receiver) the duplicate packets are eliminated. Complex configurations are supported so the mechanism can support failures at multiple points in the network.

A 5GS might need to support end-to-end redundancy as defined in FRER for TSN as well, for example by using dual connectivity to a single UE or also two PDU sessions to two UEs deployed in the same industrial device (can be called “Twin UEs”). Anyway, redundancy in the 5GS might not base on exact the same principles as a TSN network (which means complete physical end-to-end redundancy using separate equipment). The latter ones rely on fixed wired links, while 5G relies on a dynamic radio environment. Nevertheless, redundancy as defined by FRER is rather pointing at failures in the equipment (such as an error in a gNB that leads to a connection loss, etc.), but obviously also helps to overcome influences of changing radio conditions and connection losses due to handovers.

If “Twin UEs” are used they should be connected to two BSs anytime to supports full redundancy and in case of a handover, not perform it both at a time and not to the same BS.

It is an open discussion whether a physical redundancy needs to be implemented in the 5GS or whether traffic can be carried over for example a single User Plane Function (UPF) or server hardware respectively. If for example some 5GS functions are so reliable that it is not required to be deployed in a redundant way, then it might be sufficient to just use physical redundancy for some parts of the 5GS.

Some inventions have been worked on describing how this FRER type of redundancy can be supported in the 5GS, both on the RAN and on the core. As a configuration point for redundancy we also suggest using the Application Function (AF). The 5GS could announce different redundant paths to the TSN network and internally in the 5GS could support the redundancy in a way it is sufficient with or without physical redundancy of certain components. So, the actual 5G interpretation of redundancy can be hidden from the CNC/TSN definition of redundancy this way.

5G and TSN—Network Configuration

In TSN, the IEEE 802.1Qcc extension supports the runtime configuration and reconfiguration of TSN. At first, it defines a user network interface (UNI). This interface enables the user to specify stream requirements without knowledge of the network, thereby making the network configuration transparent to the user. This of course as well relevant to achieve a plug-and-play behavior as it is common for home and office networking but especially not in today's industrial Ethernet networks.

There are three models that enable this transparency. Specifically, the fully distributed model, where stream requirements propagate through the network originating from the talker till the listener. Therein the UNI is between an end station and its access switch. The fully distributed model is illustrated in FIG. 37 , where the solid arrow represents UNI interfaces for exchange of user configuration information between talkers, listeners and bridges. The dashed arrow in the figure represents a protocol carrying TSN user/network configuration information, as well as additional network configuration information.

The centralized network/distributed user model introduces an entity, called the centralized network configurator (CNC), with complete knowledge of all streams in the network. All configuration messages originate in the CNC. The UNI is still between the end station and access switch, but in this architecture, the access switch communicates directly with the CNC. FIG. 38 depicts the centralized network/distributed user model.

Finally, the fully centralized model allows a central user configurator (CUC) entity to retrieve end station capabilities and configure TSN features in end stations. Here, the UNI is between the CUC and the CNC. This configuration model might be most suitable for the manufacturing use cases, where listener and talker require a significant number of parameters to be configured. The CUC interfaces and configures end stations, while the CNC still interfaces bridges. The fully centralized model is illustrated in FIG. 39 . The following discussion provides more details for the fully centralized model, since it is likely the most suitable for the manufacturing use cases.

CUC and CNC

The CUC and CNC, at the fully centralized model, are part of a configuration agent (e.g., a PLC in a factory automation context) that executes both tasks, as shown in FIG. 40 , which illustrates a configuration agent consisting of CUC and CNC. (In the figure, “SW” refers to a switch, “ES” refers to and end station, and “UNI” refers to a user network interface.) The standard IEEE 802.1Qcc does not specify protocols to be used between CUC and CNC as shown in FIG. 40 . OPC UA (Open Platform Communications Unified Architecture) might be a possible selection for the interface between CUC and end stations, Netconf between bridges and CNC. For TSN stream establishment, a CUC will raise a join request to the CNC, as depicted in FIG. 41 , which shows interaction between CNC and CUC.

The communication between talker and listener happens in streams as introduced above. A stream has certain requirements in terms of data rate and latency given by an application implemented at talker and listener. The TSN configuration and management features are used to setup the stream and guarantee the stream's requirements across the network. The CUC collects stream requirements and end station capabilities from the devices and communicates with the CNC directly. FIG. 42 shows the sequence diagram between different entities to conduct a TSN stream setup.

The steps to setup a TSN stream in the TSN network in the fully centralized model are as follows:

1) CUC may take input from e.g. an industrial application/engineering tool (e.g. a PLC) which specifies for example the devices which are supposed to exchange time-sensitive streams. 2) CUC reads the capabilities of end stations and applications in the TSN network which includes period/interval of user traffic and payload sizes. 3) CNC discovers the physical network topology using for example LLDP and any network management protocol. 4) CNC uses a network management protocol to read TSN capabilities of bridges (e.g. IEEE 802.1Q, 802.1AS, 802.1CB) in the TSN network. 5) CUC initiates a join requests towards the CNC to configure TSN streams. CNC will configure network resources at the bridges for a TSN stream from one talker to one or more listener(s). 6) CNC configures the TSN domain. 7) CNC checks physical topology and checks if required features are supported by bridges in the network. 8) CNC performs path and schedule (in case Qbv is applied) computation of streams. 9) CNC configures TSN features in bridges along the path in the TSN network. 10) CNC returns status (success or failure) for streams to CUC. 11) CUC further configures end stations (protocol used for this information exchange is not in the scope of the IEEE 802.1Qcc specification) to start the user plane traffic exchange as defined initially between listener(s) and talker.

The 5GS Application Function (AF) is seen as the potential interface for the 5GS to interact with TSN control plane functions (i.e., CNC and CUC). The AF, according to 3GPP TS 33.501, can influence traffic routing, interact with the policy framework for policy control for 5G links and also further interact with 3GPP core functions to provide services, which can be utilized to setup and configure TSN streams in the 5G TSN interworking scenario. FIG. 43 illustrates the potential interfacing of the AF with the TSN control plane.

The FRER setup sequence stream in a TSN network is shown in FIG. 44 . A CUC sets the values of the parameters (NumSeamlessTrees greater than 1) in request to join message from CUC to CNC. A CNC then calculates disjoint trees based on this input in the path computation step. It uses management objects of IEEE 802.1CB (FRER) to configure redundant paths in the bridges.

As introduced in the FRER part above, the AF could implement the interface that signals redundancy support towards the CNC and accepts redundant path computations from it. This is illustrated in FIG. 45 , which illustrates interaction between AF, CUC, and CNC to setup FRER. Furthermore, the AF might also be used to interact with the CNC for other TSN features beyond FRER.

TSN is now in a research and development phase. Early products are available on the market that support only a subset of TSN features listed in here. Also, the TSN standardization is ongoing and some features are not yet finalized. Especially it is not clear which features will be relevant for industrial use cases and which not. IEC/IEEE 60802 is an ongoing effort to define a TSN profile for industrial usage. Nevertheless, it is a wide spread vision that TSN will be the major communication technology for wired industrial automation in the next years.

In the preceding paragraphs, the concept of the Time Sensitive Networks (TSN) was introduced and the vision of improving Ethernet communication for industrial applications was explained. Then the technical introduction provided some of the performance goals of a TSN that needs to handle not only best effort type of traffic but also critical priority streams. These critical streams require very low bounded latencies that TSN must support. This allows TSN to enable new use-cases in the area of industrial automation.

Then, more details on the TSN operating principles were provided, to explain how TSN can provide deterministic communication. The issue of integrating 5G with TSN core features, was also discussed. This integration requires the support of a specific set of TSN features from a 5G network. This feature set was explained and also some inventive techniques were described, for enabling a smooth interworking between the two networks.

Core Network

The core network is the part of the system that resides between the Radio Access Network (RAN) and one or more Data Networks (DN). A data network could be Internet or a closed corporate network. We assume the core network to be fully virtualized, running on top of a cloud platform. Tasks of the core network include: subscriber management; subscriber authentication, authorization and accounting; mobility management; session management including policy control and traffic shaping; lawful interception; network exposure functions. The 5G core network is described in the 3GPP document “System Architecture for the 5G System (5GS),” 3GPP TS 23.501, v. 15.4.0 (December 2018). FIG. 46 illustrates components of the 5G core network and its relationship to the radio access network (RAN) and the UE, as described in 3GPP TS 23.501.

In today's Mobile Broadband (MBB) deployments, core network functions are often deployed on large nodes serving millions of subscribers. The nodes are often placed in few centralized data centers, giving an economy of scale.

In 5G, many other use cases will arise besides MBB. These new use cases may require different deployments and different functionalities. For example, in manufacturing, lawful intercept and many charging and accounting function may not be needed. Mobility can be simplified or, in case of small factory sites, may not be needed at all. Instead new functions are needed including support for native Ethernet or Time-Sensitive Networking (TSN). Preferably, new functions can be added quickly without having to go through a lengthy standardization process.

For reasons of latency, data locality and survivability, a core network for manufacturing should not necessarily need to run in a large centralized data center. It should instead be possible to deploy a small-scale core network at the factory site. What is needed for 5G, and for manufacturing, is a core network that is flexible in terms of deployment and in terms of functionality.

These issues can be addressed by decomposing the user plane of the core network into a small function called a micro user plane function (μUPFs). Depending on the use case, different sets of μUPFs are re-composed into a user plane service for a subscriber. The service may change over time, and the μUPFs are hosted on execution nodes, depending on service requirements like latency. The control plane of the core network requests a service by describing it on an abstract level. A chain controller translates this high-level service description into a set of μUPFs and instantiates those μUPFs on the correct execution nodes. FIG. 47 illustrates the chain controller concept.

This approach gives flexibility in terms of deployment and functionality and can be used as a basis for use cases like manufacturing. As an important aspect of flexibility, this approach allows implementations that can scale down to very small footprints.

One core network deployment alternative for the core network in manufacturing is a local, possibly stand-alone, deployment at the factory. Another deployment alternative is to run parts of the core network at a more centralized cloud. Such cloud could be at an operator site or at some corporate site. If the core network is provided by an operator, then such deployment could give an economy-of-scale advantage. Processes for this manufacturing customer could be hosted on nodes that are also used for other customers. The same management systems may be used to serve multiple customers.

In the latter deployment, special care needs to be taken for latency, data locality and local survivability. Parts of the user plane will always need to run on the local factory cloud for latency. But the control plane may very well run remotely, since this device control plane signaling is mainly for authentication (not frequent and not time-critical), session setup (typically only once for factory devices), and mobility across base stations (which may not happen at all for small deployments).

Signaling is mainly for authentication (not frequent and not time-critical), session setup (typically only once for factory devices), and mobility across base stations (which may not happen at all for small deployments). FIG. 48 shows a high-level functional view of this deployment.

Some core network functions used for MBB are not needed in manufacturing. This imposes a requirement on a core network for industrial applications to scale down to a very minimum of features. Some new features will be needed. New features that will be required are basic Ethernet support (native Ethernet PDU sessions), and more advanced Ethernet features (for example, TSN).

It must be possible to differentiate traffic within a factory. For example, production-critical devices require a different service then “office” devices. There are several techniques to achieve such differentiation; including PLMNs, slicing, APNs or μUPF chaining.

More features can be envisioned in the following areas:

Resilience. Redundancy (multiple UEs). Data locality. Ability to access factory floor network from outside the factory.

New features for manufacturing will impact several interfaces to the core network. For example, running production-critical core network services requires a production-critical cloud to run on. Or, a network deployment with some parts running locally under the responsibility of the factory owner, and some parts running centrally under the responsibility of the operator will require changes in management systems. Furthermore, additional network exposure interfaces will be needed if the 5G (core) network system is modelled as a single logical TSN switch.

Radio Access Network

In recent years, the cellular radio access capabilities necessary for enabling support for Industrial IoT have been greatly improved, resulting in both LTE and NR becoming suitable technologies for providing this support. Several architecture options supporting reliable delivery as well as new MAC and PHY features to enable URLLC have been added to the specifications in LTE and NR Release 15. Additional URLLC enhancements are being studied for NR Release 16 with a goal to achieve 0.5-1 ms latency and reliability up to 1-10 −6 . Furthermore, improvements especially targeting support for Ethernet PDU transport and TSN by the NR RAN are envisaged for Release 16.

The following describes the specified LTE and NR URLLC features introduced in 3GPP Release 15 as well as our proposed RAN concept for NR Release 16. First, how 5G RAN architecture options may be used to support data duplication for achieving higher reliability is discussed. Then, layer-1 and layer-2 features for URLLC are described, including features that are currently being considered in Rel-16 work on NR-Industrial IoT and enhanced URLLC (eURLLC). The following continues to describe how LTE and NR deliver precise time references to UEs as well as how Ethernet compression works when Ethernet PDUs are delivered through the 5G RAN. For industrial IoT use cases such as factory automation, reliability needs to be ensured for both data and control planes. Further, a description of how reliable control plane and reliable mobility can be achieved. A technology roadmap is described, highlighting the feature sets specified in Release 15 LTE and Release 15 NR as well as planned for Release 16 NR, and is concluded with a summary.

5G RAN Architecture Options

This sub-section introduces the 5G RAN architecture on which the subsequent description of features to support Industrial IoT is based on.

The 5G standardization work in 3GPP concluded for Release 15 for NR, for LTE, and for Multi-Connectivity including both NR and LTE. Release 15 is the first release for the newly developed radio access technology 5G NR. In addition, several LTE features necessary to enable 5G use cases have been specified. These new Rel-15 NR and LTE standards support integration of both technologies in multiple variants i.e. LTE base stations (eNBs) interworking with NR base stations (gNBs) with E-UTRA core network (EPC) and 5G core network (5GC), respectively. In such integration solutions, the user-equipment (UE) connects via different carriers with two radio base stations, of LTE or NR type, simultaneously, which is generally denoted Dual Connectivity (DC) and in the case of LTE+NR denoted EN-DC/NE-DC. The network architectures allowing for LTE and NR interworking are illustrated in FIGS. 49 , 50 , and 51 .

FIG. 49 shows the control plane of the RAN in case of Multi-Connectivity. In the EN-DC case, shown on the left of the figure, the LTE master eNB (MeNB) is the anchor point towards the MME of the EPC. In this case the NR node, gNB, is integrated into the LTE network (therefore denoted en-gNB). In the NR-NR DC case, shown on the right, both master and secondary node (MN and SN) are of NR gNB type, where MN terminates the control plane interface to the 5GC, i.e. to the AMF.

FIG. 50 shows the user plane network architecture, again with the EN-DC case on the left and the NR-NR DC case shown on the right. In the user plane, data can be directly routed to the secondary node (en-gNB in EN-DC, and SN in NR-NR DC) from the core network or routed via the MeNB/MN towards the secondary node. Transmission/Reception to/from the UE may then happen from both nodes.

The protocol architecture for the radio access in LTE and NR is largely the same and consists of the physical layer (PHY), medium access control (MAC), radio link control (RLC), packet data convergence protocol (PDCP), as well as (for QoS flow handling from 5GC for the NR) the service data adaption protocol (SDAP). To provide low latency and high reliability for one transmission link, i.e. to transport data of one radio bearer via one carrier, several features have been introduced on the user plane protocols for PHY and MAC, as we will see further in the respective sections below. Furthermore, reliability can be improved by redundantly transmitting data over multiple transmission links. For this, multiple bearer type options exist.

In FIG. 51 , the different radio bearer types for NR, which both user plane and control plane bearers (DRB or SRB) can assume, are illustrated. In the Master cell group (MCG) or secondary cell group (SCG) bearer type transmissions happen solely via the cell group of the MeNB/MN or en-gNB as secondary node/SN respectively. Note that MCG and SCG are defined from viewpoint of the UE. However, from the network point of view, those bearers may be terminated in MN or SN, independently of the used cell group.

In the split bearer type operation, data is split or duplicated in PDCP and transmitted via RLC entities associated with both MCG and SCG cell groups. Also, the split bearer may be terminated in MN or SN. Data can be conveyed to the UE via one more multiple of those bearers. Duplication of data is possible for MCG or SCG bearer when additionally employing CA within a cell group, or by employing the split bearer for duplication among cell groups; which is further described below. Furthermore, redundancy can also be introduced by transmitting the same data over multiple bearers, e.g. MCG terminated bearer and SCG terminated bearer, while handling of this duplication happens on higher layer, e.g. outside of RAN.

URLLC Enablers in User Plane

For the operation of URLLC services, i.e. provisioning of low latency and high-reliability communication, several features have been introduced for both LTE and NR in Rel-15. This set of features constitutes the foundation of URLLC support, e.g. to support 1 ms latency with a 1-10{circumflex over ( )}-5 reliability.

In the RAN concept described, these URLLC features are taken as a baseline, with enhancements developed for both Layer 1 and Layer 2. These are on the one hand serving the purpose of fulfilling the more stringent latency and reliability target of 0.5 ms with 1-10{circumflex over ( )}-6 reliability, but on the other hand also allowing more efficient URLLC operation, i.e., to improve the system capacity. These enhancements are also particularly relevant in a TSN scenario, i.e. where multiple services of different (mostly periodic) traffic characteristics must be served with a deterministic latency.

In this section, URLLC enablers for user plane data transport, i.e. the Layer 1 and Layer 2 features, are described. This is one part of the overall RAN concept only; to support the 5G TSN integration from RAN, further aspects are considered, such as reliability in the control plane and mobility, as well as accurate time reference provisioning.

Note that in most cases, the main descriptions herein are based on NR, although in certain cases LTE descriptions are provided as baseline while the features are conceptually also applicable to NR. Further below, a table is provided identifying whether the features are specified for LTE/NR. Whether a feature is required depends on the specific URLLC QoS demand in terms of latency and reliability. Furthermore, some of the features can be seen not as enablers for URLLC itself but enabling more efficient realization of URLLC requirements by the system, i.e., features that enhance capacity will result in an increased number of URLLC services that can be served. Therefore, these features can be roughly grouped as essential features for low latency, essential features for high reliability, and others, as follows.

Essential features for low latency:

Scalable and flexible numerology Mini-slots and short TTIs Low-latency optimized dynamic TDD Fast processing time and fast HARQ Pre-scheduling on uplink with configured grants (CG) (Layer 2);

Essential features for high reliability:

Lower MCS and CQI for lower BLER target

Furthermore, the following features have been considered as well:

Short PUCCH: e.g. for fast scheduling request (SR) and faster HARQ feedback DL pre-emption: for fast transmission of critical traffic when other traffic is ongoing DL control enhancements: for more efficient and robust transmission of downlink control Multi-antenna techniques: improving the reliability Scheduling request and BSR enhancements: for handling of multiple traffic types PDCP duplication: for carrier-redundancy i.e. even more reliability

The following discussion will review these features as specified in Release 15, a description of enhancements suitable for Release 16, as well as new feature descriptions suitable for Release 16, starting with Layer 1 and continuing with Layer 2.

URLLC Enablers in User Plane

In NR, a slot is defined to be 14 OFDM-symbols and a subframe is 1 ms. The length of a subframe is hence the same as in LTE but, depending on OFDM numerology, the number of slots per subframe varies. (The term “numerology” refers to the combination of carrier spacing, OFDM symbol duration, and slot duration.) On carrier frequencies below 6 GHz (FR1) the numerologies 15 kHz and 30 kHz SCS (Sub-Carrier Spacing) are supported while 60 kHz SCS is optional for the UE. 15 kHz SCS equals the LTE numerology for normal cyclic prefix. For frequency range 2 (FR2) the numerologies 60 and 120 kHz SCS are supported. This can be summarized in Table 8.

The possibility of using different numerologies has the benefit of adapting NR to a wide range of different scenarios. The smallest 15 kHz subcarrier spacing simplifies co-existence with LTE and gives long symbol duration and also long cyclic prefix length, making it suitable for large cell sizes. Higher numerologies have the benefit of occupying a larger bandwidth, being more suitable for higher data rates and beamforming, having better frequency diversity, and, important for URLLC, having a low latency thanks to the short symbol duration.

The numerology itself can thus be considered as a feature for URLLC, since transmission time is shorter for high SCS. However, one needs to consider signalling limitations per slot, such as PDCCH monitoring, UE capabilities, and PUCCH transmission occasions, which can be a limiting factor, since UE is less capable per slot basis at high SCS.

NR provides support for mini-slots. There are two mapping types supported in NR, Type A and Type B, of PDSCH and PUSCH transmissions. Type A is usually referred to as slot-based while Type B transmissions may be referred to as non-slot-based or mini-slot-based. Mini-slot transmissions can be dynamically scheduled and for Release 15:

Can be of length 7, 4, or 2 symbols for DL, while it can be of any length for UL Can start and end in any symbol within a slot.

Note that the last bullet means that transmissions may not cross the slot-border which introduces complications for certain combinations of numerology and mini-slot length.

Mini-slots and short TTI both reduce the maximum alignment delay (waiting time for transmission opportunity) and transmission duration. Both the maximum alignment delay and the transmission duration decrease linearly with a decreased TTI and mini-slot length, as can be seen in FIG. 52 , which shows latency from the use of mini-slots, compared to the “normal” 14 OFDM symbol slots. The results in FIG. 52 are based on downlink FDD one-shot, one-way latency, assuming capability-2 UE processing. In certain wide-area scenarios, higher numerology is not suitable (the CP length is shortened and may not be sufficient to cope with channel time dispersion) and use of mini-slots is the main method to reduce latency.

A drawback with mini-slot is that a more frequent PDCCH monitoring needs to be assigned. The frequent monitoring can be challenging for the UE, and also uses up resources that otherwise could be used for DL data. In NR Rel-15 the number of monitoring occasions that can be configured will be limited by the maximum number of blind decodes per slot and serving cell the UE can perform and the maximum number of non-overlapping control channel elements (CCEs) per slot and serving cell.

To maintain efficiency for data symbols we can expect higher L1 overhead with mini-slots due to the higher fraction of resources used for DMRS. Even if only a fraction of the OFDM symbols is used for DMRS, it could be one symbol out of e.g. 4 instead of 2 symbols out of 14 for a slot.

Based on formulated drawbacks, the following challenges related to mini-slots are being addressed in NR Release 16:

Mini-slot repetitions (including repetitions crossing slot border); Reduction of DMRS overhead; Enhanced UE monitoring capabilities; Fast processing in UE and gNb.

Release 16 solutions for these challenges are described below.

With regards to min-slot repetitons, since URLLC traffic is very latency sensitive, the most relevant time allocation method is type B, where one can start transmission at any OFDM-symbol within a slot. At the same time the reliability requirements can lead to very conservative link adaptation settings, hence, lower MCSs may be selected which requires more RBs. Instead of having wider allocation in frequency, gNB can decide to allocate longer transmission in time which can help to schedule more UEs at the same time. Unfortunately, due to restrictions in Release-15 NR, the transmission must be delayed in time if it overlaps with the slot border. The illustration of this issue is presented in FIG. 53 , which is an illustration of long alignment delay due to transmission across slot border restriction in NR Release 15. Here the alignment delay is a time between two events: when UE is ready for transmission and when transmission is taking place in the beginning of the next slot.

To illustrate the latency gains possible by allowing scheduling of a transmission to cross the slot border using mini-slot repetition we look at the average latency gains compared to scheduling transmissions that are constrained to fit in one slot. One way of using mini-slot repetitions to achieve this is illustrated in FIG. 54 , but other ways give the same overall latency.

Given an assumption that data packets are equally likely to arrive at the UE at any symbol within a slot, Tables 9-11 show the worst case latency for different combinations of transmission durations and SCS for non-cross-border and cross-border scheduling respectively, considering UL configured grant with HARQ-based retransmissions. Since there are 14 symbols in a slot and we typically target very low block error probabilities, we need to ensure that the latency bound can be achieved when data arrives at the symbol that gives the worst-case latency. We evaluate the latency assuming capability 2 UE, and that the gNB processing time is the same as the processing time at the UE. We assume that the gNB uses half of the processing time for decoding, i.e., if the transport block is decoded correctly it can be delivered to higher layers after half the processing time. Since allowing HARQ retransmissions can lower the amount of resources used considerably by targeting a higher BLER in the first transmission we evaluate the latency after the initial transmission, 1st, 2nd, and 3rd HARQ retransmission, taking into account the time needed to transmit PDCCH scheduling the retransmission and the time needed to prepare the PUSCH retransmission. We assume that any retransmissions use the same length as the initial transmission.

In Tables 9-14 we show the worst case latency for HARQ-based retransmission achievable with Release 15 (transmission not crossing the slot border) and the worst case latency when using mini-slot repetition to allow crossing the slot border. We consider SCS=15, 30, or 120 kHz, and a total PUSCH length of 2 to 14 symbols, counting any repetitions, i.e., a 2-symbol mini-slot repeated 4 times shows up in the tables as a length 8 transmission. To make the tables easier to interpret, they focus on target latencies of 0.5, 1, 2, and 3 ms respectively. In the tables showing the worst-case latencies using mini-slot repetitions, the shaded cases show cases where one of these target latency bounds can be met using mini-slot repetitions but cannot be achieved using Release 15.

In comparison to Release 15 scheduling, the following gains can be reached:

For a latency bound of 0.5 ms, using mini-slot repetitions allows an additional 5 cases. The gains occur for the initial transmission for 30 and 120 kHz SCS. For a latency bound of 1 ms, using mini-slot repetitions allows an additional 6 cases. The gains occur for the initial transmission for 15 and 30 kHz SCS. For a latency bound of 2 ms, using mini-slot repetitions allows an additional 11 cases. The gains occur for the initial transmission, the 1st, or 2nd retransmission for 15, 30, or 120 kHz SCS. For a latency bound of 3 ms, using mini-slot repetitions allows an additional 7 cases. The gains occur for the 2nd or 3rd retransmission for 15 or 30 kHz SCS.

The mini-slot repetition in UL can be used together with other features, enabling higher reliability, such as frequency hopping according to certain pattern or precoder cycling across repetitions.

PUCCH enhancements include the use of Short PUCCH. For DL data transmission, the UE sends HARQ feedback to acknowledge (ACK) the correct reception of the data. If the DL data packet is not received correctly, the UE sends a NACK and expects a retransmission. Due to strict latency constraint of URLLC, short PUCCH format with 1-2 symbols (e.g., PUCCH format 0) are expected to be of high relevance. Short PUCCH can be configured to start at any OFDM symbol in a slot and therefore enables fast ACK/NACK feedback suitable for URLLC. However, there exists a trade-off between low latency and high reliability of HARQ feedback. If more time resources are available, it is also beneficial to consider a long PUCCH format which can have a duration of 4 to 14 symbols. With the use of longer time resources, it is possible to enhance PUCCH reliability.

Another enhancement is UCI multiplexing with PUSCH. For a UE running mixed services with both eMBB and URLLC, the reliability requirements on UCI transmitted on PUSCH can differ significantly from the PUSCH data. The reliability requirement on the UCI can either be higher than the requirement on the PUSCH data, e.g., when transmitting HARQ-ACK for DL URLLC data at the same time as eMBB data, or lower, e.g., when transmitting CQI reports meant for eMBB at the same time as URLLC data. In the case where UCI has lower requirement than PUSCH data, it may be preferable to drop some or all of the UCI.

The coding offset between UCI and PUSCH data is controlled through beta factors for different types (HARQ-ACK, CSI) of UCI. An offset larger than 1.0 means that corresponding UCI is coded more reliable than data. The beta factors defined in Release 15 have a lowest value of 1.0. This value might not be low enough when considering URLLC data together with eMBB UCI. The better solution would be an introduction of special beta-factor value allowing to omit UCI on PUSCH to ensure URLLC reliability. This approach is illustrated in FIG. 55 , which shows the use of a beta-factor in DCI signals to “omit” UCI transmission. A related issue is when a scheduling request (SR) for URLLC mini-slot transmission comes during slot-based transmission. This issue is analyzed further below.

Other enhancements are in the area of power control. When UCI is transmitted on PUCCH the reliability requirement can differ significantly if UCI is related to eMBB or URLLC/eURLLC. For Format 0 and Format 1, the number of PRBs equals one and an attempt to increase reliability by using more PRBs makes PUCCH sensitive to time dispersion. Therefore, for Format 0 and Format 1, different reliability can be achieved by different number of symbols and/or power adjustment.

The number of symbols can be dynamically indicated in downlink DCI using the field “PUCCH resource indicator,” wherein two PUCCH resources are defined with different number of symbols. Power adjustments are however limited to a single TPC table and/or possibly using the PUCCH spatial relation information wherein multiple power settings (such as P 0 ) and up to two closed-components can be defined. But, the different PUCCH power settings can only be selected using MAC CE signaling. This is clearly too slow in a mixed services scenario where the transmitted HARQ-ACK may be changed from related to eMBB to related to URLLC/eURLLC between two consecutive PUCCH transmission opportunities. As a solution for this issue, PUCCH power control enhancements can be introduced in NR Release 16 to enable larger power difference between PUCCH transmission related to eMBB and PUCCH transmission related to URLLC:

New TPC table allowing larger power adjustment steps, and/or Dynamic indication of power setting (e.g., P 0 , closed-loop index) using DCI indication

Further enhancements regard HARQ-ACK transmission opportunities. For URLLC with tight latency requirements there is a need to have several transmission opportunities within a slot when using mini-slot based PDSCH transmissions and hence also a need for several opportunities for HARQ-ACK reporting on PUCCH within a slot. In Release 15, at most one PUCCH transmission including HARQ-ACK is supported per slot. This will increase the alignment time for sending the HARQ-ACK and therefore the DL data latency. To reduce the downlink data latency, it is necessary to increase the number of PUCCH opportunities for HARQ-ACK transmission in a slot, especially if multiplexing of eMBB and URLLC traffic is supported on the downlink. While a UE processing capability gives the minimum number of OFDM symbols from the end of a PDSCH transmission until the beginning of the corresponding HARQ-ACK transmission on a PUCCH, the actual transmission time of HARQ-ACK is further limited by the allowed number of PUCCHs within the slot.

In Release 15, a UE can be configured with maximum four PUCCH resource sets where each PUCCH resource set consisting of a number of PUCCH resources, can be used for a range of UCI sizes provided by configuration, including HARQ-ACK bits. The first set is only applicable for 1-2 UCI bits including HARQ-ACK information and can have maximum 32 PUCCH resources, while the other sets, if configured, are used for more than 2 UCI bits including HARQ-ACK and can have maximum 8 PUCCH resources. When a UE reports HARQ-ACK on PUCCH it determines a PUCCH resource set based on the number of HARQ-ACK information bits and the PUCCH resource indicator field in last DCI format 1_0 or DCI format 1_1 that have a value of PDSCH-to-HARQ feedback timing indicator indicating same slot for the PUCCH transmission. When the size of the PUCCH resource set is to at most 8, the PUCCH resource identity is explicitly indicated by the PUCCH resource indicator field in the DCI. If the size of PUCCH resource set is more than 8, the PUCCH resource identity is determined by the index of first CCE for the PDCCH reception in addition to the PUCCH resource indicator field in the DCI.

For URLLC with tight latency requirements, there is a need to have several transmission opportunities within a slot for PDSCH transmission and hence also a need for several opportunities for HARQ-ACK reporting on PUCCH within a slot as mentioned earlier.

This means that a UE needs to be configured with several PUCCH resources to enable the possibility for multiple opportunities of HARQ-ACK transmissions within a slot although that only one of them may be used in each slot. For example, a UE running URLLC service may be configured with possibility of receiving PDCCH in every second OFDM symbol e.g. symbol 0,2, 4, . . . , 12 and PUCCH resources for HARQ-ACK transmission also in every second symbol, e.g. 1, 3, . . . , 13. This means that UE need to be configured with a set of 7 PUCCH resources just for HARQ-ACK reporting for URLLC for a given UCI size range. Since there may be a need to have other PUCCH resources for other needs the list of at most 8 PUCCH resources that can be explicitly indicated by PUCCH resource indicator in the DCI may likely be exceeded. If there are more than 8 PUCCH resources in the set in case of 1-2 HARQ-ACK bits the index of first CCE will control which PUCCH resource is indicated. Hence, the locations where the DCI can be transmitted may be limited to be able to reference an intended PUCCH resource. Consequently, this may impose scheduling restrictions where the DCI can be transmitted and may also cause “blocking” if the DCI cannot be sent on desired CCE (due to that it is already used for some other UE). Therefore, instead of configuring 7 PUCCH resources in the example above, one can assume one PUCCH resource with a periodicity for transmission opportunity of every 2 symbols within a lot. This approach is illustrated in FIG. 56 , which shows a short PUCHH that occupies one OFDM symbol (i.e., Ns=1), with a period (P) of two OFDM symbols. Here, a total of 7 periodic PUCCH resources are defined in a slot.

The solution and problem described above apply for FDD as well as TDD. However, for fixed “mini-slot” TDD pattern 8 PUCCH resources that can be explicitly indicated may be enough since only the UL part of the slot can comprise PUCCH resources.

With regards to PDCCH enhancements, with high reliability requirement for URLLC, it is important that transmission of downlink control information (DCI) is sufficiently reliable. It can be achieved by several means including improved UE/gNB hardware capabilities, enhanced gNB/UE implementation, and good NR PDDCH design choices.

In terms of design choices, NR PDCCH includes several features which can enhance reliability. These include:

Being DMRS-based which allows the use of beamforming; Support of distributed transmission scheme in frequency; Aggregation level 16; Increased CRC length (24 bits).

NR supports two main DCI formats, namely the normal-sized DCI formats 0-1 and 1-1 and the smaller-size fall-back DCI formats 0-0 and 1-0. Although scheduling flexibility can be limited, it may still be reasonable to consider fall-back DCI for data scheduling to obtain PDCCH robustness due to lower coding rate for a given aggregation level. Moreover, it can be noted that normal DCI contains several fields which are not relevant for URLLC such as bandwidth part indicator, CBG-related fields, and the second TB related fields.

One possible enhancement is a URLLC specific DCI format. Both aggregation level (AL) and DCI size can have impact on PDCCH performance. Aggregation levels have different channel coding rate and are used in link adaptation for PDCCH, while DCI payload size is rather fixed for configured connection. To make PDCCH transmission more robust, one can use high AL and/or small DCI payload size to lower PDCCH code rate. PDCCH performance comparison between different DCI sizes is summarized in Table 15. Here, DCI size 40 bits serves as a reference for the Release 15 fallback DCI size, while DCI sizes 30 and 24 may be referred to as compact DCI sizes. One can see that the gains of reducing DCI size from 40 to 24 bits are small especially at high AL, the gain is even smaller when reducing DCI size from 40 to 30 bits. The gain essentially depends on the level of code rate reduction.

With high reliability requirement for URLLC, it is important that transmission of downlink control information (DCI) is sufficiently reliable. It can be achieved by several means including improved UE/gNB hardware capabilities, enhanced gNB/UE implementation, and good NR PDDCH design choices.

In terms of design choices, NR PDCCH includes several features which can enhance reliability. These include:

Being DMRS-based which allows the use of beamforming; Support of distributed transmission scheme in frequency; Aggregation level 16; Increased CRC length (24 bits).

NR supports two main DCI formats namely the normal-sized DCI formats 0-1 and 1-1 and the smaller-size fall-back DCI formats 0-0 and 1-0. Although scheduling flexibility can be limited, it may still be reasonable to consider fall-back DCI for data scheduling to obtain PDCCH robustness due to lower coding rate for a given aggregation level. Moreover, it can be noted that normal DCI contains several fields which are not relevant for URLLC such as bandwidth part indicator, CBG-related fields, and the second TB related fields.

When a URLLC UE operates with good channel condition, it is reasonable to use low AL for PDCCH. It was argued that compact DCI can have positive impact on PDCCH multiplexing capacity as more UEs with good channel conditions can use low AL, and thus reducing blocking probability. To check this, the impact of using compact DCI on PDCCH blocking probability is studied as a function of DCI size, number of UEs, and CORESET resources. Number of URLLC UEs in a cell is considered from 4 to 10. CORESET resources are determined based on CORESET duration and bandwidth. CORESETs are assumed to occupy 1 or 2 OFDM symbols with 40 MHz BW.

FIG. 57 shows the blocking probability per monitoring occasion as a function of DCI size, average number of UEs, and CORESET sizes. The simulation assumption is for Release 15 enabled use case. It can be seen from FIG. 57 that PDCCH blocking probability per monitoring occasion depends on several parameters such as DCI size, number of UEs, and CORESET sizes. In terms of blocking probability improvement for a given number of UEs, it is evident that using small DCI size provide much smaller gain compared to using larger control resources.

Additionally, due to demodulation and decoding complexity constraint at the UE, there exists a budget on the number of DCI sizes UE should monitor per slot, i.e., 3 different sizes for DCI scrambled by C-RNTI and 1 additional for other RNTI as agreed in Release 15. So, introducing another DCI format with smaller size will be even more challenging for satisfying the DCI size limitation.

An alternative to compact DCI for PDCCH enhancement in Release 16 may be considered. In NR Release 15, there are two main DCI formats for unicast data scheduling, namely the fall-back DCI formats 0-0/1-0, and the normal DCI formats 0-1/1-1. The fall-back DCI supports resource allocation type 1 where the DCI size depends on the size of bandwidth part. It is intended for a single TB transmission with limited flexibility, e.g., without any multi-antenna related parameters. On the other hand, normal DCI can provide flexible scheduling with multi-layer transmission.

Due to high reliability requirement of URLLC, we see that it is beneficial to use a small size fallback DCI for good PDCCH performance. At the same time, it can be beneficial to have parameters such as multi-antenna related ones to support high reliability transmission. This can motivate a new DCI format having the same size as the fallback DCI but improved from the fallback DCI to swap in some useful fields, e.g., some fields that exist in the normal DCI but are absent in fallback DCI. By having the new DCI formats with the same size as existing DCI formats, blind decoding complexity can be kept the same. It can be noted that its use may not be limited to URLLC. Any use cases which require high PDCCH reliability with reasonable scheduling flexibility should be able to leverage the new DCI format as well.

Another area for improved performance regards limits on the number of blind decodes and CCE. As discussed above, PDSCH/PUSCH mapping type B (mini-slot with flexible starting position) is a key enabler for URLLC use cases. To achieve the full latency benefits of type B scheduling, it is necessary to have multiple PDCCH monitoring occasions within a slot. For example, to get the full benefits of 2 OFDM symbol transmissions, it is preferable to have PDCCH monitoring every 2 OFDM symbols. The limits in Release 15 on the total number of blind decodes (BD) and non-overlapping CCEs for channel estimation in a slot strongly restricts the scheduling options for these kinds of configurations, even when limiting the number of candidates in a search space.

Current limits for 15 kHz SCS in NR coincide with limits for 1 ms TTI in LTE, while these limits were extended after introduction of short TTI in LTE. These Release 15 limits as shown in the first row of Table 9 and 10 can be expected to be revised in NR Release 16 in scope of URLLC framework. For example, with the current number of CCE limits, there are only at most 3 transmission opportunities per slot if AL16 is used.

Rather than specifying multiple new UE capability levels, it is proposed to specify one additional level of support for PDCCH blind decodes, for which the numbers are doubled compared to Release 15. For this additional level of support, instead of simply defining it per slot basis, it makes more sense to take into account how the BDs/CCEs are distributed in a slot for mini-slot operation. One possible choice is to define the BD/CCE limit for each half of the slot. For the first half of the slot, it is natural to assume the same number as the other cases. For the second half of the slot, assuming that UE has finished processing PDCCH in the first half of the slot, the UE should have the same PDCCH processing capability in the second half of the slot. Therefore, it is reasonable to assume the same number as in the first slot.

Considering all of the above, the corresponding increase in the BD limits can be summarized in Table 16

Similarly, a corresponding increase in the CCE limits can be summarized in Table 17.

As an alternative solution to Tables 16 and 17, one can consider introducing a limitation per sliding window, where sliding window size and number of blind decodes or CCE per window can be further defined in specification.

A consequence of increases in numbers of blind decodes and CCE limits is more PDCCH occasions in a slot, and thus a UE has higher chance of eventually being scheduled. Table 18 shows the PDCCH blocking probability after certain number of PDCCH occasions for different number of UEs per cell. (DCI size=40 bits, CORESET duration=1 symbol.) It is evident that the PDCCH blocking probability within a slot can be reduced significantly with more PDCCH occasions.

While limits on PDCCH can improve alignment delay, the processing delay reduction can additionally contribute to total latency decrease. Thus, UE processing capabilities are addressed in the following.

The downlink data transmission timeline is illustrated in FIG. 58 with one retransmission. The UL data transmission timeline is illustrated in FIG. 59 , for PUSCH via configured UL grant, with one retransmission. The delay components are:

T UE,proc : UE processing time for UL transmission. T UE,proc varies depending on DL data vs UL data, initial transmission vs retransmission, etc. In UE Capability #1 and Capability #2 discussion, variables N 1 and N 2 are used:

N 1 is the number of OFDM symbols required for UE processing from the end of PDSCH to the earliest possible start of the corresponding ACK/NACK transmission on PUSCH or PUCCH from UE perspective. N2 is the number of OFDM symbols required for UE processing from the end of PDCCH containing the UL grant reception to the earliest possible start of the corresponding the same PUSCH transmission from UE perspective.

T UL,tx : transmission time of UL data. This is roughly equal to PUSCH duration. T UL,align : time alignment to wait for the next UL transmission opportunity. T gNB,proc : gNB processing time for DL transmission. T gNB,proc varies depending on DL data vs UL data, initial transmission vs retransmission, etc. For example, for PDSCH retransmission, this includes processing time of HARQ-ACK sent on UL. For PUSCH, this includes reception time of PUSCH. T DL,tx : transmission time of DL data. This is roughly equal to PDSCH duration. T DL,align : time alignment to wait for the next DL transmission opportunity.

T UE,proc is an important latency component to improve. In Release 15, UE processing time capability #1 and #2 have been defined, where capability #1 is defined for SCS of 15/30/60/120 kHz, and capability #2 defined for SCS of 15/30/60 kHz. The more aggressive capability #2 is still inadequate for the 1 ms latency constraint. Since the latency requirements for eURLLC are in order of 1 ms (e.g., 0.5 ms), a new UE capability #3 can be defined in Release 16 NR to fulfil the latency requirements. The proposed UE capability #3 is summarized in Table 19. The impact of the proposed capability can be seen in FIG. 60 , FIG. 61 , and FIG. 62 . FIG. 60 shows downlink data latency comparison between Release 15 and the new UE capability #3 shown in Table 19. FIG. 61 shows a comparison of grant-based uplink data latency for Release 15 versus the new UE capability #3. FIG. 62 shows a comparison of configured grant uplink latency, between Release 15 and the new UE capability #3.

Another delay component T DL,align is significantly influenced by PDCCH periodicity. The worst case T DL,align is equal to the PDCCH periodicity. In Release 15, PDCCH periodicity is affected by several constraints, including: (a) blind decoding limits, (b) # CCE limits), (c) DCI sizes. To provide shorter PDCCH periodicity for eURLLC, it is necessary that the number of blind decoding limits and CCE limits be relaxed in Release 16.

Another important UE capability is related to time of CSI report generation. The faster UE can provide the CSI report the more accurate a scheduling decision will be from link adaptation perspective. In Release 15 specification there are two key values defined:

Z corresponds to the timing requirement from triggering PDCCH to the start of the PUSCH carrying the CSI report and it should thus encompass DCI decoding time, possible CSI-RS measurement time, CSI calculation time, UCI encoding time, and possible UCI multiplexing and UL-SCH multiplexing. Z′ on the other hand corresponds to the timing requirement from aperiodic CSI-RS (if used) to the start of the PUSCH carrying the report.

The difference between Z and Z′ is thus only the DCI decoding time.

In Release 15, there exists no “advanced CSI processing capability”, that is, there is only a baseline CSI processing capability defined that all UEs must support. There was a discussion to include such an advanced CSI processing capability in Release 15, but it was not included due to lack of time.

Three “latency classes” for CSI content are defined in Release 15.

Beam reporting class: L1-RSRP reporting with CRI/SSBRI Low latency CSI: Defined as a single wideband CSI report with at most 4 CSI-RS ports (without CRI reporting), using either the Type I single panel codebook or non-PMI reporting mode High Latency CSI: All other types of CSI content

For each of these three classes, different requirements on (Z, Z′) are defined (according to CSI computation delay requirement 2). There also exist a more stringent CSI requirement, CSI computation delay requirement 1, which is only applicable when the UE is triggered with a single Low Latency CSI report without UL-SCH or UCI multiplexing and when the UE have all its CSI Processing Units unoccupied (i.e., it is not already calculating some other CSI report).

In NR Release 15, the mandatory UE CSI processing capability requires a UE to support calculation of 5 simultaneous CSI reports (which may be across different carriers, in the same carrier or as a single report with multiple CSI-RS resources). The values of (Z,Z′) is CSI processing requirement 2 where thus determined so that all UEs should be able to calculate 5 CSI reports within this timeframe. As some UE implementations calculate multiple CSI reports in a serial fashion, this implies that, roughly speaking, the CSI requirement 2 is about 5× longer than what it would be if the requirement were that only a single CSI report would need to be computed.

In a typical URLLC scenario, and indeed, in many typical deployments and scenarios, the gNB is only interested in triggering a single CSI report at the time. It is thus a bit unfortunate that the timing requirement is 5× longer than it has to be for that case. This excessively long CSI calculation time puts additional implementation constraints for the scheduler, as the N2 requirement for data triggering and data to HARQ-ACK delay (K1) requirement is much lower than the CSI processing requirement.

Further improvements are possible. For CSI processing timeline enhancements for eURLLC, the introduction of a new CSI timing requirement (“CSI computation delay requirement 3”) is beneficial for sporadic traffic for the purpose to quickly get channel state at gNb. It may be sued when the UE is triggered with a single CSI report. A starting position could be to take the values defined for CSI timing requirement 2 and divide by a factor 5.

Another possible CSI processing timeline enhancement is to introduce an advanced CSI processing capability. That is, to introduce a new set of tables for the two existing CSI timing requirements (as well as for the third one just proposed). A UE could then similarly to the advanced processing capabilities for PDSCH/PUSCH indicate in its capability that is supports the more aggressive CSI timeline.

Fast HARQ is another improvement. The faster processing and UE capabilities discussed in the previous sections enable faster HARQ re-transmissions. We assume that gNB can operate with similar processing speed as the UE. To operate with HARQ re-transmissions and keep latency low there need to be frequent PDCCH monitoring occasions but also PUCCH occasions where the HARQ-ACK can be transmitted. For simplicity reasons we will be assuming zero timing advance although that cannot be assumed in reality. With non-zero timing advance the latency values may change.

Here one can focus on comparison between Release 15 and Release 16. The evaluation results are shown below. For Release 15 capability #2 we assume a PDCCH periodicity of 5 OFDM symbols (os). Note that with the CCE limit per slot of 56, it is allowed up to 3 PDCCH monitoring occasions per slot where each occasion contains at least one AL16 candidate. For Release 16 we assume improved values of N 1 and N 2 (capability #3 which was discussed in previous sections) and PDCCH periodicity of 2 symbols as a consequence of potential improvement on the limits of number of blind decodes and CCEs.

Inter-UE pre-emption is another improvement. Dynamic multiplexing of different services is highly desirable for efficient use of system resources and to maximize its capacity. In the downlink, the assignment of resources can be instantaneous and is only limited by the scheduler implementation, while in the uplink, standard specific solutions are required. Below, the existing solutions in Release 15 and additional solutions for Release 16 are discussed.

Dynamic multiplexing of different services is highly desirable for efficient use of system resources and to maximize its capacity. In the downlink, the assignment of resources can be instantaneous and is only limited by the scheduler implementation. Once low-latency data appears in a buffer, a base station should choose the soonest moment of time when resources could be normally allocated (i.e. without colliding with the resources allocated for an already ongoing downlink transmission for that UE). This may be either beginning of the slot or a mini-slot where the mini-slot can start at any OFDM symbol. Hence, downlink pre-emption may happen when long term allocation(s) (e.g. slot based) occupy resources (particularly wideband resources) and there is no room for critical data transmission which can by typically mini-slot. In this case a scheduler can send DCI to critical data UE and override ongoing transmission in downlink. When slot eMBB transmission is pre-empted, the pre-empted part of the original message pollutes the soft buffer and should be flushed to give good performance in retransmissions, which will likely happen. NR Release 15 specification allows to indicate about the pre-emption by explicit signalling, which is carried either:

Option 1. By special DCI format 2_1 over group common PDCCH or; Option 2. By special flag in multi-CBG retransmission DCI “CBG flushing out information”.

Option 1 gives an indication as a 14-bits bitmap, which addresses to reference downlink resource domains in between two pre-emption indication messages. Highest resolution of this signaling in time is 1 OFDM symbol and in frequency ½ of BWP (BandWidth Part), but not at the same time. The longer periodicity of messages, the coarser resolution. Since this is a group common signaling, all UEs within the BWP may read it.

Option 2 is a user specific way of signaling. The HARQ retransmission DCI, which contains a set of CB/CBGs, may have a special bit to indicate that the UE must first flush related parts of the soft-buffer and then store retransmitted CB/CBGs in the soft buffer.

During 3GPP discussions of Release 15 URLLC, the uplink pre-emption feature was down-scoped due to lack of time in 3GPP URLLC working item. However, the feature is under discussion of Release 16. UL pre-emption may happen where a longer eMBB UL transmission is interrupted with urgent URLLC UL transmission. Further, it can have two flavors:

Intra-UE pre-emption, where both transmissions belong the same UE. The intra-UE pre-emption is similar to DL pre-emption case where instead of gNB, UE prioritizes the transmission in the UL direction. For this some sort of indication is necessary to the gNB of incoming URLLC transmission instead of eMBB transmission. Inter-UE multiplexing, where based on the request from some UEs for urgent transmission of high priority UL traffic (URLLC traffic), the gNB needs to provide resources to accommodate transmissions as soon as possible to meet the delay requirements. It can happen that the gNB has already assigned the suitable UL resources to one or multiple other UEs for UL transmissions with less stringent requirements in terms of delay (eMBB traffic). Hence, the gNB needs to re-schedule those resources for the prioritized URLLC transmissions.

The Intra-UE pre-emption is discussed further, below, since it more implies MAC mechanisms, while the second option has clear physical layer scope.

Given two enabling mechanism based on power control and muting, the pre-emption would be achieved at the cost of 1) additional signalling and complexity both at UE and gNB due to changing ongoing or planned UL transmissions and 2) impact to the performance of eMBB traffic. For the cost to be worth investing, it is important to adopt a mechanism that ensures best the required quality of the URLLC transmissions. Both approaches can be illustrated by FIG. 63 .

A drawback with power control-based schemes is that the URLLC transmissions would suffer from the interference originating from transmissions controlled by the serving gNB where in fact those transmissions could have been de-prioritized. Moreover, power boosting of URLLC transmissions would not only increase the interference for neighbouring cells, but also impact the performance of eMBB traffic. Hence, with pre-emption-based schemes, by cancelling the on-going or pre-scheduled eMBB UL transmissions on the suitable resources that the gNB intends to use for URLLC transmissions, the gNB at least avoids possible degradation of the URLLC traffic performance due to its self-inflicted interference. It should be noted that the discussion here relates to PUSCH transmissions where other options are more suitable for controlling reliability. For PUCCH the options are more limited.

The performance of power control-based scheme is shown in FIG. 64 for 4 GHz, TDL-C with DS 100 ns, 4×2 antenna configuration and MMSE-MRC receiver when slot based eMBB transmission interferes with mini-slot URLLC. Low SE MCS table is in use.

Based on above discussion, the indication-based scheme can ensure URLLC reliability, while power control-based scheme can be considered as backward compatible solution in Release 15/16 interworking scenario. However, the former comes with an extensive signalling cost.

This implies that although the UL pre-emption indication is in fact effective in a UE-specific manner, it is a better design choice to consider a group common UL pre-emption indication with the flexibility to adjust the group size depending on the scenario, from a single UE to multiple UEs, as needed. This approach preserves the properties for the single UE case while reducing signalling overhead and blocking probability in case multiple UEs need to be pre-empted.

Aiming to reuse the already existing mechanism, when possible, the two following options are mainly considered for group common signalling of UL pre-emption:

Option 1: UL pre-emption indication based on DCI format 2_0 (dynamic SFI) Option 2: UL pre-emption indication design similar to DCI format 2_1 (DL pre-emption indication)

In option 1, it is proposed to use the existing dynamic SFI and define a new (or extended) UE behaviour as follows. When a UE detects an assignment flexible (or DL) for the symbols that have already scheduled by UE specific signalling for UL transmissions, the UE completely cancels the UL transmissions. This design choice is based on two assumptions, i.e., for the purpose of UL pre-emption, 1) dynamic SFI overrides UE specific signalling and 2) the pre-empted UL transmission is not delayed and resumed but simply cancelled. This approach is simple and requires less processing time at the UE due to the need for only cancelling UL transmissions. However, it requires the defining of a new behavior, which is based on the assumption that a later SFI over-riding a prior UE-specific DCI which by itself is contradictory with the design philosophy used in Release 15. Moreover, relying on the existing SFI regime for the simplicity reasons implies that the specified SFI table for Release 15 should be used. With careful examining the entries of this table, one can observe limitations on where the UL transmission cancellation can occur as compared to a bit map pattern that provide full flexibility.

In option 2, the DL pre-emption mechanism can be adopted for the UL pre-emption indication. This approach enables a gNB to indicate to a UE with finer granularity which resources are needed to be pre-empted by using a bit map pattern. This mechanism is flexible in the sense that depending on how the UE behaviour is defined or its capability, the bit map pattern can be used to indicate when the UL transmission should be stopped without resuming transmission afterwards. Or alternatively, it can be used to indicate to the UE when to stop and then resume the UL transmission if the UE is capable of such operation in reasonable time.

Lower MCS and CQI for lower BLER target are additional issues. Based on the evaluation presented above, it can be observed that depending on a latency requirement for URLLC there could be time only for one radio transmission. In this example an air interface must be able to guarantee very low BLER required for URLLC service. For this purpose, there were several enhancements in Release 15:

New 64QAM CQI table has been introduced for reporting at target BLER 10{circumflex over ( )}-5. The new table contains lower spectral efficiency (SE) values. Low spectral efficiency 64QAM MCS table has been introduced to use without transform precoding. Low spectral efficiency 64QAM MCS table for DFT-spread OFDM waveform table has been introduced.

As an example, we consider TBS=256 bits (=32 bytes), transmission duration of 4 OFDM symbols with 1 DMRS symbol overhead. PDSCH BLER for different MCSs supported within 40 MHz BW are given in FIG. 65 . Here, the coding rate of MCS 6 corresponds to coding rate of MCS 0 in legacy 64QAM table.

The network can configure highlighted MCS tables semi-statically by RRC. Moreover, dynamic signalling for MCS table is also supported by configuring UE with MCS-C-RNTI in addition to regular C-RNTI where MCS-C-RNTI is always associated with Low SE MCS table. UE always applies Low SE MCS table when it detects MCS-C-RNTI scrambled with PDCCH CRC and it applies semi-statically configured MCS table (64QAM or 256QAM) otherwise. As an alternative, the MCS table can be configured semi-statically when UE has only URLLC traffic, while the dynamic way is preferable in case when UE is eMBB and URLLC capable at the same time. A drawback of dynamic MCS-table signalling is higher PDCCH CRC false alarm rate due to new MCS-C-RNTI introduction.

It must be noted that CQI and MCS tables can be configured independently, e.g., legacy 64QAM MCS table can be used with new 64QAM CQI table 10{circumflex over ( )}-5 BLER reporting.

Multi-antenna techniques are another issue. There is a well-known trade-off between increased data rates (multiplexing) and increased reliability (diversity). This means that increases in one necessarily come at the cost of some degradation of the other. In mobile broadband, MIMO techniques are typically used to increase the data rates and the spectrum efficiency of the network. On the other hand, for URLLC, it may be better to spend the degrees of freedom afforded by MIMO to increase reliability. Thus, instead of using the throughput as a metric to be optimized, the network can optimize reliability metrics such as the outage probability. For example, UL performance can be improved by both UL pre-coding and intra-site UL CoMP (joint reception) as shown in FIG. 66 , which shows UL SINR for different multi-antenna techniques with and without UL CoMP (3-sector intra-site joint reception) and UL precoding (Rel-10 rank 1 4-port precoders). For “No precoding”, single-antenna transmission is used, while for “Precoding” 4 antenna elements are used (1×2 X-pol, separation=0.5 lambda).

Cyclic-delay diversity (CDD) or space-time codes can also be considered to provide additional frequency diversity in a spec-transparent manner. Multiple receive antennas provide receive diversity and provide means to maximize the received signal-to-interference-noise-ratio (SINR) after reception combining at the receiver. Diversity schemes has the benefit that they require less channel knowledge than precoding does.

Multiple antenna elements can also be used to create directional antenna beams at the transmitter and/or receiver side to increase the received SINR and thus reliability. Clearly, improved SINR is provided that the beam is pointing in correct direction and hence beamforming requires at least some channel knowledge to determine the correct direction of the beam.

L2 Features

In this section, Layer 2 features in the RAN are described to support the provisioning of URLLC. While multiple features for LTE and NR have been introduced for Release 15, providing the fundamental URLLC support, current studies for Release 16 standardization seek for enhancements to improve the system's efficiency when providing URLLC and also in particular targeting the support of TSN integration i.e. support of multiple traffic flows of different QoS requirements. Assumed here is that not only should non-critical traffic be efficiently transmitted, other critical traffic flows should be served with a deterministic latency. In a TSN scenario, these traffic flows are typically periodical but not necessarily. In general, we address scenario where full knowledge of when, which size and with which pattern/period traffic arrives at the gNB or UE is not available a-priori. We investigate the Release 15 baseline and enhancements in the following sections on SR and BSR, Pre-scheduling for cyclic traffic, UE multiplexing, as well as PDCP duplication.

It shall be noted that the L2 features are generally independent of whether FDD or TDD is used.

Buffer Status Reports (BSR) and Scheduling Requests (SR) are the two methods which the UE can use to indicate that data is available in the transmission buffer. These indications may result in that the network provides a grant, i.e., UL-SCH resources to the UE to allow data transmissions. This is commonly known as dynamic scheduling. An example of SR and BSR operation is shown in FIG. 67 .

In a nutshell, one of the major differences between SR and BSR is that the SR is a one-bit indication in PUCCH which signals that the UE has data for transmission, while the BSR explicitly provides an approximate value of the amount of data that the UE has in its buffer on a per logical channel group basis. The BSR is transmitted in a MAC Control Element (CE) which is transmitted in the PUSCH.

In NR Release 15, one SR configuration can be configured per each logical channel, and several logical channels may be configured with the same SR configuration. The SR is transmitted in the PUCCH. In one bandwidth part (BWP), an SR may be configured with, at most, one PUCCH resource. This means that, in NR, the network may configure multiple SR configurations which could, potentially, be used for different types of traffic.

The procedure can be summarized as follows:

Data from a certain logical channel arrives.

A Regular BSR is triggered due to the arrival, given the triggering specified criteria are met.

No PUSCH resources are available to transmit the BSR.

An SR is triggered and transmitted in the SR resource associated to the logical channel which triggered the BSR.

Dynamic scheduling introduces a delay to the data transmissions, as shown in FIG. 67 . This delay depends on the periodicity/offset of the SR configuration and the time the network takes to allocate resources and transmit a grant.

Some Industrial IoT services and traffic may need to meet tight delay requirements. “Multiple SR configurations” as specified in Release 15, is thus a feature which can play a key role to ensure traffic differentiation and to ensure that delay requirements are met. An example is depicted in FIG. 68 , which shows multiple SR configurations mapped to different traffic.

Buffer Status Report (BSR), as specified, is transmitted by the UE in the PUSCH. The BSR is transmitted as a MAC Control Element in the MAC PDU. The purpose of the BSR is to indicate the approximate amount of data in the buffers. This report is indicated per Logical Channel Group (LCG). Each logical channel will be associated to a LCG. There are 8 LCG. In scenarios in which there is a need to differentiate among a limited set of traffic profiles (DRBs), the number of LCG may be sufficient to provide a 1-to-1 mapping between logical channels and LCG.

There are 4 different BSR formats and depending on the selected format, the UE may be able to indicate the buffer status of one or more logical channels groups.

The BSR can be triggered by one of the following mechanisms:

Regular BSR: A regular BSR is triggered when a logical channel which belongs to a certain LCG receives new UL data for transmission. In addition, this new data must fulfill one of the following two conditions: the new data belong to a logical channel with higher priority than any of the other logical channels which have data; or, there is no other data available for transmission in the LCG in any of the logical channels.

A regular BSR will never be triggered if more data is received in another certain logical channel and that logical channel has already data in the buffer. A regular BSR can only use the Short and Long BSR formats.

Periodic BSR: A periodic BSR is triggered periodically following the configuration provided by the network.

A periodic BSR can only use the Short and Long BSR formats.

Padding BSR: When the UE receives a larger grant than what it needs to transmit the data, the UE may be able to transmit a BSR instead of padding bits. Depending on the number of padding bits, the UE will transmit a different BSR format.

Padding BSR can use all the BSR formats.

SR and BSR will play an important role to assist Industrial IoT traffic to meet the different requirements of each traffic, especially when the traffic periodicity and size is unpredictable.

“Multiple SR configuration” may be a key feature to differentiate traffic which has strict delay requirements and dynamic scheduling as the preferred method to allocate UL network resources. A specific SR configuration could be mapped to a specific Logical Channel (which could carry traffic with specific requirements e.g. very low latency requirement). When the network receives this specific SR (which can be identified by the specific resources allocated to it), the network can identify that there is traffic with low latency requirements waiting for transmission. Then, the network may prioritize the allocation of resources to this traffic.

One possibility is that predictable Industrial IoT traffic (known periodicity/packet sizes) is mapped to a specific SR configuration. The SR configuration would then identify the traffic which would allow the network to allocate the appropriate resources for that specific traffic. On the other hand, LCHs with non-predictable traffic (packet sizes are unknown) would then be mapped to a generic SR configuration, a generic SR shared by a number of other LCH. In this case, the SR configuration cannot assist the network to identify the traffic and, therefore, the LCH needs to rely on the BSR indications to provide relevant information to the network which could assist to the scheduling decisions. Thus, Buffer Status Reporting will also be a key feature especially in scenarios in which non-predictable traffic is expected.

It is expected that Industrial IoT is based on the SR procedure designed in Release 15, but minor enhancements might be introduced in Release 16. For example, it is up to the UE to decide which SR configuration is used when there are several pending SRs. This UE behavior could be changed so that the SR configuration linked to the highest priority logical channel is selected by the UE. However, this was discussed during Release 15 without reaching any possible agreement. Furthermore, currently even though a frequent PUCCH resource is allocated for allowing quick SR transmissions when critical data arrives, when a long PUSCH transmission is ongoing, the SR can only be sent at the PUCCH resource after this long PUSCH duration, as PUCCH and PUSCH cannot overlap according to current specification. BSR might be transmitted in this case instead via PUSCH, but given the PUSCH is long (slot length, low OFDM numerology), it may also be associated with a long decoding/processing delay. This is shown in FIG. 69 , which shows delayed SR due to ongoing long UL-SCH. Therefore, it is envisaged in Release 16 to allow parallel PUCCH transmission for SR on overlapping PUSCH resources, reducing the latency for the SR.

BSR for Industrial IoT will also be based on Release 15 and minor enhancements might be also introduced. During the development of Release 15, it was proposed that new data would always trigger a BSR. This behavior was not accepted and the LTE behavior was adopted. That means that new data coming to a logical channel does not trigger a regular BSR if the logical channel group already had buffered data, or the new data belongs to a lower priority logical channel. Nevertheless, for Industrial IoT Release 16, it has been discussed again whether new data would always trigger a BSR, which would have the advantage that otherwise required frequent periodical BSR transmissions can be avoided.

Another aspect not discussed in these SR/BSR sections is the priority of the MAC CE for BSR in the logical channel prioritization procedure. MAC CE for BSR, with exception of padding BSR, has a higher priority than data from any DRB. In other words, MAC CE for BSR is transmitted before any user data per current operation. However, some optimizations targeted for NR Industrial IoT are possible:

The priority of the MAC CE for BSR is configurable, i.e., it can be modified (reduced) by the network. In this manner, certain DRBs e.g. DRBs carrying data with very low delay requirements can have a higher priority than MAC CE BSR.

In the following, we address pre-scheduling grants which is used in both Release 15 and 16. Such grants removes the delay introduced by waiting for SR transmission occasions and the corresponding response (i.e. grant).

In Release 15, when a UE does not have UL resources allocated and data becomes available, the UE needs to undergo the scheduling request procedure, i.e., request UL resources from the gNB, which are then granted. This comes with an additional UL access delay, unwanted for transmission of critical traffic, such as TSN stream data. Pre-scheduling of grants is a technique to avoid the extra latency resulting from SR-to-grant procedures when using dynamic scheduling, as illustrated in FIG. 70 .

Pre-scheduling can be done by implementation by the gNB pro-actively sending out multiple UL grants for potential UL transmissions. The standard in LTE and NR Release 15 supports this concept by allowing pre-scheduling of multiple, periodically recurring UL grants. It builds on the semi-persistent scheduling concept (SPS) originally introduced for LTE VOIP. In NR, such pre-scheduling scheme is called semi-persistent scheduling in the downlink (DL), whereas it is called configured grant (Type 1 and Type 2) in the uplink (UL).

The NR DL SPS assignment is the same as in LTE, which is a configured assignment provided by PDCCH/L1 signal (can also be deactivated/activated).

The NR UL configured grant (CG) has been specified in two variants, configured grant type 1 and type 2. In both variants gNB pre-allocates the resources of the grants (via different signaling) including:

Time-frequency resources (via RRC for Type 1 and DCI for Type 2) Period (via RRC), offset (via RRC for Type 1 and implicitly at DCI reception for Type 2) MCS, Power parameters (via RRC for Type 1 and DCI for Type 2) DMRS, repetitions (via RRC for Type 1 and DCI for Type 2) HARQ configuration; (via RRC) Activate/Deactivate message (via DCI for Type 2).

Both configured grants type 1 and type 2 share several commonalities, such as:

“Configured Scheduling” CS RNTI used on PDCCH for activation/deactivation and retransmissions. Retransmission for both Type 1 and Type 2 are based only on dynamic grant to CS RNTI (i.e. retransmissions are not sent using the periodically recurring UL grants). The dynamic grant with C-RNTI overrides a configured grant for initial transmission in case of overlap in time domain. There is at most one active Type 1 or Type 2 configuration per serving cell and BWP

One difference between Type 1 and Type 2 is the setting up procedures. The procedures of Type 1 CG are illustrated in FIG. 71 , whereas, the procedures of Type 2 CG are illustrated in FIG. 72 . It can be argued that since Type 1 CG is activated via RRC, it is best suited for traffic with deterministic arrival periodicity (on of TSN characteristics). On the other hand, Type 2 CG are suited to support streams with uncertain mis-alignment, where the grant can be reconfigured quickly with DCI (PHY signal).

A disadvantage of configured grant is the low utilization of granted resources when used to serve unpredictable yet critical traffic, because gNB will allocate resources without knowing if the traffic will arrive or not.

TSN traffic handling will be an important issue in Release 16. Several approaches to support multiple traffic flows, i.e., TSN streams are discussed here, where each stream has specific characteristics, i.e., periodicity, time offset, target reliability, latency, etc., as illustrated in FIG. 73 and FIG. 74 . FIG. 73 illustrates industrial deterministic streams with different arrival and payload sizes. FIG. 74 illustrates industrial deterministic streams with different patterns and periodicity, and differing latency and reliability requirements.

Each of the TSN stream characteristics plays a major role in scheduling the users. For instance, a TSN stream with periodical data yet ultra-low latency requirement can be best accommodated (with minimum possible network resources) if the network knows exactly the periodicity and arrival of such TSN stream data. However, if the network does not know such characteristics it will over dimension the grant to avoid violating the tight latency requirement, thereby potentially resulting in inefficient radio resource management. Furthermore, it is assumed a target reliability of the UE's TSN data stream can be reached with specific MCS index and number of repetitions. Only if the radio network accurately knows such requirements it will not over or under allocate the resources. It is assumed in the following that these traffic characteristics are not necessarily known, especially when it comes to multiple overlapping TSN streams and other non-critical traffic. Therefore, features are investigated in the following, giving the gNB the possibility to still efficiently as well as robustly schedule the traffic mix.

In Release 15, a single CG configuration within a cell/BWP can support industrial streams/flows with similar periods and other requirements (such as, latency, reliability, jitter, etc.). However, in industrial networks, as targeted in Release 16, multiple streams (data flows) generated at a node is a very common use-case, e.g., robot arm with several actuators, sensors and monitoring devices.

As a result, such multiple streams differ in its characteristics, e.g., arrival time, and payload size as shown in FIG. 73 . One of the streams has a medium size payload (in comparison to the others. Also, the packet from this stream arrives at offset zero, followed by the packets from the other two streams, which arrive at T and 2T offsets, respectively.

Furthermore, multiple streams can be characterized by different periodicity, latency and reliability requirements, as shown in FIG. 74 . Suppose the stream with the dashed outline requires not so critical reliability and latency, whereas both of the other streams require demanding reliability and latency performance. The grant's configuration parameters, such as MCS and repetition, will differ for the former, compared to the latter. Also, some streams differ in their arrival pattern and periodicity than others. Because of their different stream characteristics, all of these streams cannot be supported with a single configuration (CG), even if the CG is supported using very short periodicity, because the CG will have a single set of configuration parameters, e.g., MCS index, latency, slot period, K-repetition.

Since gNB is responsible to allocate the CG's configurations, any overlap among the configurations occurs with the knowledge of the radio network. gNB might allocate overlapping configurations to address several scenarios: 1) overcoming the mis-alignment of critical data arrival 2 ) accommodating multiple TSN streams with different characteristics. Depending on the characteristics of the configurations, the overlaps can be divided into several cases:

Case a) similar characteristics (e.g., MCS, period, K-Rep) except the starting symbol (offset). Case b) similar starting time and same periodicity (completely overlapping configurations) but different (MCS and K-Rep). Case c) different offsets and/or different priority and MCS/K-Rep.

A problem in overlapping configurations is the undefined UE decision basis for selecting which of the overlapping configurations. Assume a gNB allocates similar overlapping configuration with different offset in time to overcome the mis-alignment in critical data arrival, as shown in FIG. 75 . In such case, the UE selects the closest (in time) configuration, upon arrival of critical traffic.

Industrial applications raise additional considerations related to logical channel prioritization (LCP) restrictions and multiplexing. Following, the baseline LCP procedures are described. Then, techniques to enhance multiplexing for industrial mixed services scenarios are described.

Mixed services communication systems should address both Inter-UE and Intra-UE scenarios, however, in this section we focus on the Intra-UE one. In such systems a UE is assumed to have several traffic types that are categorized as critical and non-critical traffic. It is assumed that critical traffic is served better with configured grants, because this traffic requires very low latency high reliability in the uplink. It is further anticipated that gNB would overprovision configured grant resources to serve such traffic, because of uncertainty about traffic pattern. On the other hand, non-critical traffic has loose latency and reliability requirement and does not benefit from too robust transmissions; on the contrary: system resources might be wasted transmitting large volumes of non-critical traffic with robust grants in a capacity limited scenario. A common use-case that represents and motivates such mixed services case is an industrial robot arm that has actuators, sensors, and cameras integrated and connected to the same communication device/UE. Several RAN1/2 issues surface when such critical traffic overlaps with non-critical one.

The LCP procedures are applied whenever a new transmission is to be performed, and it is mainly used to specify how and what LCHs are going to fill the MAC PDU which is going to be sent over the PUSCH via PHY. There are mainly two parts in LCP procedures, one focuses on selecting the LCHs to be included in the MAC PDU, the other one focuses on the prioritization and the amount of each LCH's data (among the selected ones) to fill the MAC PDU.

The selection of LCHs is called LCP restriction procedures. Such procedure is controlled by several restrictions configured via RRC. Each of these restrictions allow/forbid the LCH to be included in the constructed MAC PDU. The following are the existing LCP restrictions in Release 15:

allowedSCS-List which sets the allowed Subcarrier Spacing(s) for transmission; maxPUSCH-Duration which sets the maximum PUSCH duration allowed for transmission; configuredGrantType1Allowed which sets whether a configured grant Type 1 can be used for transmission; allowedServingCells which sets the allowed cell(s) for transmission.

Logical channel priority is configured per MAC entity per logical channel. RRC configures the LCP parameters to control the multiplexing of the uplink LCH's data within the MAC. Such LCP parameters are expressed as,

priority where an increasing priority value indicates a lower priority level; prioritisedBitRate which sets the Prioritized Bit Rate (PBR); bucketSizeDuration which sets the Bucket Size Duration (BSD).

An example of how LCP multiplexing occurs is illustrated in FIG. 76 . In this example, only the “maxPUSCHDuration” restriction is considered. Higher to lower priority logical channels are located from left to right in the figure. Higher priority LCHs are placed first in the MAC PDU, followed by lower priority ones. Also, Priority bit rate (PBR) control the number of bits to be included in the MAC PDU per LCH.

Below, several scenarios that result from intra-UE mixed-services assumption are addressed. In such scenario, we assume that a single UE has to serve both critical and non-critical traffic. The critical traffic may be a-periodic or periodic and require more robust coding with relatively small size grant, compared to the non-critical traffic grant requirement. A requirement of the critical data is that it be scheduled using a periodic, robustly coded configured grant to avoid the latency induced from SR and its response procedure.

We further assume that no perfect knowledge of critical data arrival is present at the scheduler. This means that the critical traffic is a-periodic or not entirely periodic, i.e., the periodic arrival of the traffic may be affected by some jitter, or some periodic transmission opportunities may simply be skipped (due to unavailable data). In such cases, the network/scheduler cannot ideally align scheduling of periodic configured grants to the packet arrival occurrences, which results in the problems described in the next sub-sections.

Furthermore, if short periodicities of the configured grant are required to cater for very low latency-requirements of critical traffic, the short periodicity configured grant will result in imposing scheduling limitations on other non-critical traffic in the UE. Examples of such imposed scheduling limitations are 1) only short dynamic grant duration can be allocated in-between the configured grants, 2) dynamic grant has to be overlapping with the configured grant.

Problem 1: Non-Critical Traffic Sent on Robust Configured Grant

In this sub-section, we address the problem that arises when non-critical traffic is accommodated using a robust configured grant (i.e. intended for critical traffic). We assume the existence of non-critical traffic with sporadic available. Such traffic would be scheduled on robust configured grant resources that needed to be provided for sporadic critical traffic with short periodicity. As illustrated in FIG. 77 , if eMBB traffic (labeled 10 KB) is accommodated in such configured grant (1 KB per transmission occasion), the eMBB transmission takes too long (e.g. up to factor 10, or until BSR is received by network) and leads to unnecessary UL interference, which is in particular harmful if the configured grant resources were shared among users.

New LCH restrictions on the logical channel (LCH) holding non-critical traffic, as shown in FIG. 78 , can be introduced to mitigate this issue. For example, applying restrictions like “ConfiguredGrantType2Allowed” or “max ReliabilityAllowed” to a LCH supporting critical traffic enable the UE to avoid data from a non-critical LCH being sent using too robust resources.

Problem 2: Critical Traffic on Non-Robust Dynamic Grants

Another arises when a gNB needs to schedule a spectrally efficient dynamic grant for non-critical traffic in addition to robust configured grants intended for sporadic critical traffic. This is shown in FIG. 79 , which shows the extra latency when critical traffic is sent over a non-robust short grant. Assuming the same PUSCH duration of configured and dynamic grant, the existing “maxPUSCHDuration” restriction is not effective/sufficient. The critical traffic will be prioritized to be sent on a non-robust dynamic grant and hence the transmission might fail, leading to retransmission delays.

To overcome such issue, a new LCH restriction, i.e., “DynamicGrantAllowed”* or “minimumReliabilityRequired” can be introduced. Such restriction will block the critical LCH from being sent on non-robust dynamic grant, as illustrated in FIG. 80 .

Problem 3: Issues on Dynamic Grant Overriding Configured Grant

According to the current specification a configured grant is always overridden if an overlapping dynamic grant was allocated. In some scenarios a non-robust dynamic grant might overlap with a robust configured grant, as illustrated in FIG. 81 . A reason for such scenario is that gNB has to allocate a short periodicity configured grant to accommodate sporadic low-latency critical traffic.

To solve this problem, a configured grant may be conditionally prioritized, i.e. if critical data is available for transmission over the robust configured grant when there is an overlapping dynamic grant, then critical data is always prioritized as illustrated in FIG. 82 , which shows the benefit of enabling configured grant to override dynamic grant conditionally on arrival of critical data. Otherwise, the dynamic grant may be prioritized. This way, overlapping large spectrally efficient resources can be scheduled for non-critical data without risking that critical data may be transmitted on them. However, to employ this methodology, a gNB needs to decode two potential transmissions: dynamic grant and configured grant. It is noteworthy that this issue could also be solved with the solution of problem 2, i.e. providing the critical traffic LCH with restriction to not transmit on dynamic grant. Without this solution there can be cases where frequent dynamic grants are scheduled and result in unavoidable delays for the critical traffic.

Problem 4: Intra-UE UL Pre-Emption Between Grants of Different PUSCH Durations

In the industrial mixed traffic scenario, in order to enable high spectral efficiency, gNB may want to allocate longer grants to accommodate non-critical traffic. This will increase the delay of sending any sporadic critical data, as illustrated in FIG. 83 , which shows an example of overlapping grants with different PUSCH durations, since in Release 15 the current transmission cannot be interrupted by another transmission. To solve this, the physical layer (PHY) should allow stopping an ongoing (long) PUSCH and transmit new (short with higher priority) PUSCH according to overlapping short grant, as illustrated in FIG. 84 , which shows how enabling intra-UE pre-emption enhances network efficiency, depending on the scenario.

PDCP duplication is another issue to be discussed. As a method to improve reliability in LTE, NR and EN-DC, multi-connectivity within the RAN is considered. While these features previously focused on improving the user throughput, by aggregating resources of the different carriers, the focus in 3GPP has shifted recently and new features are developed for LTE (and likewise for NR) to improve the transmission reliability.

3GPP introduced carrier aggregation (CA) in Release 10, as a method for the UE to connect via multiple carriers to a single base station. In CA, the aggregation point is the medium access control (MAC) entity, allowing a centralized scheduler to distribute packets and allocate resources e.g. according to the channel knowledge among all carriers, but also requiring a tight integration of the radio protocols involved. With DC or Multi-Connectivity resource aggregation happens at PDCP. This way, two MAC protocols with their separate scheduling entities can be executed in two distinct nodes, without strict requirements on their interconnection while still allowing for realizing increased user throughput.

In 3GPP Release 15 LTE and NR, both architecture concepts of CA and DC are reused to help improve reliability as a complement to the reliability enhancements provided by PHY features. This is achieved by packet duplication, which has been decided to be employed on PDCP layer. An incoming data packet, e.g. of an URLLC service, is thereby duplicated on PDCP and each duplicate undergoes procedures on the lower layer protocols RLC, MAC, PHY, and hence individually benefits from e.g. their retransmission reliability schemes. Eventually the data packet will thus be transmitted via different frequency carriers to the UE, which ensures un-correlated transmission paths due to frequency diversity, and in case of DC transmissions from different sites thereby providing macro diversity. The method is illustrated in FIG. 85 for both CA and DC.

Frequency diversity among carriers goes beyond diversity schemes offered by the physical layer on the same carrier. Compared to time-diversity, e.g. repetition schemes, it has the advantage of mitigating potential time-correlations of the repetitions, which could e.g. occur on a carrier by temporary blocking situations. Furthermore, carrier-diversity allows, as shown in FIG. 85 for DC, the placement of transmission points in different locations, thus further reducing potential correlations of the transmission by the introduced spatial diversity.

Multi-connectivity with packet duplication on PDCP has the advantage of relying less on utilizing lower layer retransmission schemes (hybrid automated repeat request (HARQ), and RLC-retransmission) to achieve the target reliability metric, and by this lowering the latency to be guaranteed with a certain reliability. For example, let us assume the PHY achieves for each HARQ transmission a residual error probability of 0.1%. In 0.1% of the cases a retransmission is required, increasing the transmission latency by an extra HARQ round trip time (RTT). With packet duplication, the probability that both un-correlated HARQ transmissions fail is 0.1%*0.1%. That means that in 1-10{circumflex over ( )}-6 of the cases the low latency without the extra HARQ RTT is achieved, since simply the first decodable packet duplicate is accepted and delivered, while the second is discarded (at PDCP). An illustration of this relation can be found in FIG. 86 , which shows residual errors with and without duplication.

Packet duplication is considered to be applicable to both user plane and control plane, meaning that also RRC messages can be duplicated on PDCP layer. This way, latency/reliability of the RRC message transferal can be improved, which is e.g. important for handover-related signaling to avoid radio link failures.

Furthermore, multi-connectivity has the potential to enable reliable handovers without handover interruptions for user plane data. Thereby, the handover can be done in two steps, i.e. one carrier is moved at a time from source to target node, and hence the UE maintains always at least one connection. During the procedure, packet duplication may be employed, so that packets are available at both nodes for interruption-free transmission to the UE.

To support PDCP duplication in CA, a secondary RLC entity is configured for the (non-split) radio bearer used in support of duplication. See FIG. 85 . To ensure the diversity gain, restrictions can be defined for the logical channels associated with these two RLC entities, so that transmission of each RLC entity are only allowed on a configured carrier (primary or secondary cells).

Furthermore, to allow using PDCP duplication as a “scheduling tool” i.e. allowing to activate and deactivate duplication only when necessary, i.e. dynamically be the scheduler, MAC control elements had been specified.

In Release 16, within NR-Industrial IoT, enhancements to PDCP duplication in NR are envisaged, which allow duplication over more than two links, i.e. DC-based and CA-based duplication may be used together, or CA-based duplication with more than two carriers are considered. Furthermore, enhancements regarding the duplication efficiency are investigated: instead of always duplicating, the transmitter may defer from sending the duplicate if the original had been in flight already for a certain time. The reasoning is that a duplicate serves its purpose of increasing the reliability of reaching a latency bound only if both original and duplicate are received within this latency bound. One could envisage also a scenario where duplicates are only transmitted together with a retransmission, i.e., NACK-based. I.e. retransmission reliability is improved, while initial transmission reliability remains the same.

Table 20 illustrates for which bearer options, UP, CP, etc., duplication is supported.

Reference Time Provisioning

An NR-Industrial IoT feature of interest is that of providing UE based applications (e.g. residing in Industrial IoT devices connected to a UE via ethernet ports) with clock information derived from source clocks residing in networks external to the 5G network. The external source clocks can be provided in addition to the 5G system clock which is internal to the 5G system. The clocks derived from external sources can be viewed as working clocks corresponding to working domains that reside within the context of a “universal domain” as indicated by FIG. 87 .

The “universal domain” is based on the 5G system clock and is used to align operations and events chronologically within a factory (the universal domain). The working clocks are used for supporting local working domains within the universal domain wherein each local working domain consists of a set of machines. Different working domains may have different timescales and synchronization accuracy thereby requiring support for more than one working clock within the universal domain.

Within the scope of Release 15 RAN2 has focused primarily on the method by which a single reference time value can be delivered over the radio interface from a gNB to a UE and has not been concerned about or aware of any use cases wherein multiple reference time vales would need to be conveyed to a UE. The ongoing discussion within SA2/RAN3 regarding the potential need for delivering multiple reference time/working clock values to a UE continues to drive further enhancements in this area.

A 5G system supports an internal 5G clock which can be based on a very accurate and stable clock source (e.g. a GPS time) and distributed throughout the 5G network as needed, including delivery to eNBs and UEs as reference time information. It is also possible for a 5G system to acquire reference time information from an external node (not further considered herein). LTE Release 15 supports a method for delivering a single instance of reference time information (assumed to be available at an eNB) to UEs using both RRC message and SIB based methods as follows and as illustrated in FIG. 88 , which shows BS SFN transmissions:

An eNB first acquires a reference time value (e.g. from a GPS receiver internal to the 5G network) The eNB modifies the acquired reference time to the value it is projected to have when a specific reference point in the system frame structure (e.g. at the end of SFNz) occurs at the BS Antenna Reference Point (ARP) (see reference point tR in FIG. 88 ). A SIB/RRC message containing the projected reference time value and the corresponding reference point (the value of SFNz) is then transmitted during SFNx and received by a UE in advance of tR. The SIB/RRC message may indicate an uncertainty value regarding the value of reference time applicable to the reference point tR. The uncertainty value reflects (a) the accuracy with which an eNB implementation can ensure that the reference point tR (the end of SFNz) will actually occur at the ARP at the indicated reference time and (b) the accuracy with which the reference time can be acquired by the eNB.

The uncertainty introduced by (a) is implementation specific but is expected to be negligible and is therefore not further considered. When a TSN node is the source of reference time information (i.e. the TSN node serves as a GrandMaster node) the use of hardware timestamping at the GrandMaster node and eNB is assumed to be used for (b) in which case a corresponding uncertainty is expected to be introduced when conveying the GrandMaster clock to an eNB.

For NR Release 16 a method similar to LTE Release 15, as described above, is expected to be used for sourcing and delivering reference time information from a gNB to one or more UEs. However, NR Release 16 is also expected to introduce support for one or more working clocks (sourced by external nodes in the TSN network) as supplemental clock information (i.e. supplemental to the reference time provided for the universal time domain). FIG. 89 shows an industrial use case with three time domains, where an internal 5G clock serves as the reference time applicable to the universal time domain (in the 5G time domain) as well as two supplemental working clocks applicable to TSN working domain 1 and TSN working domain 2.

The internal 5G clock (shown as a 5G Grand Master) is used for serving radio related functions and is therefore delivered to both the gNB and UE (but not made available to the UPF). Once the gNB acquires the internal 5G clock (implementation specific) it can convey it to the UEs using either broadcasting (e.g. SIB) or RRC unicasting methods. The SFNs sent over the Uu interface will be synchronized to 5G internal clock and in this sense the UE will always be synchronized to the 5G internal clock even if it is not explicitly conveyed to the UEs.

The gNB receives working clock information from different external TSN nodes (i.e. directly from the TSN nodes controlling the TSN work domain clocks), thereby requiring the gNB to support PTP signalling and multiple PTP domains (multiple PTP clock instances) for communicating with TSN network. The gNB then conveys the working clocks (as standalone clocks or as offsets relative to the main internal 5G clock) to the corresponding UEs using one of two methods as follows:

a) Method 1: SFN Based Synchronization

This method of delivery is supported within the context of FIG. 89 and is the same one used for delivery of the internal 5G clock (black clock) to UEs wherein clock information is synchronized to a specific point in the SFN frame structure.

The gNB may not need to refresh the working clocks in the UEs every time it receives PTP based signalling providing it with updated values for these working clocks. This is because UEs may be able to manage the drift of these clocks with enhanced accuracy (using the internal 5G clock) compared to the rate of clock drift that may be ongoing at the source TSN Node. The net result is that the radio interface bandwidth consumed for working clock maintenance can be lower as the gNB will not need to send working clock updates to the UEs every time such updates occur within the TSN network.

In this method the gNB directly adjusts the value of the working clock information it has received according to the location within the SFN structure the working clock is mapped and then sends the adjusted value within a SIB16 or RRC message.

b) Method 2: Timestamping

In this method (also supported within the context of FIG. 80 ) the gNB supports a boundary clock function per 802.1AS and therefore obtains a working clock from the TSN network (using PTP sync message exchange) whenever the working clock source node decides to send it.

The gNB then relays the PTP message containing working clock information (or a subset of the information therein) to the UEs as higher layer payload.

The relayed PTP message also includes a time stamp providing the value of the internal 5G clock at the point where the PTP message was received by the gNB.

Upon receiving the relayed PTP message the UE adjusts the value of the working clock contained therein according to the difference between the current value of the internal 5G clock and the time stamped value also included in the relayed PTP message, thereby obtaining the current value of the working clock.

As per method 1, the gNB may not need to relay the PTP message containing the working clock to the UEs every time it receives it from the TSN network (because UEs may be able to manage the drift of these clocks with enhanced accuracy).

In this method the gNB does not adjust the value of the working clock information it has received but supplements it with time stamp information inserted directly into the same PTP message used for sending working clock information. It can then send the modified PTP message within a SIB or RRC message or, to reduce the payload size in the interest of bandwidth efficiency, the gNB can instead only map the unmodified working clock information and the corresponding time stamp into a SIB16 or RRC message.

For methods 1 and 2 above, the frequency with which a UE distributes working clocks to End Stations can be seen as implementation specific. When performed it makes use of PTP sync message exchange as performed in the TSN network. In other words, the UE acts as master clock to the TSN end stations using the (g)PTP protocol and decides when working clock values need to be refreshed in the End Stations. The UE forwards all working clocks it receives to all end stations it manages (i.e., the end stations determine which working clocks they are interested in).

For NR Release, a UPF Continuous PTP Chain Method may be used. For this method, which is illustrated in FIG. 90 , the TSN network interfaces with the UPF for the purpose of delivering working clock information wherein the UPF to UE path emulates a PTP link so that there is a virtual continuous PTP chain between the TSN Working domains on the right of the 5G network and the End Stations on the left of the 5G network (i.e. PTP sync message exchange is performed between the UE and the TSN Node supporting the working clock).

The UPF transparently relays the working clocks (e.g. green clock on the right hand side of FIG. 90 ) to each UE wherein the UPF time stamps these working clocks with the value of the internal 5G clock applicable to the point where the PTP message is relayed.

The 5G network will need some awareness of when it is relaying PTP messages containing working clock information since it will need to provide supplemental time stamp information to these PTP messages. The transport layer PDUs used to relay PTP messages from the UPF to a gNB can potentially be enhanced to support an indication of when PTP messages comprise the upper layer payload carried by these PDUs. This opens up the possibility of a gNB using SIB based transmission of the PTP message payload in the interest of radio interface bandwidth efficiency (i.e. in addition to using a RRC based option for delivering PTP messages). Upon receiving the relayed PTP message the UE adjusts the value of the working clock contained therein according to the difference between the current value of the internal 5G clock and the time stamped value included in the relayed PTP message, thereby obtaining the current value of the working clock. The UE acts as a master clock to the TSN end stations using the (g)PTP protocol and decides when working clock values need to be refreshed in the End Stations. The UE forwards all working clocks it receives to all end stations it manages (i.e. the end stations determine which working clocks they are interested in). This method does not require the use of equalized uplink and downlink delays which is an advantage (since symmetrical uplink and downlink delays impose additional complexity). However, one potential disadvantage is that the frequency with which the working clocks are refreshed by their corresponding source node within the TSN network will determine how often they are relayed through the 5G network to the UEs (e.g. this could have a significant impact on the radio interface bandwidth if each UE is individually sent user plane payload containing clock refresh information whenever any working clock is refreshed in the TSN network).

Ethernet Header Compression

For traditional IP transport over 3GPP systems header compression has been specified, i.e. robust header compression (RoHC) to reduce the volume of data sent over the radio interface, thereby RoHC is applied to the UDP/TCP/IP layers, and RoHC compression/decompression is performed by PDCP layer at UE and gNB.

In the TSN, where Ethernet transport is envisaged, header compression could potentially also be applied. This would be the case for the Ethernet PDU session type, where Ethernet frames should be conveyed between gNB and UE.

Generally, given that robust transmissions with a very low residual error rate are required for URLLC, used code rates are naturally very low, meaning that URLLC transport is resource-costly over the radio interface. Therefore, removal of unnecessary redundancy such as potentially Ethernet headers, is important to be studied as part of the Release 16 NR-Industrial IoT 3GPP study. In the following, an analysis of the Ethernet/TSN header structure and gains from compressing them is done.

Forwarding in Layer 2 (L2) networks is usually based on information available in L2 frame headers. Each Ethernet frame starts with an Ethernet header, which contains destination and source MAC addresses as its first two fields. Further header fields of an Ethernet frame are constructed quite simply using tagging. Some of the header fields are mandatory some are optional, and they depend on network scenario.

There are multiple formats of Ethernet frames (e.g., 802.3, 802.2 LLC, 802.2 SNAP). They are identified based on the value of the EtherType vs. Length field. FIG. 91 shows an example of the frame format.

Regarding Ethernet frame transmission over 3GPP networks, some parts of the Ethernet frame do not need transmission (e.g., Preamble, SFD (Start of Frame Delimiter), FCS (Frame Check Sum), see also existing specification for PDU session type, TS 23.501). Fields of the Ethernet header can be compressed but the gain achieved by compression are dependent on the network scenario. The Ethernet link can be either an access link or a trunk link. For a trunk link, the number of sessions is significantly larger and can be affected by Ethernet topology change that results in temporary flooding. On the other hand, an access link is more stable from L2 session perspective. Ethernet header compression must be L2 link specific, i.e., covering a single L2 hop (a.k.a. link-by-link basis), as illustrated above.

We consider the following fields for Ethernet header compression: MAC source and destination address (6 bytes each), tag control information (6 bytes), holding information such as VLAN-tag and Ethertype. Ethernet frame transmission over 3GPP networks does not need forwarding of some parts of the Ethernet frame (i.e., Preamble, SFD, FCS). So, in total 18 bytes can be compressed.

Assuming the 5G system is used as an Ethernet access link, only a limited number of L2 sessions would exist, and compression down to 3-5 bytes (conservative assumption) is possible, which leads to significant gains for small packet sizes (as typical in URLLC) as shown in FIG. 92 .

Regarding how and where header compression for Ethernet can be supported, the following questions may be raised.

Which protocol and standardization body: In 3GPP, RoHC as defined by IETF is used for IP header compression. There is no profile considering Ethernet. Furthermore, the standardization group dissolved. Static Context Header Compression (SCHC), also IETF is still active and considers Ethernet header compression, for the use case of low-power WAN. Also a 3GPP-based solution can be thought of. Anchor point: Current RoHC network anchor point is the gNB with PDCP. Another possibility would be the UPF, where the Ethernet PDU session is setup. FIG. 93 illustrates possible Ethernet header compression anchor points. With and without IP: Whether Ethernet header compression should be considered integrated or separate with IP header compression.

Reliable Control Plane

In this section, methods for reliable control plane provisioning i.e. for robustly maintaining the radio resource control (RRC) connection between UE and gNB, are described.

First of all, control plane, i.e. RRC signalling (SRB) transmission is handled the radio protocols as user plane data transmission, i.e. RRC signalling robustness can be established with the same features as describe for Layer 1 and Layer 2 above. Furthermore, also PDCP duplication, in the case of the split bearer in DC, as well as for CA, is also applicable to RRC signalling (SRB).

As we will see in the following, beside SRB signalling robustness, also resilience against node failures and handling of radio link failure (RRC) can be addressed: In case of a failure of the network node terminating RRC, the UE would lose its connection. Furthermore, in current Release 15 LTE and NR the radio link failure handling is not symmetrically handled, i.e. in case of failure related to the primary cell, radio link failure (RLF) is triggered, leading to a connection interruption, where the UE disconnects and searches for a new node to connect to. In case of failure related to the primary cell of the secondary cell group (SCG), however, only a failure indication is sent to RRC, while the connection continues. A similar procedure is also implemented for a secondary cell failure in case of CA duplication.

There are two failure cases that can be handled with RRC diversity today (Release 15). Specifically, for a DC architecture with PDCP duplication, both in case of secondary radio link failure, as well as in case of entire SgNB outage, the connectivity with the UE can be maintained. In case of primary cell failure or MgNB failure, this would however not be the case. These failure cases in Release 15 are illustrated in FIG. 94 .

To enable “True RRC diversity”, therefore, further enhancements need to be considered, i.e. either a fast/pro-active handover or failover of the RRC context to another node, in case of node failure, and generally symmetrical handling of radio link failures independent of in which cell the failure happened. This symmetrical handling of RLF is considered within NR WI DC in Release 16. The approach contemplated here is that instead of triggering a failure and UE interrupting data and control signaling when a failure associated with a primary cell occurs, the UE informs the network via a secondary cell, and continues its communication of data and control via this secondary cell, until reconfigured by the network.

An alternative, however costly, method would be an approach where multiple companion UEs are used for the same industrial device. Duplication and duplicate elimination would in this case happen on higher layers of the UEs. On the network side, the UEs would (as a configuration option) be connected to different eNBs/gNBs so that in case of link failure, UE failure or also node failure, the connection could be maintained via the independent companion UE.

Handover and Mobility Enhancements

For a UE in RRC-connected mode, the NR mobility mechanisms in Release 15 follow its LTE baseline, which is illustrated in FIG. 95 . The Source gNB decides (e.g. based on UE measurement reports) to handover the UE to the Target gNB. If the Target gNB admits the UE, a handover acknowledgement indication is sent to the Source gNB, which thereupon sends the handover command to the UE. The UE then switches to the new cell, indicated in the handover command and sends a handover complete indication to the Target gNB. During the switch, the UE resets MAC, re-establishes RLC and if needed re-establishes PDCP and changes security keys. The involved RACH procedure can be configured to be contention free, i.e. the RACH pre-amble to be used provided to the UE during the procedure.

For the handover to be interruption-free, i.e., in order to achieve 0 ms handover interruption time, the switching time by the UE must be minimized. For this, in LTE Release 15 (not NR), it was agreed that a dual Tx-Rx UE shall be capable of performing an enhanced make-before-break solution to ensure 0 ms handover interruption time. In this solution the UE maintains the connection to the source gNB until the UE starts to transmit/receive data from the target eNB. The details of how the dual protocol stacks are handled at the UE were left to UE implementation.

For Release 16, both in LTE and NR, some further mobility enhancements are envisaged. For reducing the handover interruption time in NR, there are several solutions under discussion for dual Tx-Rx UEs. One of which is the LTE-like enhanced make-before-break approach (described above). Other approaches, relying on the DC architecture consider a role switch operation between MN and SN and thus enabling 0 ms ‘handover’, i.e. maintaining always a connection to one of the nodes while undergoing handover. For scenarios where UEs do not have dual Tx-Rx functionalities, other approaches are envisaged such as improved i.e. faster RACH-less handover based on an improved TA calculation approach, or also faster fall-back possibilities to the source node. To improve the general handover robustness (applicable to various scenarios from URLLC or aerials domain), a solution is foreseen based on a conditional handover command (performing handover execution when a certain network configured condition is met) which reduces the handover failure/ping-pong possibility traded for higher network resource usage overhead.

One way to provide mobility without increasing latency due to handover and without requiring any capability enhancements at the UE is to deploy so-called combined cells. Combined cell is a feature that is commercially available in Ericsson's LTE networks. In a combined cell, multiple remote radios are connected to the same baseband digital unit, and serve the same cell. It allows multiple sector carriers to belong to the same cell. Combined cell can be used to extend the coverage of a cell, and provides the following additional advantages:

Reduces or eliminates coverage holes, by enabling overlapping coverage areas from different antenna sites. Increases received signal strength at UE. Provides uplink macro-diversity. Eliminates the need for inter-cell handover within the combined cell.

URLLC benefits from all of the advantages listed above. Shadowing can be a problem in factory floors, due e.g. to the presence of large metal surfaces. Combined cell can help decrease or eliminate this problem, by careful selection of the antenna sites. Increasing the signal strength at the UE is clearly beneficial for increased reliability, as is macro diversity. Avoiding or reducing the need of handovers, is also greatly beneficial for moving UEs, as handovers typically result in significant latency increase. Furthermore, combined cell provides seamless coverage in transition areas between indoor-outdoor, or indoor-indoor (e.g. multi-story halls), which would otherwise require (more) handovers. It provides a robust mechanism to grow the coverage area of the network, desirable e.g. when the factory floor is expanded.

Finally, combined cell can be used together with carrier aggregation, which provides its own benefits.

URLLC feature introduction in 3GPP Release 15 and 16 is summarized in Table 21 Table. The shading indicates features that are needed for supporting industrial IoT use cases that have stringent URLLC requirements, and the ones without shading are considered as features for efficiency optimization or scheduling flexibility.

Release 15 establishes core URLLC features enabling LTE FDD and NR, both FDD and TDD, to fulfill the IMT-2020 URLLC requirements of 99.999% reliability with 1 ms latency. For LTE, essential features for industrial IoT consist of short TTI, automatic repetitions without HARQ feedback, UL semi-persistent scheduling (SPS), reliable PDCCH, RRC diversity for achieving control-plane reliability, as well as high-precision time synchronization for allowing isochronous operation between multiple devices in the network. Although LTE FDD achieves the IMT-2020 URLLC requirements, LTE TDD however does not due to the limitation of TDD configuration. The lowest one-way user-plane latency for data transmission is limited to 4 ms in LTE TDD.

Release 15 NR meets IMT-2020 URLLC requirements with higher efficiency than LTE. One key enhancement is the scalable OFDM numerology used in NR, which combined with short TTI substantially shortens the transmission time. Another key enhancement in NR is dynamic TDD and faster DL and UL switching. NR TDD can achieve one-way user-plane latency as short as 0.51 ms.

The evolution of industrial IoT support continues in NR Release 16. One major enhancement is TSN integration, which enables NR to work with established industrial Ethernet protocols. NR Release 16 will also introduce URLLC enhancements to enable NR to meet even more stringent requirements, e.g. 99.9999% reliability with 0.5 ms latency.

In summary, NR has been designed with clear objectives of achieving low latency and ensuring high reliability from the outset. An array of layer-1 and layer-2 features in Release 15 enables URLLC:

Scalable numerology and short TTI. With scalable numerology, OFDM symbol and slot duration can be reduced by employing a larger subcarrier spacing. The transmission time interval can be further reduced by using mini-slot scheduling, which allows a packet to be transmitted in a time unit as small as 2 OFDM symbols. Scheduling design. NR supports frequent PDCCH monitoring, which increases scheduling opportunities for both DL and UL data. This helps reduce latency. For the UL, configured grant can be used to eliminate the delay incurred by UE having to first send a scheduling request and waiting for an uplink grant. In a mixed traffic scenario, NR allows URLLC traffic to be prioritized; and in events when the scheduler does not have sufficient radio resources to serve a URLLC UE, NR has a mechanism to pre-empt already allocated eMBB type resources for the use of serving DL URLLC traffic. Fast HARQ. A DL data transmission is completed by a HARQ acknowledgement, and thus a fast HARQ turn-around time is needed for achieving low latency. In NR, this is facilitated by defining a more stringent UE receiver processing time requirement (i.e. UE capability 2) and also making it possible for the UE to complete HARQ transmission in a short time interval through the use of short PUCCH. Not only does fast HARQ turn-around time contribute to low latency, it can also be used to improve the reliability or spectral efficiency of data transmissions by allowing more HARQ retransmissions within a given latency budget. Furthermore, HARQ-less repetition (sender transmits K repetitions before expecting HARQ feedback) is also adopted in NR to improve reliability without delays introduced by HARQ turn-around time. Low-latency optimized dynamic TDD. NR supports very flexible TDD configuration allowing DL and UL assignment switching at a symbol level. Robust MCS. Reliability is enhanced by including lower MCS and CQI options for lower BLER targets.

In addition, RAN architecture options are available to enhance reliability beyond the beforementioned features, i.e. by duplicating data through multiple gNBs and/or through multiple carriers.

Release 15 NR thus lays a solid foundation for supporting URLLC services. It has been also verified in the work of 3GPP IMT-2020 self-evaluation that Release 15 NR fully fulfils the IMT-2020 URLLC requirements, 99.999% reliability with 1 ms latency.

Building on the solid URLLC foundation in Release 15, Industrial IoT is in focus now in Release 16. The prioritized use cases include factory automation, electrical power distribution, and transport. Although the requirements of these prioritized use cases vary, the most demanding use cases call for 99.9999% reliability with latency as small as 0.5 ms. Furthermore, a key aspect of NR Industrial IoT is to enable NR to work with established industrial Ethernet protocols. As TSN emerges as the foundation of the industrial Ethernet protocols, a flagship Release 16 feature is “NR and TSN integration”.

NR in Release 16 will support accurate time reference provisioning to the UE, in order to synchronize TSN devices on the UE side wirelessly with the TSN working time domain on the network side. Configured grant scheduling and UE multiplexing and pre-emption are proposed to be enhanced to more efficiently cope with the mixed TSN traffic scenario. PDCP duplication is designed to handle reliability provisioning more efficiently. Ethernet header compression is being studied for overhead reduction in RAN. Layer-1 URLLC enhancements are also being considered in Release 16 to further reduce latency, improve reliability and spectral efficiency, and improve handling of multiplexing uplink control and data from different services types (e.g. control for URLLC multiplexed with data from eMBB, or vice versa).

With TSN integration and further URLLC enhancements, NR Release 16 will make great strides toward enabling smart wireless manufacturing and ushering in a new era in industry digitalization and transformation.

Industrial Communication Technologies and Protocols Alongside TSN and 5G

It is a widely accepted thinking that TSN and 5G will be the fundamental connectivity technologies for future factories and other industrial use cases. Nevertheless, most industrial players do not start their industrial IoT story from scratch in a greenfield deployment. Rather, many industrial processes already involve connected devices using their own industry defined connectivity mechanisms. These deployments are commonly referred to as brownfield. While most brownfield deployments (94%) are wired, also many wireless solutions exist, especially for data collection. Industry is conservative and already made investments are guarded. Thus, many times a new technology needs to be introduced as a complementary solution to the already existing infrastructure at the industrial site unless significant added value can be shown.

The protocol stack for industrial IoT can look very different depending on choices on different protocol layers. FIG. 96 depicts some possible protocol alternatives on the different layers mapped to the Open Systems Interconnect (OSI) protocol stack layers.

To get a complete picture, this chapter introduces both wired and wireless communication technologies that are used today. Regarding the use of 4G and 5G for brownfield deployments, two aspects are important and covered:

Interworking with legacy wired technologies (like e.g. Profinet) Competing with other wireless technologies (like any IEEE 802.11 technology)

Furthermore, OPC-UA and SECS/GEM are introduced as two communication protocols being used in factory automation today and assumed to play a major role in the future.

With regards to the physical and medium access layers, many wired communication technologies dedicated to industrial usage have been developed in the past. Initially so-called Fieldbus technologies have been used as e.g. standardized in IEC 61158. Nowadays a shift towards industrial Ethernet solutions has happened and Profinet is one example of such. A main trait of these technologies is that they are designed to deliver data under tight time constraints, 1 ms or less. A disadvantage of Fieldbuses and some industrial Ethernet variants is a general incompatibility with each other and the need for special hardware to run these technologies beyond standard office-Ethernet equipment. Time-sensitive networking (TSN) is a set of IEEE standards, which add reliability and deterministic low latency capabilities on top of standardized Ethernet (IEEE 802.3). It is the ambition to establish a common standard into the splintered wired communication technology market for industries. A lot of industrial equipment vendors right now have started or at least indicated to move to TSN for their portfolio.

Industrial Ethernet has become quite popular and gains market share over legacy fieldbus technologies as Ethernet has also become a major communication standard in other domains. One reason might be the cheap and common parts and cables etc. It was already mentioned that different industrial Ethernet technologies are incompatible and don't allow an interworking without the use of special gateways or similar additional equipment. This is because they use different concepts to satisfy the requirements for industrial use cases. Nevertheless, there are some common facts for industrial Ethernet:

Industrial Ethernet is almost always ‘switched Ethernet’ 100 Mbit/s and full-duplex links Different topologies possible (line, star, ring etc.) but might be strictly defined by the technology Redundancy methods (e.g. Parallel Redundancy Protocol (PRP)) Master/controller-slave architecture Functions to detect communication errors (like timers and counters for packet losses)

FIG. 97 shows the concept of industrial Ethernet and how it is built-up on standard Ethernet. On layer 2, some industrial Ethernet technologies are based on time scheduled transmissions (like Profinet RT) to achieve deterministic latencies in the network. The network cycle time is a metric that is widely used to promote and compare technologies—the lower the network cycle time that is supported, the better. Usually the network cycle time is the minimum application cycle time that is supported (i.e. the application transmits a certain message in every network cycle). Very challenging use cases require incredible small application cycle times below 50 microseconds, to achieve a sufficient accuracy for e.g. motion control. EtherCat for example defines a new Ethernet layer 2 to achieve very low network cycle times.

As can be seen in FIG. 97 , Profinet has different flavors:

Profinet CBA (component based automation)—only for process automation with less stringent transmission characteristics and requirements Profinet-IO RT (realtime) Profinet-IO IRT (isochronous realtime)—this variant supports application cycle times down to 31.25 microseconds (by using network cycle times of 31.25 microseconds)

FIG. 98 shows an example of time scheduled transmissions in Profinet—the figure depicts one network cycle that is repeated periodically. Therein the network access time is shared between a cyclic IRT phase and a cyclic RT phase both to provide strict QoS and a Non-RT phase, which is equivalent to a best effort phase without guarantees on QoS. Profinet uses a time synchronization protocol like IEEE1588 to establish a common notion of time between all nodes. For very strict applications there might be no RT or Non-RT phase involved. IRT communication is always pure layer 2 communication, not using UDP/TCP/IP. A Profinet IRT frame is illustrated in FIG. 99 .

The network management in case of Profinet (as well as for other technologies) is manually pre-configured and usually no devices can be added on-the-fly—so plug-and-play is mostly not possible, but instead there is some expertise needed to set up these networks which is definitely a pain-point for industries.

Industrial Ethernet equipment differs from standard Ethernet as well:

Specific switches—rugged, QoS optimized, highly available implementation Most technologies require specific ASICs, some are based on software, some vendors sell also multi-technology ASICs (e.g. HMS, Hilscher, AD) Usually PLCs offer different communication modules to support multiple technologies Devices (from sensors to robots) usually offer only a limited set of technology interfaces

Some of the industrial Ethernet technologies will probably disappear and being replaced sooner or later by TSN products. Nevertheless, the product life cycles are very long in industries. TSN adopts many features also used in existing industrial Ethernet technologies. Furthermore, organizations behind Profinet and EtherCat already published whitepapers explaining how an operation alongside TSN will work. They might see TSN as a common infrastructure where Profinet and other technologies might coexist on.

The way industrial Ethernet is nowadays deployed is similar to islands. High QoS can only be guaranteed within such an island. One island is deployed using one communication technology, e.g. Profinet. Usually a Programmable Logic Controller (PLC) is used as a master of an island (e.g. a Profinet master). An island usually consists of a few devices on the same shopfloor only. The interworking of e.g. Profinet and cellular is especially relevant if one of these devices (e.g. the PLC) is virtualized (central link) or if one device (device-to-island) or a group of devices using a gateway (inter-island link) is separated from the island on the shopfloor. The interworking of Profinet and for example LTE has been showed already in some research proof-of-concept studies—it is possible if the applications cycle time is above a certain limit (e.g. 32 ms as an example), depending upon the configuration of the LTE network.

Requirements for the cellular network in terms of latency and packet error rate (PER) are not set by the communication technology like e.g. Profinet but the applications using them, or the application cycle times used respectively. Usually the lowest supported network cycle time is a KPI for industrial communication technologies. Although Profinet IRT supports network cycle times down to 31.25 usec it is also being used for applications with much lower requirements (i.e. application cycle times much higher that this). Profinet IRT can be used for application cycle times up to 4 ms. In case of Profinet, the RT version that supports only higher network cycle times seems anyway more relevant, at least was always used in any trials with industrial partners.

Other wireless solutions are trying to enter the field at the same time with 5G. One interesting technology is MulteFire, which is marketed heavily for industrial connectivity. MulteFire as a technology is very similar to LTE but only runs on unlicensed spectrum, so the benefits of scheduling and mobility within the system are there. Device availability is a challenge for MulteFire at current time. WiMAX has partly been used as wireless technology in industrial but is challenged due to the low economy of scale.

Industrial grade Wi-Fi has a small footprint in connecting industrial devices. Reliability and latency issues are addressed through implementation. No global certification exists, but rather the solutions are vendor specific and do not interoperate. More commonly, regular Wi-Fi is deployed in industrial spaces to allow employee Internet access from laptops, tablets and mobile phones. This connectivity is valuable and important for shop floor personnel.

FIG. 100 depicts an estimated difference between Wi-Fi, MulteFire, LTE and 5G NR with regards to increasing reliability demands and increasing end to end (E2E) latency demands. Example use cases are placed on the figure to show what kind of requirements roughly need to be fulfilled for each.

Wireless sensor networks are used to collect sensor data and monitor machinery. Industrial Bluetooth implementations exist as vendor specific solutions. Typically, Bluetooth is used as a connectivity for personnel to acquire reading from machinery when at close distance. There is increasing interest in deploying gateways for continuous connectivity. Also, many different variants of the IEEE802.15.4 protocol exist for industrial use. Most well-known are WirelessHART and ISA100.11a, which are defined and certified by industry players. 6TiSCH is being standardized by the IETF to bring determinism and reliability into the IEEE802.15.4 radio interface.

10-Link Wireless standard might be interesting as well, as it is said to achieve a PER of 10{circumflex over ( )}-9 and can support down to 5 ms cycle time. It has a limited scalability, however, and is limited in communication range.

MulteFire is LTE based technology which fully operates in the unlicensed spectrum. The main goal of MulteFire is to provide LTE-like performance with Wi-Fi-like ease of deployment in unlicensed spectrum. Compared to eLAA, the MulteFire RAN was designed to have independent operation. In particular, MulteFire performs all the control signaling and data signaling in unlicensed spectrum. Today MulteFire also includes eMTC-U and NB-IoT-U as Radio Access Technology (RAT) to support a wide range of applications from mobile broadband to machine type communication.

MulteFire (MF) uses principles of carrier selection, discontinuous transmission and listen before talk (LBT) that are based on 3GPP release 13 and 14 LAA. MulteFire targets 5.0 GHZ global spectrum and utilizes the Release-13 LBT procedure with some additions. Compared to LTE protocol stack, MF is unique in UL, DL physical layer, DRS transmission, SIB-MF broadcasts and its content, RACH procedures and has additional S1, X2 information signaling.

MulteFire 1.0 was further extended with additional features such as Grant Uplink Access, Wideband Coverage Extension (WCE), Autonomous Mobility (AUM), sXGP 1.9 GHZ support, eMTC-U and NB-IoT functionality. These features target more industrial deployments and support for machine type communications.

Grant Uplink Access further reduces UL control signaling overhead, which works very well in low load scenarios. This feature is based on the 3GPP feature Autonomous UL as defined in 3GPP Release 15. The WCE feature aims to increase coverage with up to 8 dB compared to legacy MF MBB operations. Compared to licensed spectrum, LBT and few measurements for RRM and RLF makes mobility very challenging. To address this, MF has specified AUM to deal with fast changing channel quality and handover, in which UE and potential eNBs can be pre-configured with handover related parameters. In particular, UEs may be configured with AUM related mobility assistance information for up to 8 AUM cells. These cells are basically potential candidate target cells, which have been prepared with potential UE context. Parameters which are shared to the UE includes frequency and physical Cell ID (PCI) of the candidate target cells.

To support massive IoT use cases, MF adapted the 3GPP Rel-13 eMTC technology based on 1.4 MHZ carrier bandwidth applied to the 2.4 GHZ frequency band. However, in the 2.4 GHZ frequency band, regulations are unique to USA, Europe, Japan and China. Among this, ETSI in Europe has strict rules and to adhere the regulations, frequency hopping mechanism was adopted. To enable eMTC-like performance, a new time-frequency frame structure is defined which has two fixed time-periods, first time-period being anchor channel and second being a data channel dwell. The data channel usually contains UL/DL transmission which are preceded by LBT and it always starts with a DL transmission. The anchor channel always remains on the same channel, several anchor channels are defined out of which eNB can select one of them to transmit. Data channel dwells transmission using frequency hopping, it is done by splitting 82.5 MHZ into 56 channels, with spacing of 1.4 GHZ between hoping channels. Specifications are currently being finalized to further extend Rel-13 NB-IoT support in unlicensed bands.

The IEEE 802.11 technology family, commonly referred to as Wi-Fi, is a popular technology to provide wireless Internet access in our homes. The industrial grade Wi-Fi solutions listed in the previous section are typically modifications to the IEEE 802.11 Wi-Fi. Industrial grade Wi-Fi is usually based on IEEE 802.11 Wi-Fi certified chipsets, with primarily stripped-down MAC layer. In particular, the LBT mechanism in Wi-Fi, albeit necessary for spectrum regulations, is often removed. The problem with the industrial grade Wi-Fi is interoperability as each industrial grade Wi-Fi is developed independently of the others. In contrast, IEEE 802.11 is a well-known standard, and you can expect products from different vendors to operate well between each other. We will in this section briefly consider a few mechanisms: channel access (largely impacts latency), quality of service (impacts priorities), and link adaptation (spectrum efficiency).

To understand the channel access of Wi-Fi, we must understand the background to some of the design principles in unlicensed bands. In unlicensed bands, as opposed to licensed bands, there is typically no physical controlling entity. There are a set of spectrum rules, but anyone adhering to these rules have the same right and priority to access the wireless medium. Therefore, a major design principle in unlicensed bands is the uncoordinated, competition-based channel access. This is called CSMA/CA—Carrier Sense Multiple Access with Collision Avoidance. The fundamental idea is that there is a random number associated with each channel access, the random number deciding a backoff time. For each failed channel access, this random number becomes larger. The result of this channel access is that round-trip latencies contain a random factor. When the wireless medium is to a large extent unused, the latency is very low, but when the wireless medium is very occupied, the latency can become very large. In industrial scenarios this uncertainty in latency is a concern. We show a typical channel access and data transmission in FIG. 101 . The channel access in Wi-Fi is the main reason why guaranteed latency is not possible, and this is a feature necessary to comply with the regulations. The strength of cellular technologies is that they are designed for exclusive use of the spectrum, meaning that guaranteed latency can be obtained.

In addition to the random backoff, there is in Wi-Fi an interframe spacing time (IFS). There are 3 main interframe spacing times: short IFS (SIFS), PCF (Point Coordination Function) IFS (PIFS), DCF (Distributed Coordination Function) IFS (DIFS). In summary, IFS<PCF<DIFS, where IFS are used for special response frames, i.e. ACK. PCF is used for certain priority frames, and DIFS for standard frames.

Wi-Fi has a quality of service (QoS) mechanism called Enhanced Distributed Channel Access (EDCA). EDCA is mainly based on adjusting the random backoff time when performing channel access, but it also introduces a new IFS, called arbitration IFS (AIFS). A higher priority will in average get priority access due to reduced backoff times. However, note that there is still randomness in each channel access, and no guarantees can be provided. There are 4 priority classes introduced in EDCA: voice, video, best effort, and background. An illustration of how the different priority classes may obtain channel access is shown in FIG. 102 . Note that each priority class has an individual IFS and that the random backoff is different.

The link adaptation in Wi-Fi is based on full data re-transmissions. If a packet fails to be decoded, the packet is sent again (possibly with another coding and modulation). Note that the data packets in Wi-Fi are self-contained, and if a packet fails all information is typically trashed. This is a main disadvantage compared to LTE or NR, where the soft information received during the initial transmission is combined with the soft information received during the retransmission. The gain by soft combining is in the order of 3-6 dB, depending on if the retransmission is a repetition of the previous coded bits (called Chase combining) or if additional parity bits are transmitted (called incremental redundancy).

The coding and modulation chosen is typically selected by the Minstrel algorithm. The Minstrel algorithm works by keeping trial and error statistics over packets sent with different coding and modulations and attempts to maximize the throughput. The algorithm works well in static environments with little to no interference but suffers when channel characteristics change fast. This results in that Minstrel is typically slow to adopt to an improved channel, as shown in FIG. 103 , which illustrates a simulation of the Minstrel algorithm in a single-link radio simulator.

Industrial services above the IP/Ethernet layer use a variety of protocols to accomplish the tasks at hand. The reference introduces protocols such as Constraint Application Protocol (CoAP), Hypertext Transfer Protocol (HTTP) and HTTP/2, Message Queue Telemetry Transport (MQTT), Open Connectivity Foundation (OCF), Data Distribution Service (DDS) for real-time systems, etc. In the following we give a short introduction to one of the main protocols in the industrial area that is OPC UA. Finally, we take a brief look at SECS/GEM used in the semiconductor industry.

As introduced above, usually an interworking between legacy industrial communication technologies is not possible. As a result, end customers and device manufacturers are faced with a multitude of technologies that need to be produced, run, diagnosed, maintained and kept in stock. While the availability of products and services is largely satisfactory, dealing with multiple solutions generates prohibitive costs and limits IoT capability. The OPC-UA (Open Platform Communication-Unified Architecture) tries to address these problems. OPC-UA is the next generation of OPC technology. It should provide better security and a more complete information model than the original OPC, “OPC Classic.” OPC Classic is a well-established protocol for automation from (primarily) Microsoft. OPC-UA is said to be a very flexible and adaptable mechanism for moving data between enterprise-type systems and the kinds of controls, monitoring devices and sensors that interact with real-world data. OPC-UA is platform independent and ensures the seamless flow of information among devices from multiple vendors. The OPC Foundation is responsible for the development and maintenance of this standard. FIG. 104 illustrates the OPC-UA protocol stack.

For the use in TSN, the OPC-UA standard is adapted to be more deterministic and support certain TSN features. FIG. 105 illustrates the use of OPC-UA over TSN. In general, a TSN network infrastructure is simultaneously able to carry all types of industrial traffic, from hard real-time to best effort, while maintaining the individual properties of each type. The OPC-UA TSN initiative uses a publisher-subscriber communication model and the use of OPC-UA without TCP/UDP/IP.

OPC-UA is also assumed to be used as a configuration-protocol in TSN.

Regarding the time line of OPC-UA and TSN: in Q4 2018 there was an announcement that most industrial automation suppliers (incl. Siemens, Bosch, Cisco, ABB, Rockwell, B&R, TTTEch etc.) are supporting the ‘OPC UA including TSN down to field level’ initiative. It is said that the work will be closely aligned to the work in IEC/IEEE 60802, which defines a common profile for TSN for industrial automation. It is currently planned to conclude the work in 60802 during 2021 which might be probably the same date to publish some final documents describing OPC-UA and TSN.

SEMI (previously known as Semiconductor Equipment and Materials International) standards define the SEMI Equipment Communications Standard/Generic Equipment Model (SECS/GEM) that in turn provides a protocol interface for equipment to host data communications. SEMIs purpose is to serve the manufacturing supply chain for electronics production in semiconductor fabrication plants, aka, fabs.

SECS/GEM is an alternative to OPC UA used in the semiconductor industry. The specification defined how equipment communicates with host in the factory.

Specific Applications to Industrial IoT

Following are detailed discussions of several applications of the technology and techniques described above in the Industrial IoT context. It will be appreciated, of course, that these applications are not limited to this context. Several different applications are described, including techniques for scheduling resources, handling time-sensitive data streams in a 5G network, detecting system support for TSN, handling different timings from different, and data compression and decompression. Further, a few combinations of these techniques are described. It should be appreciated, however, that any of these techniques may be combined with any of the other techniques, as well, as with any one or more of the other techniques and technologies described above, to address the special needs of a factor or other industrial setting.

Scheduling Resources in the RAN

As discussed above, while 5G is based on wireless communications using Long-Term Evolution (LTE) and/or New Radio (NR) technologies, TSN is based on the IEEE 802.3 Ethernet standard, a wired communication standard that is designed for “best effort” quality of service (QoS). TSN describes a collection of features intended to make legacy Ethernet performance more deterministic, including time synchronization, guaranteed low-latency transmissions, and improved reliability. The TSN features available today can be grouped into the following categories (shown below with associated IEEE specifications):

Time Synchronization (e.g., IEEE 802.1AS); Bounded Low Latency (e.g., IEEE 802.1Qav, IEEE 802.1Qbu, IEEE 802.1Qbv, IEEE 802.1Qch, IEEE 802.1Qcr); Ultra-Reliability (e.g., IEEE 802.1CB, IEEE 802.1Qca, IEEE 802.1Qci); Network Configuration and Management (e.g., IEEE 802.1Qat, IEEE 802.1Qcc, IEEE 802.1Qcp, IEEE 802.1CS).

The configuration and management of a TSN network can be implemented in different ways, as illustrated in FIGS. 106 , 107 , and 108 . More specifically, FIGS. 106 - 108 are block diagrams that respectively illustrate Distributed, Centralized, and Fully Centralized Time-Sensitive Networking (TSN) configuration models, as specified in IEEE Std. 802.1Qbv-2015. Within a TSN network, the communication endpoints are called “Talker” and “Listener.” All the switches and/or bridges between a Talker and a Listener can support certain TSN features, such as IEEE 802.1AS time synchronization. A “TSN domain” includes all nodes that are synchronized in the network, and TSN communication is only possible within such a TSN domain.

The communication between Talker and Listener is in streams. Each stream is based on data rate and latency requirements of an application implemented at both Talker and Listener. The TSN configuration and management features are used to set up the stream and to guarantee the stream's requirements across the network. In the distributed model from FIG. 106 , the Talker and Listener can, for example, use the Stream Reservation Protocol (SRP) to setup and configure a TSN stream in every switch along the path from Talker to Listener in the TSN network.

Nevertheless, some TSN features may require a central management entity called Centralized Network Configuration (CNC), as shown in FIG. 107 . The CNC can use, for example, Netconf and YANG models to configure the switches in the network for each TSN stream. This also facilitates the use of time-gated queueing (defined in IEEE 802.1Qbv) that enables data transport in a TSN network with deterministic latency. With time-gated queueing on each switch, queues are opened or closed according to a precise schedule thereby allowing high-priority packets to pass through with minimum latency and jitter. Of course, packets may arrive at a switch ingress port before the gate is scheduled to be open. The fully centralized model shown in FIG. 108 also includes a Centralized User Configuration (CUC) entity used as a point of contact for Listener and Talker. The CUC collects stream requirements and endpoint capabilities from the devices and communicates with the CNC directly. Further details about TSN configuration are given in IEEE 802.1Qcc.

FIG. 109 shows a sequence diagram of an exemplary TSN stream configuration procedure based on the fully centralized configuration model shown in FIG. 108 . The numbered operations shown in FIG. 109 correspond to the description below. Even so, the numerical labels are used for illustration rather than to specify an order for the operations. In other words, the operations shown in FIG. 109 can be performed in different orders and can be combined and/or divided into other operations than shown in the figure.

1 CUC can receive input from, e.g., an industrial application and/or engineering tool (e.g., a programmable logic control, PLC) that specifies devices and/or end stations to exchange time-sensitive streams. 2 CUC reads the capabilities of end stations and applications in the TSN network, including a period/interval of user traffic and payload sizes. 3 Based on this above information CUC creates StreamID as an identifier for each TSN stream, a StreamRank, and UsertoNetwork Requirements. In the TSN network, the stream ID is used to uniquely identify stream configurations and to assign TSN resources to a user's stream. The streamID consists of the two tuples: 1) MacAddress associated with the TSN Talker; and 2) UniqueID to distinguish between multiple streams within end stations identified by MacAddress. 4 CNC discovers the physical network topology using for example Link Layer Discovery Protocol (LLDP) and any network management protocol. 5 CNC uses a network management protocol to read TSN capabilities of bridges (e.g., IEEE 802.1Q, 802.1AS, 802.1CB) in the TSN network. 6 CUC initiates join requests to configure network resources at the bridges for a TSN stream from one Talker to one Listener. 7 Talker and Listener groups (group of elements specifying a TSN stream) are created by CUC, as specified in IEEE 802.1Qcc, 46.2.2). CNC configures the TSN domain, and checks physical topology and if the time sensitive streams are supported by bridges in the network. CNC also performs path and schedule computation of streams. 8 CNC configures TSN features in bridges along the computed path in the (e.g., configuration of the transmission schedule, as explained further below). 9 CNC returns status (success or failure) for resulting resource assignment for streams to CUC. 10 CUC further configures end stations to start the user plane (UP) traffic exchange as defined initially between Listener and Talker.

In the distributed configuration model as illustrated in FIG. 106 , there is no CUC and no CNC. The Talker is therefore responsible for initiation of a TSN stream. Since no CNC is present, the bridges configure themselves, which does not allow use of time-gated queuing mentioned above. In contrast, in the centralized model shown in FIG. 107 , the Talker is responsible for stream initialization but the bridges are configured by CNC.

3GPP-standardized 5G networks are one solution for connecting wireless devices and/or end stations to an 802.1 TSN network. In general, the 5G network architecture consists of a Next Generation radio access network (NG-RAN) and a 5G core network (5GC). The NG-RAN can comprise a set of gNodeB's (gNBs, also referred to as base stations) connected to the 5GC via one or more NG interfaces, whereas the gNBs can be connected to each other via one or more Xn interfaces. Each of the gNBs can support frequency division duplexing (FDD), time division duplexing (TDD), or a combination thereof. Devices—also referred to as user equipment (UE)—communicate wirelessly with the 5G network via the gNBs.

FIG. 110 is a block diagram illustrating an exemplary division of the 5G network architecture into control plane (CP) and data or user plane (UP) functionality. For example, a UE can communicate data packets to a device and/or application on an external network (e.g., the Internet) by sending them via a serving gNB to a user plane function (UPF), which provides an interface from the 5G network to other external networks. CP functionality can operate cooperatively with the UP functionality and can include various functions shown in FIG. 110 , including an access management function (AMF) and a session management function (SMF).

Even so, there are several challenges and/or issues needing to be solved for the proper interworking of 5G and TSN networks. In particular, there are several challenges related to configuring a 5G network to handle data communications to/from an external network (e.g., a TSN network) that are subject to a time-critical schedule determined by the external network rather than the 5G network.

FIG. 111 is a block diagram illustrating an exemplary arrangement for interworking between the 5G network architecture shown in FIG. 110 and an exemplary fully centralized TSN network architecture. In the following discussion, a device connected to the 5G network is referred to as 5G endpoint, and a device connected to the TSN domain is referred to as TSN endpoint. The arrangement shown in FIG. 111 includes a Talker TSN endpoint and a Listener 5G endpoint connected to a UE. In other arrangements, a UE can instead be connected to a TSN network comprising at least one TSN bridge and at least one TSN endpoint. In this configuration, the UE can be part of a TSN-5G gateway.

Both 5G and TSN networks utilize specific procedures for network management and configuration, and specific mechanisms to achieve deterministic performance. To facilitate end-to-end deterministic networking for industrial networks, these different procedures and mechanisms must work together cooperatively.

As described in IEEE 802.1Qbv-2015, TSN provides specific time-aware traffic scheduling to facilitate deterministic low latency for industrial application, where cycle time is known in advance. This traffic scheduling is based on time-aware gates that enable transmissions from each queue according to a predefined time scale. FIG. 112 is a block diagram illustrating gate-based transmission selection among traffic queues based on gates, as specified in IEEE Std. 802.1Qbv-2015. For a given queue, the transmission gates can be in two states: open or closed.

Furthermore, each transmission gate relates to a traffic class associated with a specific queue, with potentially multiple queues associated with a given port. At any instance of time, a gate can be either turned on or off. This mechanism is time-aware and can be based on, e.g., a PTP application within a TSN bridge or a TSN end station. This mechanism allows execution of a gate control list to be precisely coordinated across the network, facilitating tightly-scheduled transmissions for a given class of traffic. Herein, a transmission schedule can be defined as a schedule that indicates when transmissions are to occur in time. Also, a time-critical transmission schedule can be defined as a schedule that indicates when transmissions of a Time-Sensitive Network (TSN) are to occur in time.

As described above in relation to FIG. 109 , the information about TSN stream schedules are is calculated by a CNC entity in the fully-centralized TSN model, based on the user to network requirements (e.g., IEEE 802.1Qcc § 46.2.3.6 of) provided by Talker and/or Listener (and/or via the CUC entity). In addition, standard management objects (e.g., defined in IEEE 802.1Qvc) and a remote network management protocol are used by the CNC to configure transmission schedules on TSN bridges (operation 8 in FIG. 109 ).

Nevertheless, these features are specific to TSN networks and do not take into account interworking 5G network architecture, such as illustrated in FIG. 111 . For example, 5G networks do not provide any mechanism for elements (e.g., UEs, gNBs, etc.) to take into account time-critical transmission schedules established by external networks (e.g., TSN networks) when scheduling transmissions over the wireless interface between UE and gNB. For example, even if such a time-critical transmission schedule is known to a UE (e.g., connected to a TSN endpoint), there is no mechanism for the UE to inform the gNB of such a schedule. Furthermore, there is no mechanism that enables the gNB or UE to understand and process scheduling requests, coming from the 5G network.

Exemplary embodiments of the present disclosure address these and other problems and/or shortcomings of prior solutions by providing novel techniques for predefined time scheduling for specific users and/or QoS flows based on time-aware transmission schedules (e.g., from external networks) to meet specific bounded latency requirements. For example, these techniques can provide mechanisms for a UE (or network node, e.g., gNB) to be informed of such a transmission time schedule and to inform the network node (or UE) of the schedule. In this manner, such novel techniques can provide various benefits including cooperative interworking between cellular (e.g., 5G) and TSN networks that utilize different schedulers and/or scheduling mechanisms, thereby facilitating bounded latency of time-critical transmissions between Talker/Listener endpoints via 5G networks.

FIG. 113 is a block diagram illustrating an exemplary communication scenario between two TSN talker/listener units via 5G and TSN networks, according to some exemplary embodiments of the present disclosure. In this scenario, a UE is connected to a TSN talker/listener, which in turn can be connected to plant equipment (e.g., a robot control) that is required to run an application according to a predefined cycle time. One challenge in this scenario is to facilitate timely transmission of the TSN stream packets from the gNB to the UE, according to the bounded latencies required by the equipment and/or application.

FIG. 114 shows a sequence diagram of an exemplary method and/or procedure for configuring timely transmission of TSN stream packets via the network configuration shown in FIG. 113 , according to these exemplary embodiments. The numbered operations shown in FIG. 114 correspond to the description below. Even so, the numerical labels are used for illustration rather than to specify an order for the operations. In other words, the operations shown in FIG. 114 can be performed in different orders and can be combined and/or divided into other operations than shown in the figure.

In operation 11 , the CUC sends to the CNC a join request for a user to join the TSN network. For example, this request can be based on and/or in response to a programmable logic control (PLC) application requesting to schedule a TSN stream between a sensor (Talker) and a PLC controller (Listener). In operation 12 , the CNC computes a transmission schedule based on the specific requirements of the TSN stream identified in operation 11 .

In operation 13 , the CNC configures managed objects of TSN switches that are in the path between the sensor and PLC controller. Exemplary managed object to be configured for enhanced time-aware scheduling are described in IEEE 802.1Qbv-2015 § 12. In exemplary embodiments, the CNC treats the 5G network as a TSN switch in the path, and therefore requests the 5G core network (5GC) to configure resources for this TSN stream. For example, this can be done by the CNC sending to an access management function (AMF, see FIGS. 110 - 111 ) the cycle times and gate control lists for traffic classes within the TSN stream.

In operation 14 , the receiving entity (e.g., AMF) in the 5GC can translate the requested TSN stream requirements (e.g., cycle time, gate control list, etc.) to QoS requirements for the UE that is connected to the TSN Talker/Listener (e.g., sensor). In addition, the AMF can translate the requested TSN stream requirements into a time window and periodicity for the gNB(s) to which the UE will transmit and/or receive this TSN stream.

In some embodiments, operation 14 can involve various sub-operations. For example, the UE and a PDU session corresponding to the TSN stream can be identified, and a mapping between traffic classes within this TSN stream and QoS flows of the UE can be identified. For each QoS flow (which can correspond to one or multiple traffic classes), a certain QoS requirement can be indicated to the gNB. In some embodiments, this indication to the gNB can include an indicator of a time-window during which a packet of the QoS flow should be guaranteed to be transmitted. This time window can be indicated, e.g., by providing an absolute time reference for the time window start together with a length of the window (e.g., as a latency bound). For example, the absolute time reference can be indicated as an offset to a certain absolute reference time such gNB subframe (SFN) timing or a universal time coordinate (UTC), such as provided by a global navigation satellite system (GNSS, e.g., GPS). In some embodiments, the indication to the gNB can also include a periodicity (or period) of the time window. This can be included, e.g., if the TSN stream comprises multiple transmission events that occur according to a periodic schedule.

By indicating this time-window information per QoS flow of the UE, multiple traffic classes of a TSN stream or multiple TSN streams can be independently served. In other words, this information facilitates the affected gNB(s) to reserve radio resources for each of the QoS flows during the respective time windows associated with those QoS flows. For example, this can facilitate the gNB(s) to map the various QoS flows to different radio bearers and to apply the resource allocation/reservation per radio bearer. Herein, a radio bearer takes the usual definition from the 3rd Generation Partnership Project (3GPP).

In operation 14 , after determining the information as discussed above, the AMF sends an indication and/or request the gNB(s) to confirm that the QoS, time window, and/or periodicity requirements can be met. In operation 15 , after receiving the request/indication sent in operation 14 , the gNB (or gNBs, as the case may be) determines whether it can serve this additional QoS flow with the indicated time-window requirement. For example, in making this determination, the gNB can consider resources used for current and estimated traffic load, capabilities of the UE (e.g., spectral efficiency, supported transmission/reception modes, etc.), channel quality between the RAN and the UE, and whether (and/or how many) additional guaranteed resources need to be allocated for the UE. After making this determination, the gNB responds to the 5GC function (e.g., AMF) by accepting the request (“yes”) or declining the request (“no”). In some embodiments, when declining the request, the gNB can indicate an alternative time window (e.g., by an offset to the requested time window) during which the gNB could accept a corresponding request. In situations where the gNB accepts the request, the gNB can also reserve any additional resources identified as required to meet the requested transmission schedule.

In operation 16 , after receiving the response from the gNB(s), the 5GC function may then translate this response—which is based on per QoS flow mapping—to a traffic flow/TSN stream level of granularity, and provides a response to the TSN CNC. The response may be in a format that can be decoded by the TSN CNC. In operation 17 , after receiving this response, the CNC provides to the CUC a corresponding response to the join request received in operation 11 . In operation 18 , after receiving the join response from the CNC, the CUC further configures all Talker and Listener end station associated with the original request. In some embodiments, the CUC can also request the 5GC to initiate a connection to the UE, whereas in other embodiments, the 5GC or it might use a default and/or already-existing PDU session.

FIG. 115 is a block diagram illustrating another exemplary communication scenario between a TSN talker/listener unit and a virtualized controller via a 5G network, according to other exemplary embodiments of the present disclosure. In this scenario, a TSN network is connected to UE, which acts a gateway to connect a Talker/Listener end station over a wireless link to the 5G network. One challenge in this scenario is to facilitate timely transmission of the TSN stream packets from the UE to the gNB, according to the bounded latencies required by the schedule computed by a CNC in the TSN network.

FIG. 116 shows a sequence diagram of an exemplary method and/or procedure for configuring timely delivery of TSN stream packets via the network configuration shown in FIG. 115 , according to these exemplary embodiments. The numbered operations shown in FIG. 116 correspond to the description below. Even so, the numerical labels are used for illustration rather than to specify an order for the operations. In other words, the operations shown in FIG. 116 can be performed in different orders and can be combined and/or divided into other operations than shown in the figure.

In operation 21 , the CNC calculates the transmission schedule based on the requirements provided by CUC and sends it to the TSN interface of the 5G network, which is in this case the UE. In operation 22 , the UE creates and sends a message requesting uplink (UL) radio resources according to the transmission schedule provided by the CNC, which can be included in the message. For example, the UE can send the message to the AMF in the 5GC. In operation 23 , after receiving this message, the AMF retrieves the UE profile from a user data management (UDM) function in the 5GC and, based on this information, determines to which gNB(s) the UE is connected. In operation 24 , the AMF sends a request to the gNB(s) to enable the TSN QoS feature towards the UE based on the transmission schedule, which can be included in the request. In some embodiments, the AMF can also send a modified time reference to the other Talker/Listener (e.g., a virtualized controller) connected to the 5G network (operation 24 a ).

In operation 25 , the receiving gNB(s) can perform operations substantially similar to those described above with reference to operation 15 of FIG. 114 , but with respect to the uplink rather than the downlink. After receiving the response from the gNB(s) sent in operation 25 , the AMF can respond (operation 26 ) to the request for resources received from the UE in operation 22 . Similar to operation 16 shown in FIG. 114 , the AMF can translate the gNB response—which is based on per QoS flow mapping—to a traffic flow/TSN stream level of granularity and provides a response in this format to the UE. In operation 27 , the UE can forward this information to the CNC, in response to the requested transmission schedule received in operation 21 . As discussed above in relation to certain embodiments illustrated by FIG. 114 , if gNB declines the requested transmission schedule but offers an alternate time window that it can accept, the responses sent in operations 15 - 17 of FIG. 114 and operations 25 - 27 of FIG. 116 can include such an alternate time window, formatted and/or translated according to the protocols and/or requirements of the respective recipients.

As can be understood from the above description, these and other exemplary embodiments facilitate time-aware scheduling of transmissions in a cellular network (e.g., a 5G network) according to the time-sensitive (e.g., bounded latency) requirements of an external network, such as a TSN network. The exemplary embodiments facilitate such features through novel techniques for collecting (either via the UE or a network function such as an AMF) information about timing and periodicity associated with traffic provided an external network and forwarding such information to one or more base stations (e.g., gNBs) in the cellular network. In such case, the base station(s) can determine whether the external time-sensitive requirements of the requested traffic can be supported and, if so, utilize such information for scheduling uplink or downlink transmissions between the UE and the base station(s).

FIG. 117 is a flow diagram illustrating an exemplary method and/or procedure for scheduling resources in a radio access network (RAN) according to a transmission schedule associated with an external network, according to various exemplary embodiments of the present disclosure. The exemplary method and/or procedure shown in FIG. 117 can be implemented in a core network (e.g., 5GC) associated with the RAN (e.g., NG-RAN), such as by a core network node (e.g., AMF) shown in, or described in relation to, other figures herein. Furthermore, as explained below, the exemplary method and/or procedure shown in FIG. 117 can be utilized cooperatively with the exemplary method and/or procedures shown in FIGS. 118 and/or 119 (described below), to provide various exemplary benefits described herein. Although FIG. 117 shows blocks in a particular order, this order is merely exemplary, and the operations of the exemplary method and/or procedure can be performed in a different order than shown in FIG. 117 and can be combined and/or divided into blocks having different functionality. Optional operations are represented by dashed lines.

The exemplary method and/or procedure illustrated in FIG. 117 can include the operations of block 1210 , in which the network node can receive, from the external network, a transmission schedule associated with a time-sensitive data stream. Herein, a time-sensitive data stream can be a data stream of a Time-Sensitive Network (TSN). Thus, in some embodiments, the external network comprises a Time-Sensitive Network (TSN) such as described in the IEEE standards discussed herein. In such embodiments, the data stream can comprise a TSN stream, e.g., associated with a Talker and/or Listener end station in the TSN. In such embodiments, the transmission schedule can comprise cycle times and gate control lists for one or more traffic classes comprising the TSN stream.

The exemplary method and/or procedure can also include the operations of block 1220 , in which the network node can send, to the RAN, a request to allocate radio resources for communication of the data stream between the RAN and a user equipment (UE), wherein the request further comprises information related to the transmission schedule. In some embodiments, the information related to the transmission schedule includes one or more of the following: an identifier of the UE; identifiers of one or more quality-of-service (QoS) flows associated with the data stream; and a QoS requirement associated with each of the QoS flows. In some embodiments, each QoS requirement can comprise one or more time windows during which the data stream is required to be transmitted. In some embodiments, each QoS requirement comprises an initial time window and a periodicity that identifies subsequent time windows.

The exemplary method and/or procedure can also include the operations of block 1230 , in which the network node can receive, from the RAN, a response indicating whether radio resources can be allocated to meet the transmission schedule associated with the data stream. In some embodiments, according to sub-block 1235 , if the response indicates that radio resources cannot be allocated to meet the transmission schedule of the data stream, the response further comprises an indication of one or more further time windows during which radio resources can be allocated.

In some embodiments, the response can indicate whether the QoS requirement associated with each of the QoS flows can be met. In such embodiments, the exemplary method and/or procedure can also include the operations of block 1240 , in which the network node can determine whether the transmission schedule can be met based on the indication of whether the QoS requirement associated with each of the QoS flows can be met. In some embodiments, the exemplary method and/or procedure can also include the operations of block 1250 , in which the network node can send, to the external network, an indication of whether the transmission schedule can be met.

In some embodiments, the method can be performed by an access management function (AMF) in a 5G core network (5GC). In some embodiments, the transmission schedule can be received from the external network; and the radio resources are for downlink communication from the RAN to the UE. In some embodiments, the transmission schedule is received from the UE; and the radio resources are for uplink communication from the UE to the RAN.

FIG. 118 is a flow diagram illustrating an exemplary method and/or procedure for scheduling resources in a radio access network (RAN) according to a transmission schedule associated with an external network, according to various exemplary embodiments of the present disclosure. The exemplary method and/or procedure shown in FIG. 118 can be implemented in a RAN (e.g., NG-RAN) associated with a core network (e.g., 5GC), such as by a RAN node (e.g., gNB) shown in, or described in relation to, other figures herein. Furthermore, as explained below, the exemplary method and/or procedure shown in FIG. 118 can be utilized cooperatively with the exemplary method and/or procedures shown in FIGS. 117 and/or 119 (described above and below), to provide various exemplary benefits described herein. Although FIG. 118 shows blocks in a particular order, this order is merely exemplary, and the operations of the exemplary method and/or procedure can be performed in a different order than shown in FIG. 118 and can be combined and/or divided into blocks having different functionality. Optional operations are represented by dashed lines.

The exemplary method and/or procedure illustrated in FIG. 118 can include the operations of block 1310 , in which the network node can receive, from the core network, a request to allocate radio resources between the RAN and a user equipment (UE) for communication of a time-sensitive data stream, wherein the request further comprises information related to a transmission schedule associated with the data stream. In some embodiments, the external network comprises a Time-Sensitive Network (TSN); and the data stream comprises a TSN stream.

In some embodiments, the information related to the transmission schedule includes one or more of the following: an identifier of the UE; identifiers of one or more quality-of-service (QoS) flows associated with the data stream; and a QoS requirement associated with each of the QoS flows. In some embodiments, each QoS requirement can comprise one or more time windows during which the data stream is required to be transmitted. In some embodiments, each QoS requirement comprises an initial time window and a periodicity that identifies subsequent time windows.

The exemplary method and/or procedure illustrated in FIG. 118 can also include the operations of block 1320 , in which the network node can, based on the information related to the transmission schedule, determine whether radio resources can be allocated to meet the transmission schedule. In some embodiments, determining whether radio resources can be allocated to meet the transmission schedule can be further based on one or more of the following: resources needed for current or estimated traffic load, capabilities of the UE, channel quality between the RAN and the UE, and need for additional guaranteed resources to be allocated for the UE.

In some embodiments, if it is determined in block 1320 that radio resources cannot be allocated to meet the transmission schedule associated with the data stream, the exemplary method and/or procedure includes the operations of block 1330 , where the network node can determine one or more further time windows during which radio resources can be allocated. In some embodiments, if it is determined in block 1320 that radio resources can be allocated to meet the transmission schedule associated with the data stream, the exemplary method and/or procedure includes the operations of block 1340 , where the network node can map the one or more QoS flows to at least one radio bearer between the RAN and the UE, and reserve transmission resources for the at least one radio bearer.

The exemplary method and/or procedure also includes the operations of block 1350 , in which the network node can send, to the core network, a response indicating whether the radio resources can be allocated to meet the transmission schedule. In some embodiments, if it is determined in block 1320 that radio resources cannot be allocated to meet the transmission schedule, the response sent in block 1350 can also include an indication of the one or more further time windows determined in optional subblock 1330 . This is illustrated by optional subblock 1355 .

FIG. 119 is a flow diagram illustrating an exemplary method and/or procedure for scheduling resources in a radio access network (RAN) according to a transmission schedule associated with an external network, according to various exemplary embodiments of the present disclosure. The exemplary method and/or procedure shown in FIG. 119 can be implemented, for example, by a user equipment (UE, e.g., wireless device, IoT device, M2M device, etc.) in communication with a RAN (e.g., NG-RAN) that is associated with a core network (e.g., 5GC), such as shown in, or described in relation to, other figures herein. Furthermore, as explained below, the exemplary method and/or procedure shown in FIG. 119 can be utilized cooperatively with the exemplary method and/or procedures shown in FIGS. 117 and/or 118 (described above), to provide various exemplary benefits described herein. Although FIG. 119 shows blocks in a particular order, this order is merely exemplary, and the operations of the exemplary method and/or procedure can be performed in a different order than shown in FIG. 119 and can be combined and/or divided into blocks having different functionality. Optional operations are represented by dashed lines.

The exemplary method and/or procedure illustrated in FIG. 119 can include the operations of block 1410 , in which the UE can receive, from the external network, a transmission schedule associated with a time-sensitive data stream. In some embodiments, the external network comprises a Time-Sensitive Network (TSN) such as described in the IEEE standards discussed herein. In such embodiments, the data stream can comprise a TSN stream, e.g., associated with a Talker and/or Listener end station in the TSN. In such embodiments, the transmission schedule can comprise cycle times and gate control lists for one or more traffic classes comprising the TSN stream.

The exemplary method and/or procedure can also include the operations of block 1420 , in which the UE can send, to a core network associated with the RAN, a request to allocate radio resources for communication of the data stream between the UE and the RAN, wherein the request further comprises information related to the transmission schedule. In some embodiments, the information related to the transmission schedule comprises the transmission schedule.

The exemplary method and/or procedure can also include the operations of block 1430 , in which the UE can receive, from the core network, a response indicating whether radio resources can be allocated to meet the transmission schedule associated with the data stream. In some embodiments, if the response from the core network indicates that radio resources cannot be allocated to meet the transmission schedule of the data stream, the response further comprises an indication of one or more further time windows during which radio resources can be allocated. This is illustrated by optional subblock 1435 . In some embodiments, the request (block 1420 ) can be sent to, and the response (block 1430 ) can be received from, an access management function (AMF) in a 5GC.

In some embodiments, the exemplary method and/or procedure can also include the operations of block 1440 , in which the UE can send, to the external network, an indication of whether the transmission schedule can be met. In some embodiments, if the response received in block 1430 comprises an indication of one or more further time windows during which radio resources can be allocated (subblock 1435 ), the indication sent to the external network further includes information related to the one or more further time windows. This is illustrated by optional subblock 1445 .

FIG. 120 illustrates one example of a cellular communications system and/or network, comprising various devices and/or systems usable to implement any of the exemplary methods described herein. In the embodiments described herein, the cellular communications network 1500 is a 5G NR network. In this example, the cellular communications network 1500 includes base stations 1502 - 1 and 1502 - 2 , which in LTE are referred to as eNBs and in 5G NR are referred to as gNBs, controlling corresponding macro cells 1504 - 1 and 1504 - 2 . The base stations 1502 - 1 and 1502 - 2 are generally referred to herein collectively as base stations 1502 and individually as base station 1502 . Likewise, the macro cells 1504 - 1 and 1504 - 2 are generally referred to herein collectively as macro cells 1504 and individually as macro cell 1504 .

The cellular communications network 1500 can also include some number of low power nodes 1506 - 1 through 1506 - 4 that control corresponding small cells 1508 - 1 through 1508 - 4 . The low power nodes 1506 - 1 through 1506 - 4 can be small base stations (such as pico or femto base stations), Remote Radio Heads (RRHs), or the like. Notably, while not illustrated, one or more of the small cells 1508 - 1 through 1508 - 4 may alternatively be provided by the base stations 1502 . The low power nodes 1506 - 1 through 1506 - 4 are generally referred to herein collectively as low power nodes 1506 and individually as low power node 1506 . Likewise, the small cells 1508 - 1 through 1508 - 4 are generally referred to herein collectively as small cells 1508 and individually as small cell 1508 . The base stations 1502 (and optionally the low power nodes 1506 ) are connected to a core network 6150 .

The base stations 1502 and the low power nodes 1506 provide service to wireless devices 1512 - 1 through 1512 - 5 in the corresponding cells 1504 and 1508 . The wireless devices 1512 - 1 through 1512 - 5 are generally referred to herein collectively as wireless devices 1512 and individually as wireless device 1512 . The wireless devices 1512 are also sometimes referred to herein as UEs. Wireless devices 1512 can take on various forms, including those compatible with MTC and/or NB-IoT.

FIG. 121 is a schematic block diagram of a radio access node 2200 according to some embodiments of the present disclosure. The radio access node 2200 may be, for example, a base station (e.g., gNB or eNB) described herein in relation to one or more other figures. As illustrated, the radio access node 2200 includes a control system 2202 that further includes one or more processors 2204 (e.g., Central Processing Units (CPUs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs), and/or the like), memory 2206 , and a network interface 2208 . In addition, the radio access node 2200 includes one or more radio units 2210 that each includes one or more transmitters 2212 and one or more receivers 2214 coupled to one or more antennas 2216 . In some embodiments, the radio unit(s) 2210 is external to the control system 2202 and connected to the control system 2202 via, e.g., a wired connection (e.g., an optical cable). However, in some other embodiments, the radio unit(s) 2210 and potentially the antenna(s) 2216 are integrated together with the control system 2202 . The one or more processors 2204 operate to provide one or more functions of a radio access node 2200 as described herein. In some embodiments, the function(s) are implemented in software that is stored, e.g., in the memory 2206 and executed by the one or more processors 2204 .

FIG. 122 is a schematic block diagram that illustrates a virtualized embodiment of the radio access node 2200 according to some embodiments of the present disclosure. This discussion is equally applicable to other types of network nodes. Further, other types of network nodes may have similar virtualized architectures.

As used herein, a “virtualized” radio access node is an implementation of the radio access node 2200 in which at least a portion of the functionality of node 2200 is implemented as a virtual component(s) (e.g., via a virtual machine(s) executing on a physical processing node(s) in a network(s)). As illustrated, in this example, the radio access node 2200 includes the control system 2202 that includes the one or more processors 2204 (e.g., CPUs, ASICs, FPGAs, and/or the like), the memory 2206 , and the network interface 2208 and the one or more radio units 2210 that each includes the one or more transmitters 2212 and the one or more receivers 2214 coupled to the one or more antennas 2223 , as described above. The control system 2202 is connected to the radio unit(s) 2210 via, for example, an optical cable or the like. The control system 2202 can be connected to one or more processing nodes 2300 coupled to or included as part of a network(s) 2302 via the network interface 2308 . Each processing node 2300 can include one or more processors 2310 (e.g., CPUs, ASICs, FPGAs, and/or the like), memory 2306 , and a network interface 2308 .

In this example, functions 2310 of the radio access node 2200 described herein are implemented at the one or more processing nodes 2300 or distributed across the control system 2202 and the one or more processing nodes 2300 in any desired manner. In some particular embodiments, some or all of the functions 2310 of the radio access node 2200 described herein are implemented as virtual components executed by one or more virtual machines implemented in a virtual environment(s) hosted by the processing node(s) 2300 . As will be appreciated by one of ordinary skill in the art, additional signaling or communication between the processing node(s) 2300 and the control system 2202 is used in order to carry out at least some of the desired functions 2310 . Notably, in some embodiments, the control system 2202 may not be included, in which case the radio unit(s) 2210 communicate directly with the processing node(s) 2300 via an appropriate network interface(s).

In some embodiments, a computer program including instructions which, when executed by at least one processor, causes the at least one processor to carry out the functionality of radio access node 2200 or a node (e.g., a processing node 2300 ) implementing one or more of the functions 2310 of the radio access node 2200 in a virtual environment according to any of the embodiments described herein is provided. In some embodiments, a carrier comprising the aforementioned computer program product is provided. The carrier is one of an electronic signal, an optical signal, a radio signal, or a computer readable storage medium (e.g., a non-transitory computer readable medium such as memory).

FIG. 122 is a schematic block diagram of the radio access node 2200 according to some other embodiments of the present disclosure. The radio access node 2200 includes one or more modules 2400 , each of which is implemented in software. The module(s) 2400 provide the functionality of the radio access node 2200 described herein. This discussion is equally applicable to the processing node 2300 of FIG. 123 , where the modules 2400 may be implemented and/or distributed across one or more processing nodes 2300 and/or control system 2202 .

FIG. 124 is a schematic block diagram of a UE 2500 according to some embodiments of the present disclosure. As illustrated, the UE 2500 includes one or more processors 2502 (e.g., CPUs, ASICs, FPGAs, and/or the like), memory 2504 , and one or more transceivers 2506 each including one or more transmitters 2508 and one or more receivers 2510 coupled to one or more antennas 2512 . In some embodiments, the functionality of the UE 2500 described above may be fully or partially implemented in software that is, e.g., stored in the memory 2504 and executed by the processor(s) 2502 .

In some embodiments, a computer program including instructions which, when executed by at least one processor, causes the at least one processor to carry out the functionality of the UE 2500 according to any of the embodiments described herein is provided. In some embodiments, a carrier comprising the aforementioned computer program product can be provided. The carrier can be one of an electronic signal, an optical signal, a radio signal, or a computer readable storage medium (e.g., a non-transitory computer readable medium such as a physical memory).

FIG. 125 is a schematic block diagram of the UE 2500 according to some other embodiments of the present disclosure. In these embodiments, UE 2500 can include one or more modules 2600 , each of which is implemented in software. Module(s) 2600 can provide at least a portion of the functionality of UE 2500 described hereinabove.

Transport of Data Flows Over Cellular Networks

FIG. 126 illustrates the architecture of a 5G network and introduces relevant core network functions like the User Plane Function (UPF).

In NR PDCP, header compression is used. The protocol is based on the Robust Header Compression (ROHC) framework defined in IETF RFC 5795: “The Robust Header Compression (RoHC) Framework.” The basic idea is to utilize the redundancy in protocol headers of new packets, i.e., use the fact that they are similar or the same as previously received packets. Therefore, subsequent packets do not need to include the full protocol header information since it is already known from previously received packets. A compression/decompression context is maintained to keep track of that information. Several different RoHC profiles with different header compression algorithms/variants exist and are defined/referred to in the NR PDCP specification.

The UE undergoes a handover procedure when it changes its primary cell. The source and target cell may be belonging to different gNBs. Focusing on the user plane protocol stack involved in this procedure: the UE resets MAC with HARQ processes, as well as re-establishes (flushes) the RLC entities. The PDCP protocol serves as the handover anchor, meaning that PDCP will in acknowledged mode do retransmissions of not yet acknowledged data, that might have been lost due to MAC/HARQ and RLC flushing at handover.

In dual connectivity, beside handover, a radio bearer might be changed from MCG type to/from SCG type or to/from Split type. This can be realized with the handover procedure including PDCP re-establishment, or alternatively with the PDCP data recovery procedure.

Support for Ethernet PDU sessions over 5G networks was introduced in 3GPP TS 23.501 and TS 23.502 (see, for example, versions 15.2.0 of both those specifications).

FIG. 127 shows the protocol stack for Ethernet PDU type data (user plane) as defined in release 15 of 3GPP TS 29.561, “Interworking between 5G Network and External Data Networks; Stage 3”. External data networks may include, for example, Ethernet LANs. Key characteristics for such interworking with external Data Networks (DNs) include:

UPF shall store MAC addresses received from the DN or the UE; the 5G network does not assign MAC addresses to UEs Ethernet preamble, Start Frame Delimiter (SFD) and Frame Check Sequence (FCS) are not sent over 5GS The SMF provides Ethernet filter set and forwarding rules to the UPF based on the Ethernet Frame Structure and UE MAC addresses During PDU session establishment a DN-AAA (Data Network—Authentication, Authorization and Accounting) server can provide a list of MAC addresses allowed for this particular PDU session (see release 15 of 3GPP TS 29.561). IP layer is considered as an application layer which is not part of the Ethernet PDU Session (see release 15 of 3GPP TS 29.561)

Time Sensitive Networking (TSN) is a set of features that allow deterministic networking in Ethernet based wired communication networks. Within a TSN network the communication endpoints are called Talker and Listener. All the switches (e.g., bridges) in between Talker and Listener need to support certain TSN features, like e.g. IEEE 802.1AS time synchronization. All nodes that are synchronized in the network belong to a so-called TSN domain. TSN communication is only possible within such a TSN domain. To allow for deterministic communication, TSN communication happens in streams, that are setup across the TSN domain before the data communication takes place. In the TSN network, there are different possibilities as to how frames are identified and mapped to a TSN stream, as defined in IEEE 802.1CB. The identification might be based on MAC addresses and VLAN-headers and/or IP headers. But as the TSN standard is under development now, other aspects (e.g. the Ether-Type field) might also be introduced therein to identify frames. After a TSN stream has been established in the TSN network, frames are identified in the whole TSN network based on the specific stream identifiers.

There is currently no header compression defined for Ethernet frames for a 5G network. This would lead to transmission of uncompressed Ethernet frames, which entails a significant overhead given the typically small payload sizes for certain types of traffic, such as industrial IoT/URLLC traffic.

During handover re-establishment and data recovery, RoHC performance cannot be guaranteed, which is problematic for services relying on guaranteed transmission success. Counteracting this issue by provisioning more resources for the service (e.g. not using RoHC) is likely to lead to unacceptable resource wastage.

A protocol for Ethernet header compression aligned with RoHC may sometimes be able to lead to good compression ratios but not deterministically, e.g. in the above handover situation. This leads to the disadvantage of radio access nodes (e.g., gNB) also being unable to reserve minimum-needed resources deterministically, i.e. such nodes may need to reserve more resources for the case that header compression does not lead to full compression, coming with additional resource wastage.

A RoHC compression context loss (e.g., due to a handover) will lead to delays in packet forwarding at the receiver which may be unacceptable for URLLC traffic.

Certain aspects of the present disclosure and their embodiments may provide solutions to these or other challenges.

The present disclosure is described within the context of 3GPP NR radio technology (e.g., 3GPP TS 38.300 V1.3.0). However, it will be understood by those skilled in the art that embodiments of the disclosure also apply to other cellular communication networks. Embodiments of the disclosure enable the efficient transport of data flows (e.g., time-sensitive data flows, such as those for time-sensitive networking (TSN)) over a cellular (e.g., 5G) network by compressing redundant information. This is achieved by making one or more core network nodes TSN-aware, supporting the handling of the TSN flows while reducing unnecessary overhead.

Methods are outlined in this disclosure for header compression of Ethernet/TSN stream-based transmissions in a 5G network. Compared to known methods like RoHC for IP header compression, the methods outlined herein rely on specific properties of the Ethernet/TSN stream to enable a deterministic compression ratio.

There are, proposed herein, various embodiments which address one or more of the issues disclosed herein.

Certain embodiments may provide one or more of the following technical advantage(s). Ethernet header compression in cellular networks generally lowers resource usage, increasing capacity. Embodiments of the disclosure may lead to a deterministic compression ratio, i.e. enabling deterministic minimum-needed resource reservations for the flow/UE instead of needing to account for situations where this optimum compression ratio cannot be met. In this way, the capacity of the system is improved.

As described below, embodiments of the disclosure assume that values for one or more fields in a data packet header (e.g., an Ethernet header) are static for an established data stream such as a TSN stream. In this context, a value may be considered “static” if it remains the same for multiple data packets in sequence within the data stream. Thus, this does not preclude embodiments in which the values for the fields in the header are updated as necessary (i.e. semi-static). The values for the fields may or may not remain the same for the lifetime of the data stream.

TSN streams are established and a configuration is applied across all nodes involved in the TSN stream before any data packet is transmitted. This includes also, that TSN stream identifiers are announced.

FIG. 128 shows a frame structure for a TSN data packet. Within a TSN stream, header fields are used to identify a stream. These fields comprise of e.g. DST MAC address (6 byte), VLAN-header (4 bytes) and IP-Header fields (various fields). These fields are not usually altered after a TSN stream has been setup. Therefore, these fields offer the possibility of a static compression throughout the 5G network, e.g. UPF to UE, gNB to UE, etc.

According to one embodiment of the disclosure, one or more fields within a header for the data packet are configured for the UE and/or the gNB or UPF before data transmission takes place. For example, the one or more fields may comprise the Ethernet header and maybe also other header fields as for example parts of an IP-header in case they are used for TSN stream identification.

The values for the fields in the header for packets received or transmitted in a QoS flow may be configured per QoS flow. Additionally or alternatively, the values for the fields in the header for packets received or transmitted in a PDU session may be configured per PDU session.

The procedure for downlink is illustrated in FIG. 129 .

For TSN streams in the Downlink the 5G CN (e.g., a core network node, such as the AMF or UPF, or a combination of both) may use information from a TSN network regarding TSN stream identification and which fields can be treated as static or not, or it might use a pre-configuration for this.

An identifier might be added to data packets inside of PDU sessions or QoS Flows to differentiate multiple TSN/Ethernet streams within the same session or flow (thus the identifier is for a particular TSN/Ethernet stream). For example, the identifier may be used instead of the Ethernet header fields removed statically for transmission; an 8-bit header might be sufficient to separate TSN streams inside sessions or flows.

For header compression between UPF and UE (initiated by 5G CN), NAS signaling is used. This comprises to signal the header content that is statically mapped to the UE and optionally also a stream identifier that is used within a PDU session or within a QoS Flow to differentiate between different TSN streams. The 5G CN configures the UPF for the static mapping.

For Downlink transmissions for header compression between gNB and UE, RRC signaling can be used, i.e. when a new QoS flow is established for the UE, the UE is instructed to utilize the configured header for packets received on this QoS flow. In an alternative embodiment, PDCP control signaling is employed to indicate updates to the otherwise static header context (i.e. providing the UE with a new header context), allowing a semi-static header configuration for the UE.

Furthermore, in all cases above, when an update of the static header is indicated, or the new static header is indicated, a sequence number may be indicated alongside, identifying the packet from which onwards the new header should be used for decompression.

In a further embodiment, in the receiving entity (e.g., UE in DL), reordering of received packets according to a sequence number should be applied prior to header decompression. This way, when indicating new configured headers alongside with a sequence number, the first packet for which a new configured header is valid is identified.

The procedure for uplink is illustrated in FIG. 130 .

For TSN streams in the Uplink the UE might get information from a TSN network regarding TSN stream identification and which fields can be treated as static or not and inform the 5G CN accordingly (e.g., by forwarding the request from the TSN network to the 5G CN).

An identifier might be added to data packets inside of PDU sessions or QoS Flows to differentiate multiple TSN/Ethernet streams within the same session or flow (thus the identifier is for a particular TSN/Ethernet stream). For example, the identifier may be used instead of the Ethernet header fields removed statically for transmission; an 8-bit header might be sufficient to separate TSN streams inside sessions or flows.

For header compression between UE and UPF (initiated by UE), again NAS signaling is used. The UE might request a static header compression from the SGCN by signaling the request over NAS alongside any TSN configuration data it has received from a TSN network regarding the TSN stream packet headers. The SGCN may then configure the static mapping in the UPF and possibly also assign a stream identifier that is used within a PDU session or within a QoS Flow to differentiate between multiple TSN streams. The SGCN may use NAS signaling to inform the UE about the static mapping, as well as a potential identifier to use. The SGCN configures the UPF for the static mapping.

Furthermore, in all cases above, when an update of the static header is indicated, or the new static header is indicated, a sequence number may be indicated alongside, identifying the packet from which onwards the new header should be used for decompression.

For Uplink transmissions, the UE is configured to remove the Ethernet header fields before transmission. The configuration may be indicated via RRC signaling or NAS signaling. The header removal function may be implemented in an SDAP or PDCP transmission algorithm. A sequence number may be indicated identifying the first packet from which onwards the removal of Ethernet header fields applies.

For Uplink transmissions, the UE indicates the (removed) header to the 5G network prior to any data transmission, so that the 5G network can consider the header when receiving packets from the UE. Also, in this case the header can be configured per QoS flow or per PDU session. Furthermore, a sequence number may be indicated identifying the first packet for which the header had been removed and the configured header should be applied to.

In a further embodiment, in the receiving entity (gNB or UPF in UL), reordering of received packets according to a sequence number should be applied prior to header decompression. This way, when indicating new configured headers alongside with a sequence number, the first packet for which a new configured header is valid for is identified.

To handle TSN streams over radio, radio resources may be pre-allocated using e.g. semi-persistent scheduling (SPS) or instant-uplink access (IUA). Resource pre-allocation benefits from a known payload size for transmission. In the RoHC framework the worst-case payload size is still the whole packet including all headers; as it cannot be determined when it is necessary to transmit the full context, it would be necessary to reserve resources for the worst-case. This is not the case for the static header compression method outlined above.

TSN is based on timely delivery of packets. Packets that have to be retransmitted or buffered because of a context unawareness lead to packet latencies that are most likely unacceptable. It would be better to either discard the packet or reuse an old (or as introduced in this disclosure, statically configured) context instead.

FIG. 131 depicts a method in accordance with particular embodiments. The method may be performed by one or more core network nodes. For example, the method may be performed by an AMF and/or a UPF (such as the AMF and UPF described above with respect to FIG. 126 . Further, the method may relate or correspond to the actions of the element “5G CN” in FIG. 129 described above. The method enables transport of data packets associated with a data stream (such as a TSN or other time-critical data stream) in an external data network (such as an Ethernet network or LAN).

The method begins at step VV 102 , in which the core network node(s) obtains configuration information for a data stream in an external data network. The configuration information indicates respective values for one or more fields within a header of data packets associated with the data stream which are to remain static. The core network node(s) may receive such configuration information from the external data network directly (e.g., in a request message to establish the data stream), or be pre-configured with the information. The one or more fields for which values may be static may comprise one or more Ethernet header fields, such as one or more (or all) of: destination address field; source address field; virtual LAN tag field; and type/length field. The one or more fields may additionally or alternatively comprise one or more fields in the IP header.

In step VV 104 , the core network node(s) initiates transmission of the configuration information to a wireless device which is to receive the data stream. For example, the configuration information may be transmitted via NAS signalling.

The core network node(s) may establish an identifier for the data stream to enable it to be distinguished from other data streams. In embodiments where data packets are transmitted to the wireless device as part of a PDU session or QoS flow, the identifier may be unique within the PDU session or QoS flow (and therefore in such embodiments an identifier value may be re-used for different data flows outside the PDU session or QoS flow). The configuration information may additionally include the identifier for the associated data stream.

In step VV 106 , the core network node(s) receives a data packet associated with the data stream from the external data network. The data packet may be identified as being associated with, or belonging to, the data stream via any suitable mechanism. The identification might be based on MAC addresses and VLAN-headers and/or IP headers. Alternatively or additionally, other aspects (e.g. the Ether-Type field) might also be introduced therein to identify data packets.

In step VV 108 , the core network node(s) removes the one or more fields from the data packet to generate a compressed data packet. That is, the core network node(s) removes the one or more fields which were identified in the configuration information obtained in step VV 102 . Optionally, the core network node(s) may add the identifier for the data stream to the compressed data packet. It will be understood that the identifier may be added to the data packet before or after the one or more fields have been removed.

In step VV 110 , the core network node(s) initiates transmission of the compressed data packet to the wireless device. For example, the core network node(s) may send the compressed data packet to a radio access node (such as a gNB or other base station) for onward transmission to the wireless device.

In further embodiments of the disclosure, the configuration information for the data stream may become updated after the configuration above has been established. In such embodiments, updated configuration information may be obtained for the data stream (e.g., from the external data network), comprising an indication of respective updated values for one or more fields within the header of data packets associated with the data stream which are to remain static. The one or more fields which have static values may be the same as or different to the one or more fields identified originally. The updated configuration information can then be transmitted to the wireless device (e.g., via NAS signalling) to enable the wireless device to decompress data packets which have had header information removed according to the updated configuration. The updated configuration information may comprise a sequence number, indicating the data packet in the sequence of data packets associated with the data stream from which the updated configuration is to apply.

FIG. 132 depicts a method in accordance with particular embodiments. The method may be performed by one or more core network nodes. For example, the method may be performed by an AMF and/or a UPF (such as the AMF and UPF described above with respect to FIG. 126 . Further, the method may relate or correspond to the actions of the element “5G CN” in FIG. 130 described above. The method enables transport of data packets associated with a data stream (such as a TSN or other time-critical data stream) in an external data network (such as an Ethernet network or LAN).

The method begins at step VV 202 , in which the core network node(s) obtains configuration information for a data stream in an external data network. The configuration information indicates respective values for one or more fields within a header of data packets associated with the data stream which are to remain static. The core network node(s) may receive such configuration information from the external data network directly (e.g., in a request message to establish the data stream), from a wireless device which is to transmit data packets associated with or belonging to the data stream (e.g., in a request message from the external data network forwarded by the wireless device over signalling such as NAS signalling) or be pre-configured with the information. The one or more fields for which values may be static may comprise one or more Ethernet header fields, such as one or more (or all) of: destination address field; source address field; virtual LAN tag field; and type/length field. The one or more fields may additionally or alternatively comprise one or more fields in the IP header.

An identifier for the data stream may be established to enable it to be distinguished from other data streams. In embodiments where data packets are transmitted by the wireless device as part of a PDU session or QoS flow, the identifier may be unique within the PDU session or QoS flow (and therefore in such embodiments an identifier value may be re-used for different data flows outside the PDU session or QoS flow). The configuration information may additionally include the identifier for the associated data stream. Alternatively, where the core network node(s) establish the identifier for the data stream, the identifier may be transmitted by the core network node(s) to the wireless device.

Optionally, the method may further comprise a step (not illustrated) of sending the configuration information to the wireless device which is to transmit data packets associated with or belonging to the data stream. This step may particularly apply when the configuration information in step VV 202 is not received from the wireless device, or when the wireless device is unable to process and obtain the configuration information itself (e.g., from a request message received from the external data network). The configuration information may be sent via NAS signalling, for example.

In step VV 204 , the core network node(s) receives a data packet associated with the data stream from the wireless device. The data packet is compressed by the removal of one or more fields in the header (e.g., by the wireless device following the method set out below in FIG. 133 ), according to the configuration information obtained in step VV 202 .

In step VV 206 , the core network node(s) adds the one or more fields from the data packet to generate a decompressed data packet. That is, the core network node(s) adds the one or more fields which were identified in the configuration information obtained in step VV 202 .

In step VV 208 , the core network node(s) initiates transmission of the decompressed data packet over the external data network.

In further embodiments of the disclosure, the configuration information for the data stream may become updated after the configuration above has been established. In such embodiments, updated configuration information may be obtained for the data stream (e.g., from the external data network or the wireless device), comprising an indication of respective updated values for one or more fields within the header of data packets associated with the data stream which are to remain static. The one or more fields which have static values may be the same as or different to the one or more fields identified originally. The updated configuration information transmitted to the wireless device (e.g., via NAS signalling), particularly if the updated configuration information is received from the external data network directly. Additionally or alternatively, the updated configuration information is utilized to decompress received data packets in future which have been compressed by the wireless device according to the updated configuration. The updated configuration information may comprise a sequence number, indicating the data packet in the sequence of data packets associated with the data stream from which the updated configuration is to apply. Thus the core network node(s) may add header fields according to the updated configuration for all data packets which follow the sequence number indicated in the updated configuration information. Optionally, the core network node(s) may re-order received data packets according to their respective sequence numbers to facilitate this processing.

FIG. 133 depicts a method in accordance with particular embodiments. The method may be performed by a wireless device (such as the UE described above with respect to FIG. 126 ). Further, the method may relate or correspond to the actions of the element “UE” in FIG. 129 described above. The method enables transport of data packets associated with a data stream (such as a TSN or other time-critical data stream) in an external data network (such as an Ethernet network or LAN).

The method begins at step XX 102 , in which the wireless device obtains configuration information for a data stream in an external data network. The configuration information indicates respective values for one or more fields within a header of data packets associated with the data stream which are to remain static. The wireless device may receive such configuration information from the external data network directly (e.g., in a request message to establish the data stream), or from one or more core network nodes (e.g., via a transmission from a radio access network node, such as a gNB or other base station, via NAS signalling). The one or more fields for which values may be static may comprise one or more Ethernet header fields, such as one or more (or all) of: destination address field; source address field; virtual LAN tag field; and type/length field. The one or more fields may additionally or alternatively comprise one or more fields in the IP header.

An identifier for the data stream may be established to enable it to be distinguished from other data streams. In embodiments where data packets are received by the wireless device as part of a PDU session or QoS flow, the identifier may be unique within the PDU session or QoS flow (and therefore in such embodiments an identifier value may be re-used for different data flows outside the PDU session or QoS flow). The configuration information may additionally include the identifier for the associated data stream.

In step XX 104 , the wireless device receives a data packet associated with the data stream from the radio access network node. The data packet is compressed by the removal of one or more fields in the header (e.g., by the core network node(s) or the radio access network node itself following the method set out above), according to the configuration information obtained in step XX 102 .

In step XX 106 , the wireless device adds the one or more fields from the data packet to generate a decompressed data packet. That is, the wireless device adds the one or more fields which were identified in the configuration information obtained in step XX 102 . Optionally, the decompressed data packet may be transmitted onwards over the external data network.

In further embodiments of the disclosure, the configuration information for the data stream may become updated after the configuration above has been established. In such embodiments, updated configuration information may be obtained for the data stream (e.g., from the core network node), comprising an indication of respective updated values for one or more fields within the header of data packets associated with the data stream which are to remain static. The one or more fields which have static values may be the same as or different to the one or more fields identified originally. The updated configuration information is then utilized to decompress received data packets in future which have been compressed by the core network node(s) or radio access network node according to the updated configuration. The updated configuration information may comprise a sequence number, indicating the data packet in the sequence of data packets associated with the data stream from which the updated configuration is to apply. Thus the wireless device may add header fields according to the updated configuration for all data packets which follow the sequence number indicated in the updated configuration information. Optionally, the wireless device may re-order received data packets according to their respective sequence numbers to facilitate this processing.

FIG. 134 depicts a method in accordance with particular embodiments. The method may be performed by a wireless device (such as the UE described above with respect to FIG. 126 ). Further, the method may relate or correspond to the actions of the element “UE” in FIG. 130 described above. The method enables transport of data packets associated with a data stream (such as a TSN or other time-critical data stream) in an external data network (such as an Ethernet network or LAN).

The method begins at step XX 202 , in which the wireless device obtains configuration information for a data stream in an external data network. The configuration information indicates respective values for one or more fields within a header of data packets associated with the data stream which are to remain static. The wireless device may receive such configuration information from the external data network directly (e.g., in a request message to establish the data stream), or from one or more core network nodes (e.g., via NAS or RRC signalling). The one or more fields for which values may be static may comprise one or more Ethernet header fields, such as one or more (or all) of: destination address field; source address field; virtual LAN tag field; and type/length field. The one or more fields may additionally or alternatively comprise one or more fields in the IP header.

An identifier for the data stream may be established (e.g., by the one or more core network nodes) to enable it to be distinguished from other data streams. In embodiments where data packets are received by the wireless device as part of a PDU session or QoS flow, the identifier may be unique within the PDU session or QoS flow (and therefore in such embodiments an identifier value may be re-used for different data flows outside the PDU session or QoS flow). The configuration information may additionally include the identifier for the associated data stream.

In step XX 204 , the wireless device obtains a data packet associated with or belonging to the data stream. For example, the data packet may be received from the external data network, or generated by the wireless device (e.g., in response to some user interaction or by execution of an application on the wireless device).

In step XX 206 , the wireless device removes the one or more fields from the data packet to generate a compressed data packet. That is, the wireless device removes the one or more fields which were identified in the configuration information obtained in step XX 202 . Optionally, the wireless device may add the identifier for the data stream to the compressed data packet. It will be understood that the identifier may be added to the data packet before or after the one or more fields have been removed. The header removal function may be implemented in an SDAP or PDCP transmission algorithm.

In step XX 208 , the wireless device initiates transmission of the compressed data packet over the external data network. For example, the wireless device may send the compressed data packet in a transmission to a radio access network node (such as a gNB or other base station) for onward transmission to one or more core network nodes and thereafter the external data network. The one or more core network nodes are enabled to decompress the compressed data packets prior to their transmission over the external data network, e.g., by following the methods set out above).

In further embodiments, the configuration information for the data stream may become updated after the configuration above has been established. In such embodiments, updated configuration information may be obtained for the data stream (e.g., from the external data network), comprising an indication of respective updated values for one or more fields within the header of data packets associated with the data stream which are to remain static. The one or more fields which have static values may be the same as or different to the one or more fields identified originally. The updated configuration information can then be transmitted by the wireless device (e.g., via NAS signalling) to one or more core network nodes to enable those core network nodes to decompress data packets which have had header information removed according to the updated configuration. The updated configuration information may comprise a sequence number, indicating the data packet in the sequence of data packets associated with the data stream from which the updated configuration is to apply.

It will be appreciated that the methods shown in FIGS. 131 - 134 may be implemented in one or more of the nodes shown in FIGS. 120 - 125 , as appropriate.

Combination of Resource-Scheduling and Header-Compression Techniques

As indicated above, the various techniques described herein may be combined with each other, to provide advantages with respect to latency, reliability, etc. For example, one particular combination that is advantageous is the combination of the techniques described above for scheduling resources and the techniques described for compressing headers of TSN frames.

Thus, for example, the method illustrated in FIG. 117 can be combined with the method shown in FIG. 131 , resulting in a method performed in one or more nodes of a core network associated with a radio access network (RAN) for handling a time-sensitive data stream associated with a user equipment (UE) and an external network. This method comprises, as shown at block 1210 of FIG. 117 , the step of receiving, from the external network, a transmission schedule associated with a time-sensitive data stream, and further comprises, as shown at block 1220 of that same figure, the step of sending, to the RAN, a request to allocate radio resources for communication of the data stream between the RAN and a first UE, wherein the request further comprises information related to the transmission schedule. As shown at block 1230 of FIG. 117 , the method further comprises receiving, from the RAN, a response indicating whether radio resources can be allocated to meet the transmission schedule associated with the data stream.

The method further comprises the step of obtaining configuration information for the data stream, the configuration information indicating respective values for one or more fields within a header of data packets associated with the data stream which are to remain static; this step is shown at block VV 102 of FIG. 131 . The method still further comprises the steps of initiating transmission of the configuration information to the first UE, receiving a data packet associated with the data stream from the external data network, removing the one or more fields from the data packet to generate a compressed data packet, and initiating transmission of the compressed data packet to the first UE, as shown at blocks VV 104 , VV 106 , VV 108 , and VV 110 of FIG. 131 .

It will be appreciated that any of the variations discussed above for these techniques may apply here, for the combined technique. Thus, for example, the external network comprises a Time-Sensitive Network (TSN), in some embodiments, and the data stream comprises a TSN stream. Here, the transmission schedule may comprise cycle times and gate control lists for one or more traffic classes comprising the TSN stream.

In some embodiments, the information related to the transmission schedule includes one or more of the following: an identifier of the first UE; identifiers of one or more quality-of-service, QoS, flows associated with the data stream; and a QoS requirement associated with each of the QoS flows. In some of these embodiments, each QoS requirement comprises one or more time windows during which the data stream is required to be transmitted and/or an initial time window and a periodicity that identifies subsequent time windows. In some of these latter embodiments, if the response indicates that radio resources cannot be allocated to meet the transmission schedule of the data stream, the response further comprises an indication of one or more further time windows during which radio resources can be allocated. In some embodiments, the response indicates whether the QoS requirement associated with each of the QoS flows can be met, and the method further comprises determining whether the transmission schedule can be met based on the indication of whether the QoS requirement associated with each of the QoS flows can be met.

In some embodiments, the method further comprises sending, to the external network, an indication of whether the transmission schedule can be met. In some of these and in other embodiments, the method is performed by an access management function (AMF) in a 5G core network (5GC). The transmission schedule may be received from the external network and the radio resources may be for downlink communication from the RAN to the first UE, in some embodiments, or the transmission schedule may be received from the first UE and the radio resources may be for uplink communication from the first UE to the RAN, in other embodiments or instances.

In some embodiments, the step of obtaining configuration information comprises receiving the configuration information from the external data network. In others, the configuration information is pre-configured in the one or more nodes of the core network.

In some embodiments, the compressed data packet comprises an identifier for the data stream. The identifier may be added by the one or more nodes of the core network node.

In some embodiments, the compressed data packet is transmitted to the first UE as part of a protocol data unit (PDU) session or a quality of service (QoS) flow. In some of these embodiments, the identifier mentioned above may be unique within the PDU session or QoS flow.

In some embodiments, the configuration information is transmitted to the first UE using non-access stratum (NAS) signaling. In some, the configuration information comprises an identifier for the data stream.

In some embodiments, the method further comprises obtaining updated configuration information for the data stream, the updated configuration information comprising an indication of respective updated values for one or more fields within the header of data packets associated with the data stream which are to remain static, and initiating transmission of the updated configuration information to the first UE. This updated configuration information further may comprise an indication of a sequence number identifying a data packet associated with the data stream from which the respective updated values apply.

In any of the preceding embodiments, the data packet may comprise user data, and the step of initiating transmission of the compressed data packet to the first UE may comprise forwarding the user data to the first UE via a transmission to a base station.

The decompression techniques described above may also be combined with these techniques. Thus, some methods carried out by one or more nodes of the core network may comprise receiving a data packet associated with the data stream from a second UE; adding the one or more fields to the data packet to generate a decompressed data packet; and initiating transmission of the decompressed data packet over the external data network, as shown at blocks VV 204 , VV 206 , and VV 208 of FIG. 132 .

In some embodiments, the method may further comprise initiating transmission, to the second UE, of an indication of the respective values for one or more fields within the header of data packets associated with the data stream which are to remain static. The data packet may comprise user data, and the step of initiating transmission of the decompressed data packet over the external data network may comprise forwarding the user data to a host computer over the external data network.

TSN Over a RAN

At least some units of factory automation, such as autonomous, multifunctional, and/or mobile machinery and robots, require networking by means of wireless radio communication. However, a factory unit acting as a mobile terminal of the RAN, e.g., a 3GPP user equipment (UE), would have to establish a radio connection with a radio base station of the RAN just to find out that this particular radio base station does not support TSN.

Accordingly, there is a need for a technique that enables TSN over wireless radio communication. An alternative or more specific object is to enable a mobile terminal to specifically select a radio base station that supports TSN, preferably prior to establishing a radio connection between the mobile terminal and the radio base station.

FIG. 135 shows a flowchart for a method 400 of handling TSN over a RAN. The method 400 comprises a step 402 of receiving SI from a RBS of the RAN. The SI is implicative or indicative as to support for TSN through the RBS. The SI may be RBS-specific. The method 400 further comprises a step 404 of establishing or initiating to establish, depending on the received SI, at least one TSN stream of the TSN through the RBS. The method 400 may be performed by a UE radio-connected or radio-connectable to the RAN.

FIG. 136 shows a flowchart for a method 500 of announcing TSN over a RAN.

The method 500 comprises a step 502 of transmitting SI from a RBS of the RAN. The SI is implicative or indicative as to support for TSN through the RBS. The SI may be RBS-specific. The method 500 further comprises a step 504 of supporting, according to the transmitted SI, at least one TSN stream of the TSN through the RBS. The method 500 may be performed by the RBS of the RAN, for example.

FIG. 137 shows a flowchart for a method 600 of distributing a configuration message for TSN over a RAN. The method 600 comprises a step 602 of determining at least one configuration message indicative or implicative as to support for the TSN through at least one RBS of the RAN. The method 600 further comprises a step 604 of sending the at least one configuration message from a CN to each of the at least one RBS of the RAN.

The method 600 may be performed by the CN and/or the using a network component of the CN, the AMF or MME, and/or using a TSN function. The TSN function may be a Centralized Network Configuration (CNC) or a Centralized User Configuration (CUC).

The step 404 of establishing or initiating to establish, depending on the received SI, the at least one TSN stream may comprise selectively (e.g., conditionally) establishing or selectively (e.g., conditionally) initiating to establish the at least one TSN stream. The selectivity (e.g., conditionality) may be dependent on the received SI. The UE may decide, based on the SI from the RBS, whether to attempt establishing the TSN stream, e.g., prior to accessing or connecting with the base station, or not.

The step 404 of establishing or initiating to establish the at least one TSN stream may comprise selectively performing or selectively initiating to perform at least one of a random access procedure with the RBS of the RAN; a radio resource control (RRC) connection setup with the RBS of the RAN; and a network attach procedure with a CN connected to the RAN. The selectivity may be dependent on the received SI.

The establishing step 404 may comprise performing or initiating to perform a TSN application that uses the at least one established TSN stream. The TSN application or a client of the TSN application may be performed at the UE. The selectivity (e.g., the conditionality) in the step 404 may be fulfilled if the received SI is indicative of TSN features required by the TSN application.

The step 402 of receiving the SI is performed with respect to each of a plurality of RBSs of the RAN. The step 404 of establishing or initiating to establish the at least one TSN stream may comprise selecting, among the plurality of RBSs, the RBS the SI of which is indicative of TSN features required by the TSN application.

The RBS which best fulfills the required TSN features according to the respective SI may be selected (e.g., if none of the plurality of RBSs fulfills the required TSN features). Alternatively or in addition, the RBS which SI is indicative of the most preferably TSN features may be selected (e.g., if more than one of the plurality of RBSs fulfills the required TSN features).

The method 400 may further comprise a step of sending a control message to the CN. The control message may be indicative of TSN features required by the TSN application. The control message may be a non-access stratum (NAS) message.

The control message may be indicative of a request for the TSN. The control message may be forwarded to the CUC.

The SI may be implicative or indicative of at least one TSN feature supported by or through the RBS. The SI may be RBS-specific. The selectivity (e.g., the conditionality) in the step 404 may be dependent on the at least one supported TSN feature. Alternatively or in addition, the TSN stream may be established over the RAN depending on the at least one supported TSN feature. For example, the establishing of the at least one TSN stream may comprise performing or initiating to perform the random access with the RBS depending on the at least one supported TSN feature.

Herein, the TSN feature may encompass any feature or functionality available at the RBS for the TSN. The at least one TSN feature supported through the RBS may also be referred to as TSN capability of the RBS.

The at least one TSN feature may comprise at least one of a time-synchronization, a latency bound for the at least one TSN stream, and a reliability measure for the at least one TSN stream. The time-synchronization may be a time-synchronization of RBSs and/or network components processing (e.g., transporting) the at least one TSN stream.

Alternatively or in addition, the SI may be indicative of a TSN configuration (also, TSN configuration scheme) for the TSN through the RBS. For example, the establishing 404 of the at least one TSN stream may comprise performing or initiating a TSN setup according to the TSN configuration. The TSN configuration may be indicative of an availability or unavailability of at least one of a CNC and a CUC.

The SI may be broadcasted from the RBS in the step 502 . The SI may be a broadcast message. The SI may be comprised in one or more system information blocks (SIBs).

The method 500 may further comprise a step of receiving a configuration message indicative of the support for TSN from the CN at the RBS. The SI transmitted by the RBS may be derived from the received configuration message.

The SI may be implicative or indicative of at least one TSN feature supported by or through the RBS. The SI may be broadcasted in one or more SIBs. The method 500 may further comprise any feature and/or step disclosed in the context of the UE and the method 400 , or any feature or step corresponding thereto.

The configuration message may be sent from the AMF of the CN. The configuration message may be implicative or indicative of at least one TSN feature supported or supposed to be supported by or through the RBS.

The method 600 may further comprise any feature or step of the methods 400 and 500 , or any feature or step corresponding thereto.

Embodiments of the technique maintain compatibility with the 3GPP document TS 23.501, version 15.1.0, specifying “System Architecture for the 5G System” (Stage 2), or a successor thereof.

A network (e.g., a 5G network comprising the RAN providing NR access as defined by 3GPP) is configured to support TSN transmissions through at least some RBSs. For a UE to become attached to such a TSN network over the RAN (e.g., 5G radio or NR), there is no existing way to get information as to whether the network in general, and the RBS (e.g., a gNB) specifically, supports TSN transmissions or not. In embodiments of the technique, the SI enables the UE to determine if and/or how certain TSN features are supported, before getting into radio resource control (RRC) connected mode and further signaling with the 5G network. Thus, the technique enables the UE and, therefore, also a TSN application the UE is connected to, to be aware of whether, which and/or how TSN features are supported by the network, specifically the RAN and/or the RBS transmitting the SI.

The SI may be implicative or indicative as to the support of TSN features. The TSN features may comprise at least one of time synchronization, redundancy, reliability, and latency (e.g., an estimated end-to-end latency).

Embodiments of the technique enable the UE to receive necessary TSN-related information in the SI before getting attached to the 5G network. In this way, the UE is aware of which TSN features are supported by the 5G network. Furthermore, the 5G network may inform one or more UEs in the same way about configuration details of the TSN network and/or how to, for example, perform time synchronization and network management.

For example, not all RBSs (e.g., gNBs) covering an area (e.g., deployed in a factory hall) support TSN traffic. The technique may be implemented to block those UEs (also: TSN-UEs) that require TSN traffic from certain RBSs (e.g., gNBs), e.g., from those RBSs that do not support TSN or not the TSN features required by the UE.

The SI may be implemented by one or more System Information Blocks (SIBs).

An overall functionality and structure of a Master Information Block (MIB) and SIBs for NR may be essentially the same as for LTE. A difference between NR and LTE is that in NR provides two different types of SIBs. A first type of SIBs is transmitted periodically, e.g., equal or similar to SIB transmissions in LTE. A second type of SIBs is transmitted only when there is the request from the UE.

The SIBs are broadcasted by the RBS (e.g., a gNB) and include the main part of the system information the UE requires to access a cell served by the RBS and other information on cell reselection. SIBs are transmitted over a Downlink Shared Channel (DL-SCH). The presence of the system information on the DL-SCH in a subframe is indicated by the transmission of a corresponding Physical Downlink Control Channel (PDCCH) marked with a special system-information Radio Network Temporary Identifier (SI-RNTI).

A number of the different SIBs are defined by 3GPP for LTE and NR, e.g., characterized by the type of the information included in the SIBs. This system information informs the UE about the network capabilities. Not all SIBs are supposed to be present. SIBs are broadcasted repeatedly by the RBS (e.g., the gNB).

Within a TSN network, i.e., a network supporting TSN, the communication endpoints are called TSN talker and TSN listener. At least one of the TSN talker and the TSN listener is an UE. For the support of TSN, all RBSs and network components (e.g., switches, bridges, or routers) between the TSN talker and the TSN listener support certain TSN features, e.g. IEEE 802.1AS time synchronization. All nodes (e.g., RBSs and/or network components) that are synchronized in the network belong to a so-called TSN domain. TSN communication is only possible within such a TSN domain.

TSN for a RAN or a RAN configured for TSN may comprise features for deterministic networking, which are also referred to as TSN features. The TSN features may comprise at least one of time synchronization, guaranteed (e.g., low) latency transmissions (e.g., an upper bound on latency), and guaranteed (e.g., high) reliability (e.g., an upper bound on packet error rate). The time synchronization may comprise a time synchronization between components of the RAN (e.g., the RBSs) and/or network components (e.g., in a backhaul domain and/or the CN).

Optionally, the SI is indicative of the TSN features supported through the respective RBS.

The supported TSN features may comprise or be compatible with at least one of the following group of categories. A first category comprises time synchronization, e.g. according to the standard IEEE 802.1AS. A second category comprises bounded low latency, e.g. according to at least one of the standards IEEE 802.1Qav, IEEE 802.1Qbu, IEEE 802.1Qbv, IEEE 802.1Qch, and IEEE 802.1Qcr. A third category comprises ultra-reliability, e.g. according to at least one of the standards IEEE 802.1CB, IEEE 802.1Qca, and IEEE 802.1Qci. A fourth category comprises network configuration and management, e.g. according to at least one of the standards IEEE 802.1Qat, IEEE 802.1Qcc, IEEE 802.1Qcp and IEEE 802.1CS.

The configuration and/or management of a TSN network including the RAN can be implemented in different manners, e.g., in a centralized or in a distributed setup as defined by the standard IEEE 802.1Qcc. Examples of different configuration models are described with reference to FIGS. 138 , 139 , and 140 .

FIG. 138 schematically illustrates a block diagram for a first example of a communications system 700 comprising embodiments of devices 100 , 200 and 300 , which may be configured to carry out the methods illustrated in FIGS. 135 , 136 , and 137 , respectively. The communication system 700 comprises the RAN 710 and the CN 730 . The RAN 710 may comprise at least one embodiment of the device 200 . The CN 730 may comprise at least one embodiment of the device 300 , e.g., a network component 300 - 1 . The network component 300 - 1 may be a switch, a bridge or a router. A backhaul domain 720 provides data links between the RBSs 200 of the RAN 710 and/or between the at least one RBS 200 and the CN 730 . The data links may comprise at least one of microwave links, Ethernet links and fiber optical links.

The SI 712 is broadcasted by the RBS 200 to the UE 100 according to the steps 402 and 502 . The RBS 200 is configured to broadcast the SI 712 according to the step 502 and to support the TSN stream according to the step 504 responsive to the configuration message 722 - 1 received from or through the network component 300 - 1 .

In a scheme for distributed TSN configuration, which is illustrated by the first example in FIG. 138 , there is no CUC and no CNC for the TSN network. The TSN talker 100 is, therefore, responsible for initiation of a TSN stream in the step 404 . As no CNC is present, the network components 300 - 1 (e.g., switches or bridges) are configuring themselves, which may not allow using, for example, time-gated queuing as defined in IEEE 802.1Qbv. The distributed TSN configuration may be compatible or consistent with the document IEEE P802.1Qcc/D2.3, “Draft Standard for Local and metropolitan area networks—Bridges and Bridged Networks Amendment: Stream Reservation Protocol (SRP) Enhancements and Performance Improvements”, IEEE TSN Task Group, e.g., draft status 03-05-2018.

In a first scheme for centralized TSN configuration, which is schematically depicted in FIG. 139 for a second example of the communication system 700 , the TSN talker 100 is responsible for initialization of the TSN stream in the step 404 , while the network components 300 - 1 are configured by a CNC 300 - 2 . The centralized TSN configuration may be compatible or consistent with the document IEEE P802.1Qcc/D2.3.

The SI 712 is broadcasted by the RBS 200 to the UE 100 according to the steps 402 and 502 . Alternatively or additionally to the configuration message 722 - 1 , the RBS 200 is configured to broadcast the SI 712 according to the step 502 and to support the TSN stream according to the step 504 responsive to the configuration message 722 - 2 received from or through the CNC 300 - 2 .

In a second scheme for centralized TSN configuration (also: fully centralized TSN configuration), which is schematically depicted in FIG. 140 for a third example of the communication system 700 , the network components 300 - 1 are configured by the CNC 300 - 2 and the CUC 300 - 3 with network configuration information and user configuration information, respectively. In one implementation, the CUC 300 - 3 may configure the network components to establish the TSN stream as soon as the TSN talker 100 is radio-connected to the RBS 200 . In another implementation that is combinable with the one implementation, the TSN talker 100 is responsible for initialization of the at least one TSN stream, while quality requirements of the TSN talker 100 for the at least one TSN stream and/or the number of TSN streams for the TSN talker 100 is configured by the CUC 300 - 3 . The fully centralized TSN configuration may be compatible or consistent with the document IEEE P802.1Qcc/D2.3.

The SI 712 is broadcasted by the RBS 200 to the UE 100 according to the steps 402 and 502 . Alternatively or additionally to the configuration message 722 - 1 and/or the configuration message 722 - 2 , the RBS 200 is configured to broadcast the SI 712 according to the step 502 and to support the TSN stream according to the step 504 responsive to the configuration message 722 - 3 received from the CUC 300 - 3 .

Optionally, e.g. in any of the three examples for the communication system 700 , the SI 712 is transmitted on a broadcast channel of the RAN 710 . The SI 712 may (e.g., positively) indicate the support of the TSN, e.g., without user and/or network configuration information. The UE 100 may receive the user and/or network configuration information on a downlink control channel from the RBS 200 , by TSN-specific protocols and/or from the CN 710 (e.g., the device 300 - 1 ) using a non-access stratum (NAS) protocol. Alternatively or in combination, the SI 712 may comprise (at least partly) the user and/or network configuration information.

The TSN communication between TSN talker (as an embodiment of the device 100 ) and TSN listener (which may or may not be a further embodiment of the device 100 ) happens in TSN streams. A TSN stream is based on certain requirements in terms of data rate and latency given by an application (TSN application) implemented at the TSN talker and the TSN listener. The TSN configuration and management features are used to setup the TSN stream and to guarantee the requirements of the TSN stream across the network.

In the distributed scheme (e.g., according to the first example in FIG. 138 ), the TSN talker 100 and the TSN listener 100 may use the Stream Reservation Protocol (SRP) to setup and configure the at least one TSN stream in every RBS 200 and/or every network component 300 - 1 (e.g., every switch) along the path from the TSN talker 100 to the TSN listener 100 in the TSN network. Optionally, some TSN features require the CNC 300 - 2 as a central management entity (e.g., according to the second example in FIG. 139 ). The CNC 300 - 2 uses for example a Network Configuration Protocol (Netconf) and/or “Yet Another Next Generation” (YANG) models to configure the RBS 200 and/or the network components 300 - 1 (e.g., switches) in the network for each TSN stream. This also allows the use of time-gated queuing as defined in IEEE 802.1Qbv that enables data transport in a TSN network with deterministic latency. With time-gated queuing on each RBS 200 and/or each network component 300 - 1 (e.g., switch), queues are opened or closed following a precise schedule that allows high priority packets to pass through the RBS 200 or network component 300 - 1 with minimum latency and jitter if it arrives at ingress port within the time the gate is scheduled to be open. In the fully centralized scheme (e.g., according to the third example in FIG. 140 ) the communication system 700 comprises a CUC 300 - 3 as a point of contact for the TSN listener 100 and/or the TSN talker 100 . The CUC 300 - 3 collects stream requirements and/or endpoint capabilities from the TSN listener 100 and/or the TSN talker 100 . The CUC 300 - 3 may communicate with the CNC 300 - 2 directly. The TSN configuration may be implemented as explained in the standard IEEE 802.1Qcc in detail.

FIG. 141 shows a functional block diagram for a fourth example of a communication system 700 comprising embodiments of the devices 100 , 200 and 300 . The fourth example may further comprise any of the feature described for the first, second and/or third example, wherein like reference signs refer to interchangeable or equivalent features. An optional interworking between the 5G network (e.g., comprising the RAN 710 and the CN 730 ) and the TSN network architecture (e.g., the CNC 300 - 2 and the CUC 300 - 3 ) may be based on at least one of the control messages 722 - 2 and 722 - 3 from the CNC 300 - 2 and the CUC 300 - 3 , respectively, e.g., as illustrated in FIG. 141 . At least one of the control messages 722 - 2 and 722 - 3 may be forwarded to the AMF 300 - 4 (in the CN 730 ) and/or to the RBS 200 (in the RAN 710 ) using a control plane of the 5G network. Alternatively or in addition, the CN 730 , e.g., the AMF 300 - 4 , may implement at least one of the CNC 300 - 2 and the CUC 300 - 3 .

The technique enables connecting TSN listener 100 and TSN talker 100 wirelessly to a TSN network, e.g., using a 5G network as defined by 3GPP. The 5G standard defined by 3GPP addresses factory use cases through a plurality of features, especially on the RAN (e.g., providing 5G NR) to make it more reliable and reduce the transmit latency compared to an evolved UMTS radio access network (E-UTRAN, i.e., the radio access technology of 4G LTE).

The 5G network comprises the UE 100 , the RAN 730 instantiated as the gNB 200 and nodes 300 - 4 within the core network (5G CN). An example for the 5G network architecture is illustrated on the left-hand side in FIG. 141 . An example for the TSN network architecture is illustrated on the right-hand side in FIG. 141

Both technologies, the 5G network and the TSN network, define own methods for network management and/or configuration. Different mechanisms to achieve communication determinism are arranged to enable end-to-end deterministic networking to support TSN streams, e.g., for industrial networks. A study item for upcoming 3GPP release 16 has been initiated in the 3GPP document RP-181479 to support TSN, e.g., for factory automation use cases.

Here, the UE 100 being the radio device connected to the RAN 710 (and thus to the 5G network) may also be referred to as a 5G endpoint. A device connected to the TSN network (also, TSN domain) may be referred to as a TSN endpoint.

Despite what is shown in FIG. 141 , is also possible that the UE 100 is not connected to a single endpoint but instead to a TSN network comprising at least one TSN bridge and at least one endpoint. The UE 100 is then part of a TSN-5G gateway.

The control plane of the 5G network may comprise at least one of a Network Repository Function (NRF), the AMF 300 - 4 , a Session Management Function (SMF), a Network Exposure Function (NEF), a Policy Control Function (PCF), and a Unified Data Management (UDM).

A data plane of the 5G network comprises a User Plane Function (UPF), at least one embodiment of the RBS 200 , and/or at least one embodiment of the UE 100 .

A TSN listener 1002 may be embodied by or performed (e.g., as an application) at the UE 100 . While the UE 100 operates as or is used by the TSN listener 1002 in the fourth example of the communication system 700 shown in FIG. 141 , the UE 100 may alternatively or additionally operate as a TSN talker in any example. Optionally, a TSN talker 1004 is embodied by another UE 100 connected through the same or another RBS 200 to the communication system 700 .

The step 604 of the method 600 may be implemented according to at least one of the following variants (e.g., in the context of any of the four examples of the communication system 700 in FIGS. 138 to 141 ). In a first variant, the CNC 300 - 2 configures the gNB 200 by sending the configuration message 722 - 2 . In a second variant, the CUC 300 - 3 sends the configuration message 722 - 3 to the AMF 300 - 4 and, thereby, configures the gNB 200 . For example, the AMF 300 - 4 forwards the configuration message 722 - 3 to the gNB 200 or derives the configuration message 722 - 4 from the configuration message 722 - 3 . In a third variant (not shown), the CUC 300 - 3 sends the configuration message 722 - 3 to the gNB 200 . In a fourth variant (not shown), the CNC 300 - 2 sends the configuration message 722 - 2 to the AMF 300 - 4 . Optionally, e.g., in any of the variants, the AMF 300 - 4 implements at least one of the CNC 300 - 2 and the CUC 300 - 3 .

Alternatively or in addition, the CNC 300 - 2 sends the configuration message 722 - 2 to the network component 300 - 1 (e.g., a switch or a router) and, thereby, configures the gNB 200 . For example, the network component 300 - 1 forwards the configuration message 722 - 2 to the gNB 200 or derives the configuration message 722 - 1 from the configuration message 722 - 2 .

While the technique is described herein with embodiments in the context manufacturing and factory automation for clarity and concreteness, the technique may further be applicable to automotive communication and home automation.

FIG. 142 shows a signaling diagram 1100 for TSN Stream Configuration involving exemplary embodiments of the device 100 (e.g., a UE 100 as the TSN talker and/or a UE 100 as the TSN listener) and exemplary embodiments of the device 300 (namely 300 - 1 , 300 - 2 and 300 - 3 ). While these multiple embodiments of the devices 100 and 300 are shown and described in combination, any subcombination may be realized. For example, only one of the network component 300 - 1 , the CNC 300 - 2 and the CUC 300 - 3 may embody the device 300 . Alternatively or in addition, only one of the TSN talker and the TSN listener may be an embodiment of the device 100 .

The steps for the TSN Stream Configuration (e.g., according to the signaling diagram 1100 ) may be performed after the UE 100 has decided to access (e.g., radio-connect and/or attach to) the RBS 200 (not shown in FIG. 141 for simplicity) based on the SI received in the step 402 . The step 404 may initiate at least one of the steps for the TSN Stream Configuration.

Each UE 100 implementing a TSN talker or a TSN listener is radio-connected through an embodiment of the RBS 200 to at least one of the network component 300 - 1 , the CNC 300 - 2 and the CUC 300 - 3 . The UEs 100 may be radio-connected through the same RBS 200 or different RBSs 200 . The TSN Stream Configuration may be compatible or consistent with IEEE 802.1Qcc.

The TSN Stream Configuration (i.e., setting up the at least one TSN stream in the TSN network) according to the fully centralized configuration scheme comprises at least one of the following steps.

In a first step 1102 , the CUC 300 - 3 may take input from e.g. an industrial application or engineering tool (e.g. a programmable logic controller, PLC), which specifies for example the devices, which are supposed to exchange time-sensitive streams (i.e., TSN streams). The PLC may be adapted to control manufacturing processes, such as assembly lines, or robotic devices, or any activity that requires high reliability control and/or ease of programming and process fault diagnosis.

In a second step 1104 , the CUC 300 - 2 reads the capabilities of end stations and applications in the TSN network, which includes period and/or interval of user traffic and payload sizes.

In a third step 1106 , based on this above information, the CUC 300 - 3 creates at least one of a Stream ID as an identifier for each TSN stream, a Stream Rank, and UsertoNetwork Requirements.

In a fourth step 1108 , the CNC 300 - 2 discovers the physical network topology using, for example, a Link Layer Discovery Protocol (LLDP) and any network management protocol.

In a fifth step 1110 , the CNC 300 - 2 uses a network management protocol to read TSN capabilities of bridges (e.g., IEEE 802.1Q, 802.1AS, 802.1CB) in the TSN network.

In a sixth step 1112 , the CUC 300 - 3 initiates join requests to configure the at least one TSN stream in order to configure network resources at the bridges 300 - 1 for a TSN stream from one TSN talker 100 to one TSN listener 100 .

In a seventh step, a group of the TSN talker 100 and the TSN listener 100 (i.e., a group of elements specifying a TSN stream) is created by the CUC 300 - 3 , e.g., as specified in the standard IEEE 802.1Qcc, clause 46.2.2.

In an eighth step 1114 , the CNC 300 - 2 configures the TSN domain, checks physical topology and checks if the time sensitive streams are supported by bridges in the network, and performs path and schedule computation of streams.

In a ninth step 1116 , the CNC 300 - 2 configures TSN features in bridges along the path in the TSN network.

In a tenth step 1118 , the CNC 300 - 2 returns status (e.g., success or failure) for resulting resource assignment for the at least one TSN stream to the CUC 300 - 3 .

In an eleventh step 1120 , the CUC 300 - 3 further configures end stations (wherein a protocol used for this information exchange may be out of the scope of the IEEE 802.1Qcc specification) to start the user plane traffic exchange, as defined initially between the TSN listener 100 and the TSN talker 100 .

In the TSN network, the streamID is used to uniquely identify stream configurations. It is used to assign TSN resources to the TSN stream of a TSN talker. The streamID comprises the two tuples MacAddress and UniqueID. The MacAddress is associated with the TSN talker 100 . The UniqueID distinguishes between multiple streams within end stations identified by the same MacAddress.

Any embodiment and implementation of the technique may encode the SI 712 in dedicated information elements in one or more SIBs. According to the step 402 and 502 , a UE 100 is enabled to detect TSN features that are supported by the RBS 200 of the network and/or how they are supported. The UE 100 receives the SI 712 before it attaches to the network, and can check first by listening to an SIB message comprising the SI 712 . The received SI 712 may be forward to the TSN application 1002 or 1004 the UE 100 is serving, and/or the UE 100 uses the SI 712 to setup a connection to the 5G network.

Any embodiment of the RBS 200 may implement the technique by including one or more SIBs and/or information elements in SIBs for indicating to the UE 100 the TSN features and/or TSN configuration details supported by the 5G network, e.g., specifically be the RBS 200 .

Any embodiment of the UE 100 may implement the step 402 by reading the one or more SIBs and/or the information element included therein. Optionally, the included information as to supported TSN features and/or TSN configuration are forwarded to the TSN applications it is serving. Conditionally, i.e., depending on the features indicated as supported in the SI 712 , the information is used to establish a connection to the RBS (e.g., to the 5G network).

An (e.g., expandable) example of a SIB block structure for the SI 712 in the steps 402 and 502 is outlined below using Abstract Syntax Notation One (ASN.1). The same information may also be included in the configuration message 722 of the method 600 .

Furthermore, the SIB blocks may be adapted to future versions of TSN features by, for example, introducing reserved fields to be defined in the future.

For end-to-end time synchronization (e.g., provisioning of an absolute time reference) multiple ways of implementation are possible. The SI 712 may comprise information about how the time synchronization is treated by the RAN (e.g., the 5G network).

The “FRER” parameter refers to the redundancy features that are supported by the 5G network. In case the network does not support redundancy, there is no need to establish, e.g., redundant protocol data unit (PDU) sessions.

The TSN configuration may include the presence of the CUC 300 - 3 and/or the CNC 300 - 2 in the TSN network and/or specific TSN configuration schemes that are supported.

The “Max. Latency added by 5G network” parameter may be used to signal a QoS level in terms of latency and/or reliability that can be supported by the 5G network to the UE 100 . A field representing this parameter may comprise a latency value (e.g., in milliseconds) that can be guaranteed with a sufficient reliability or a classification value (e.g., non-real-time, real-time, hard-real-time or similar). The value may be indicated by a predefined index value. This information may be used by the UE 100 (or the endpoint 1002 or 1004 of the TSN network behind the UE 100 ) to find out before connection establishment if a connection to the RBS 200 (or the 5G network) will be able to support the requirements of the TSN application 1002 or 1004 , or not.

The RBS 200 (e.g., a gNB) may further include a current cell load and/or other metrics into the calculation of that field. Optionally, the SI 712 is indicative of a traffic shaper support, which refers to a quality of service (QoS) that may be guaranteed by the RBS (e.g., the 5G network). For example, the SI 712 may be indicative of whether the shaper is based on credit (e.g., data volume per time and UE) or a time aware shaper (TAS) for TSN.

FIG. 143 shows a signaling diagram 1200 resulting from implementations of the methods 400 , 500 and 600 being performed by embodiments of the devices 100 , 200 and 300 , respectively. More specifically, the technique enables an embodiment of the UE 100 to become aware of TSN features supported by the network over the SI 712 included in one or more SIBs. While the signaling diagram 1200 (and the corresponding flowchart) for TSN stream configuration uses the fully centralized configuration scheme (e.g., as shown in FIG. 140 ), the technique is readily applicable to other configuration schemes (e.g., as shown in FIG. 138 or 139 ).

The implementations of the methods 400 , 500 and 600 enable the UE 100 to get aware of TSN features supported by the network and/or specifically by the RBS 200 over the one or more SIBs including the SI 712 .

In the step 604 , a 5G core function (e.g., the AMF 300 - 4 ) indicates by sending the configuration message 722 to specific RBSs 200 (e.g., gNBs) which TSN features (e.g., according to above non-exhaustive list) are supported or supposed to be enabled (e.g., only a subset of all gNBs might support TSN) and how these TSN features are supported.

Responsive to the reception of the configuration message 722 (e.g., any of the above implementations 722 - 1 to 722 - 4 ), the RBS 200 (e.g., a gNB) generates the SI 712 (e.g., the SIB block information as outlined above) and starts broadcasting the SI 712 , e.g. over the DL-SCH, in the step 502 .

The UE 100 receives and/or reads the SI 712 in the SIB in the step 402 . Optionally, the UE 100 transfers at least some of the information in the SI 712 to the TSN application 1002 or 1004 , e.g., a list of the TSN features supported by the RBS 200 . The TSN application 1002 or 1004 may request a TSN connection towards the UE 100 , if the supported list of TSN features is sufficient, as an example for the conditionality or selectivity in the step 404 .

For initiating the TSN stream in the step 404 , the UE 100 goes into RRC connected mode if not already in that mode and requests a PDU session, which may be of Ethernet type. UE may further provide information by means of NAS signaling on which TSN features are required.

A TSN controller (e.g., the CNC 300 - 2 ) receives a confirmation from the CN 730 and performs path computation and time scheduling. TSN stream communication starts, wherein the RBS 200 supports the TSN stream according to the step 504 .

In any embodiment, the UE 100 may defer or refrain from requesting the RRC connection setup in the step 404 , if the TSN application requires certain TSN features and the UE 100 did not receive in the SIB broadcast 402 that one or more of these features are supported, as an example for the conditionality or selectivity in the step 404 .

In same or another embodiment, the UE 100 reads the SI 712 (i.e., the TSN information included in the one or more SIBs) of multiple RBSs 200 (e.g., gNBs) and selects the RBS 200 , which best fulfills the TSN requirements of the UE 100 . If all RBS 200 fulfill the requirements, the UE 100 may act according to a selection rule, e.g. selecting the RBS 200 indicating the lowest latency.

In any embodiment, the UE 100 may store the SI 712 received in the step 402 . The technique may be implemented as described up until and including the step 402 . When the TSN application 1002 or 1004 requests a TSN communication (i.e., one or more TSN streams), the UE 100 uses the stored SI 712 to either setup the at least one TSN stream in the supported way or declines the TSN request if it is not supported.

The UE 100 may further use the SI 712 from the SIB, e.g., to initialize packet filtering of packets coming in for TSN transmission. Furthermore, the received SI 712 may be used to establish a default PDU session with the 5G network

Combination of TSN Support Detection and Header-Compression Techniques

Once more, as indicated above, the various techniques described herein may be combined with each other, to provide advantages with respect to latency, reliability, etc. For example, one particular combination that is advantageous is the combination of the techniques described above for detecting support for TSN and the techniques described for compressing headers of TSN frames.

Thus, for example, the method illustrated in FIG. 135 can be combined with the method shown in FIG. 133 , resulting in a method performed by a wireless device associated with a wireless communications network, for transport of data packets associated with a data stream in an external data network. This method includes the step of receiving system information (SI) from a radio base station (RBS) of a radio access network (RAN), the SI being indicative of support for time-sensitive networking (TSN) through the RBS, as shown at block 402 of FIG. 135 , as well as the step of establishing at least one TSN stream with the external data network, through the RBS, as shown at block 404 of FIG. 135 . The method further includes the steps of obtaining configuration information for the TSN stream, the configuration information indicating respective values for one or more fields within a header of data packets associated with the TSN stream which are to remain static, as shown at block XX 102 of FIG. 133 , and receiving, from the RBS, a data packet associated with the TSN stream, as shown at block XX 104 of FIG. 133 . The method still further includes the step of adding the one or more fields to the data packet to generate a decompressed data packet, as shown at block XX 106 of FIG. 133 .

In some embodiments, the SI is comprised in one or more system information blocks (SIBs). In some embodiments, the step of obtaining configuration information comprises receiving the configuration information from a network node of the wireless communications network. The data packet may comprise an identifier for the TSN stream; in some embodiments, this identifier is added by the core network node.

In some embodiments, the compressed data packet is received as part of a protocol data unit (PDU) session or a quality of service (QoS) flow. In some of these embodiments, the identifier for the TSN stream is unique within the PDU session or QoS flow.

In some embodiments, the configuration information is transmitted to the wireless device using non-access stratum (NAS) signaling. The configu