10.2.1 Ultra‐Reliable and Low‐Latency Communications

For Release 15, the ultra‐reliable and low‐latency communications (URLLC) service has the capability to achieve an end‐to‐end latency of around 1 ms with a reliability of 99.999% [1]. This is already good enough to support most of the closed loop control applications but that of the most extreme use cases such as, for example, motion control (see Section 10.4.1). Notwithstanding, the capability is certainly good enough to begin the transition to smart factories. Release 16 will further enhance the support for the smart factory and the first release specifically designed with factory support.

For mobility, [2] evaluated the URLLC service via outdoor trial and confirmed that the 3GPP requirements can be satisfied with 25 km/h mobility. Further enhancement will be needed in future releases to cover that use case.

10.2.2 Enhanced Mobile Broadband

The enhanced mobile broadband (eMBB) service defined in Release 15 can achieve downlink average capacity of 16.65 bit/s/Hz/transmit receive point with massive MIMO (multiple‐input multiple‐output) of 32 transmit antennas and 4 received antennas [3]. For 100 MHz channel with only one transmit and receive point (no multi‐user MIMO), the average data rate is 1.665 Gbits/s. For the uplink, with massive MIMO of 4 transmit antennas and 32 received antennas, the average capacity is 6.136 bits/s/Hz/transmit receive point. For example, if the user equipment (UE) only uses 20 MHz of the uplink channel then the average data rate is 120 Mbits/s. This type of data rate is sufficient for almost all but the most demanding use case, such as high‐end cameras (see Section 10.4.13).

Mobility support for many manufacturing services is not much of a concern except for motion control and mobile robots which demand high mobility support. The 5G eMBB service was designed from the beginning to support mobility. It has robust mobility mechanisms capable of supporting seamless mobility across the network.

5G also already has the capability to support traffic classification and prioritization natively. The eMBB services were designed from the beginning to have a flexible quality of service (QoS) framework that supports a variety of traffic flows with a range of QoS requirements.

Unlike the traditional wireless network which has been built as a monolithic network, 5G, from the beginning, has been designed to provide flexible support for different services through the network slicing feature. Loosely speaking, a network slice corresponds to a dedicated set of virtual or physical resource on the same physical network infrastructure and dedicated behavior. A more technical description of the network slice can be found in [3] but this loose definition is sufficient for our discussion here. A particular application can be served by one or more network slices; this is dependent on implementation. For our discussion here we can limit it to a single slice. The slice provides the isolation needed for the applications to scale without impacting other applications on the network thus guaranteeing the Service Level Agreement (SLA) of a particular service no matter how other services (slices) are scaled. This also means that regardless of what is happening in the other slices, certain level of security of a particular slice is maintained. From a particular application point of view, it is as if it has dedicated network resources, even though the underlying implementation is not dedicated. This concept of isolation is key to understanding the paradigm shift features that 5G brings to the table.

What it means for manufacturing use cases is that the different use cases may be implemented on their own dedicated slice. Thus, it naturally resolves the conflicting requirement of different manufacturing use cases. For example, the robotic use case in a plant can scale completely independent of the closed loop control use case. It allows, for example, different lines of communication to scale without impacting the others. For a business operating multiple plants, it will even let the communications of the different plants to scale dynamically and independently. All of this is available natively in the Release 15 eMBB services.

10.2.3 Massive Machine Type Communication

The 5G system defined in Release 15 can easily exceed the massive machine type communication (mMTC) requirements from the International Telecommunication Union (ITU). For narrowband Internet of Things (NB‐IoT), it will support around 600 000 connections per square kilometer per 180 kHz of bandwidth [3]. That means that for 1 MHz of bandwidth it can support over 3.3 million connections. This number is sufficient for all manufacturing scenarios. Since the number of connections is not really seen as a limiting factor, Release 16 development is focused on enhancing the downlink and uplink efficiency, as well as providing extreme coverage [4].

The battery life for NB‐IoT devices was calculated in [5] to be of the order of 10 years if transmission of the device is kept to 200 bytes per day with a coupling loss of −164 dB. This feature is especially important for sensors that need to be untethered such as, for example, some of those used in the logistic use case. If long‐term battery life is an issue, energy harvesting can be used.

These release 15 performance capability are sufficient for the initial senor implementations of the factory. But more capability enhancements, such as security, relay capability, data rate and latency, are needed to meet those more advanced sensor implementations in the future smart factory.

10.4.1 Motion Control

Motion control in an industrial manufacturing sense is defined as the control of moving parts in a machine in a well‐defined manner. The nature of the motion may be linear, circular, or even more complex. Wireless connection is envisioned here to replace the wired connection because of the freedom of movement that it allows and the possibility of reduction in down time from the wear and tear movement makes on the wired connectors.

A motion control feedback loop in diagrammatic form is shown in Figure 10.4. As usual, the input to the loop is the desire output. The controller, based upon the input from the sensors and the input to the loop, periodically sends a control signal to the actuator. The actuator then performs a corresponding action on one or more processes, which in this case, by definition, is a movement of a part in the process. At the same time, the sensors monitor the process output and report the results to the controller. All of this is done in a strictly deterministic cyclic manner. The current industrial Ethernet solution supports a period of around 50 μs. It is now clear that smaller periods allow for faster and more accurate movement controls. Also, the reliability of the communication may impact both safety and the quality of the output. This leads us to requirements that will highly emphasize the URLLC service that could be beyond the URLLC capability in Release 15. 3GPP RAN will study this and determine how best to support such strict URLLC requirement; for example, whether to use the U_u or the side link (PC 5¹).

Besides the transmission of time critical data in motion control, some additional non‐real time (NRT) data is exchanged between the nodes in the motion control loop. This data may be transmitted in parallel to the real time data discussed previously. Examples of this type of data are firmware/software updates or maintenance information. The capacity requirement of these NRT data is not large and data rates of around 1 Mits/s should be sufficient and can be easily handled with the current capabilities of the eMBB feature.

It is also within reason that, in a smart factory, not all devices will be connected wirelessly in the motion control system. Given that, it is then necessary for the 5G system to coexist with the industrial Ethernet system. This implies the following:

1. 5G network should be able to forward frames from the Ethernet sources to 5G destinations and vice versa.
2. Precise time synchronization between multiple motion controllers, some of which are connected to industrial Ethernet and some to 5G can be solved if both industrial Ethernet and 5G supports the Precise Time Protocol (IEEE 1588).
3. As the Ethernet device can be separated from other devices on the same physical Ethernet network via Virtual LAN (IEEE 802.1Q), it means that the 5G system must now become aware of Virtual LAN associations.
4. In order to overcome communication bottle necks, industrial Ethernet solutions use reservation protocols. This implies that 5G would need to be aware of the time aware scheduling defined in IEEE 802.1Qbv.

10.4.2 Control to Control Communication

Control‐to‐control (C2C) communication, which is the communication between controllers, is already in use today. There are two major use cases:

1. Large machines (such as newspaper printers) use several controllers to cluster machine functions together. For efficient operation, these controllers need to communicate among themselves. As can be readily seen, this type of communication uses real time communication that needs to be synchronized.
2. Communication between the controllers of individual machines is often necessary for the efficient operation of the factory. An example of this is the communication between machines that coordinate and control the handoff of work products.

It is not too difficult to imagine that in the factory of the future, with the untethering from 5G, such communication will only increase in both connection and the amount of data exchanged. From a requirements point of view, the main focus can be the C2C communication between different motion subsystems because it typically has the most demanding requirements. In general, we can say that the stringent requirements of reliability, latency, and determinism are similar to that for motion control except that the service area can be much bigger; up to 100 times.

10.4.3 Mobile Control Panels with Safety Functions

Control panels are used to provide the human interface to production machinery and interaction with moving devices. Their main function is to provide configuration, monitoring, debugging, control, maintenance, and emergency safety stop of industrial machinery ranging from simple machines, robots, and cranes to entire production lines. From the safety point of view, reliability and speed are the critical requirements. The URLLC requirements are similar to that of motion control but there are additional requirements on eMBB because the service area is usually bigger although interaction with the public cellular network is not necessarily required.

It is also clear that not all aspects of control panel communication deal with safety. Consequently, it also has the requirement for simultaneous transmission of non‐critical data with the critical safety data. For this non‐critical requirement, the 5G eMBB feature is more than sufficient for the task.

10.4.4 Mobile Robots

Robots are programmable machines that perform multiple operations. They and autonomous machinery, such as automated guided vehicles, have gained more and more usage in the manufacturing environment. Therefore, it is expected that they will play an ever‐increasing role in the factory of the future. They have functionality in a localized service area, such as inside a factory, as well as in a wide service area, such as outdoors. For the localized service area, ultra‐low latency is required but connectivity to the public network is not. For the wide service area case, the latency requirement can be relaxed but interconnection with the public network (e.g. service continuity and roaming) is additionally required. As robotic functions become more complicated, there will arise a need for synchronized ultra‐low latency and ultra‐high reliable communication between robots, for example to control and coordinate carrying and handoff of work goods, or to assemble products in highly cooperative and automatically way. However, not all communications are latency sensitive and so there is also the requirement of parallel transmission of non‐critical data.

10.4.5 Massive Wireless Sensor Networks

The factory of the future will be no different than other 5G environments. Sensors, which measure or monitor the state or behavior of a particular environment, will become ubiquitous. The types of sensors and their communication needs will be heterogeneous. The environments in which the sensors operate will be dynamic, ranging from very benign to very hostile. They may function individually or be used in a coordinated distributed system. Independent of how the sensors are deployed, the training and analysis of the sensor network may be accomplished in a centralized fashion, in a distributed fashion, or a mixture of both.

The traffic pattern of sensors is also heterogeneous and depends upon the type of sensor and the environment in which it operates. It may need low bandwidth as well as high bandwidth. The reliability ranges from strict for safety sensors to that of NRT traffic.² The traffic may exhibit cell similarity and/or periodicity. To minimize the load on the communication system, pre‐processing of the data may be employed.

The architecture of the 5G network that supports sensor networks may also be different. In some cases, such as in remote areas it may be best to have them connected to large macro cells while in other areas it may be better to take advantage of the heterogeneous network architecture.

It now becomes obvious that massive sensor networks, such as the Internet of Things (IoT), do not represent a single use case. They rather represent a class of very diverse use cases. Although the current LTE based mMTC can meet many of those requirements, there are limitations to restrict them to meet the future factory and process plants. New 5G capabilities are needed, such as low latency, high data rate, wider coverage and suitability for non‐public network deployment.

10.4.6 Remote Access and Maintenance

Remote access and maintenance will be a key paradigm shift for the factory of the future. Clearly untethering the factory would transform what can be controlled and maintained remotely, as well as what constitutes “remotely.” We can see that the freedom from the telephone wire that mobile users enjoy so much will directly translate into the smart factory environment.

Remote access and maintenance will apply to devices which:

1. Already have a communication connection, which could be cyclic, for transmitting data regularly.
2. Act almost autonomously.
3. Have local interaction with humans.
4. Sleep most of the time and are only woken up to establish the connection for transmission.

For case 1, the ad hoc communication for remote access and maintenance would be in parallel with the regular data communication. For case 2, the device must have its own local processing power, since it functions almost autonomously, and only establishes communication when necessary. In case 3, if the local personnel need further assistance, a remote connection can be established so that they can access remote expert assistance. Case 4 is similar to a sensor that only wakes up once in a while to transmit its information.

What constitutes remote need not mean extreme geographic differences. Consider, for example, where the partner is a mobile device. It may use remote access even when it is very near the device; maybe even geographically collocated. From the device point of view, it would not and have no need to differentiate that it was accessed locally. From the remote partner's point of view, however, it may be of benefit for it to know what devices are local, say, within the same plant. This could potentially be achieved in the application layer if the application also had access to the 5G location services.

Tracking of the inventory of the devices and periodic readouts of configuration data, event logs, revision data, and predictive maintenance information is also another interesting use case. The requirements in this use case can easily be met with the 5G eMBB service.

Depending on the action instructed by the remote partner, the action of the remotely accessed device may have significant impact on the functionality of the factory. With that being said, the mere act of remote access should not impact the functionality of the factory around the device being accessed, i.e. the act of just opening a remote connection should have no impact on the factory functionality.

The ability to severely impact the factory operation by the remote partner brings up the need for cyber security. Most of the current wired communications do not consider cyber security because the physical access to the device and network will restrict the access to authorized personnel. Thus, we will need a suite of security protocols to protect the device against cyber threats, ranging from turning remote access off to different restrictions on who can access the device and what they can do with the device.

The last thing to consider in this use case is the life cycle of industrial machines. Industrial devices are expected to have a life cycle of around 25 years which is longer than the typical life cycle of a generation of wireless technology (10 years). This means that either we need a way to easily upgrade the communication of the device or that we will need to consider that older generations of wireless technology will need to coexist with newer generations. This is not necessary a standards issue but mainly an implementation issue.³ Nevertheless, the standards will need to take this into account when designing the system.

10.4.7 Augmented Reality

AR will be a major game changer, like everywhere else, on how we will interact with our environment. On the factory floor of the future smart factory, workers will be optimally supported in their new tasks and activities, as well as in ensuring smooth operation. The following applications are foreseen:

1. Monitoring processes and production flows.
2. Expert assistance, maybe even step‐by‐step instructions, for a specific task.
3. Ad hoc support from remote experts.
4. Training and education.

As comfort and ergonomics will be of prime importance in AR devices for the workplace, energy efficient communication and processing comes to the forefront. One solution is to make use of the edge computing offered in 5G to offload all complex processing tasks to the edge.

Another key requirement is that the augmented image should track with the movement of the worker. This puts some limitation on the location of the edge processing as well as the latency of the transmission. Consequently, this use case, similar to motion control, has strict requirements for latency. The required service area, however, is bigger than that for motion control. It does not require interaction with the public network for local service but remote expert support may need the support of the remote access and maintenance use case depending upon the level of support.

10.4.8 Process Automation – Closed Loop Control

This use case consists of the case where multiple sensors are installed in a plant, the sensors send their measurements to the controller, usually periodically, and the controller then decides on how to control the actuator to give the desired output of the process. This use case is different to motion control in the sense that the controller is not controlling moving parts. Consequently, it has stringent requirements for latency but not as extreme as that in motion control. The service area is bigger than that for motion control but interaction with the public network is usually not required.

10.4.9 Process Automation – Process Monitoring

This use case consists of multiple sensors installed in the plant whose main purpose is not to control the plant in the short term but rather to give insights into the process, environmental conditions, or inventory of material. The key requirement of this use case is wide area coverage and interaction with the public network may be required. The current 5G eMBB service seems to be sufficient for this use case. In the case of an extreme number of sensors, augmentation with, e.g. NB‐IoT, would solve the connection issues [3].

10.4.10 Process Automation – Plant Asset Management

For a factory to run optimally, each part of the factory must also be running optimally. This use case involves using sensors to monitor the performance of plant assets, such as pumps, valves, heaters, etc. and to maintain them in a timely fashion. Most of the requirements of this use case can be met by the eMBB service.

For the monitoring of processes, environmental conditions inventory of material, and asset management maintenance, the positioning requirement of the sensors used shows some similarity to that for IoT devices; for indoors it should be better than 1 m, 99% of the time.

10.4.11 Inbound Logistics

Logistics deals with the organization and management of things, including nonphysical things such as information between different points, e.g. between points of origin and points of consumption. It includes all aspects of transportation and storage of things as they move between the points. Inbound logistics has to do with the logistics of things coming into a business. The major use case here includes the tracking of goods, and transportation assets. For example, heavy goods vehicles can be connected to the public 5G network to enable real time tracking and telematics. The container or pallets of goods can be connected to the public 5G network for tracking and inventory control purposes. Both of these use cases can be supported by the current eMBB service. The pallet or container may also detect the presence of a local 5G non‐public network as it moves into the intended plant. This implies that it must know which non‐public network it should connect to; for example, in consolidated trucking, as a truck moves into a business, those goods intended for that business should be able to connect to the non‐public network. However, the non‐public network should not be able to have any knowledge of other goods on the truck; ideally, they would not even attempt to attach to the non‐public network. In this case there are several options as the 5G or IoT device attached to the pallet detects the presence of the local 5G non‐public network:

1. Dual connection: The 5G or IoT device has an independent subscription to both the 5G public network and the local 5G non‐public network.
2. Dual connection: The 5G or IoT devices remains connected and registers on the public 5G network and establishes a simultaneous parallel connection to the local 5G non‐public network.
3. Manual PLMN selection: The 5G or IoT device performs a manual PLMN selection procedure, which may be initiated automatically. This requires that the local network has a human readable identifier to enable manual selection.

The moving of goods and materials inside a plant is also of considerable interest. In this use case, these goods and materials are wirelessly connected and tracked as they move through the plant. In this case the major requirements are seamless service continuity between 5G public and 5G non public networks and indoor positioning accuracy that needs to be less than 1 m (maybe in the range of 20–30 cm) because the communication needs are rather benign.

10.4.12 Wide Area Connectivity for Fleet Maintenance

In this use case, the interest is in using 5G for automatic wide area data collection and tuning of an automotive fleet. Besides the usual use case for telemetry, the 5G capability allows for the downloading of electronic control unit (ECU) tuning to the heavy goods vehicle in a semi static manner, for example to optimize the performance based upon the load, telemetries, and environmental conditions. The major requirement here is the wide area coverage capability because the data requirements are benign. The data itself can be quite delay tolerant; maybe even longer than 30 min because the ECU tune will not be changed that frequently. The reliability of individual uplink telemetries is not so critical as long as periodicity of the data is frequent enough; it is not that big a deal if you miss an individual message.

10.4.13 High‐End Camera

Recently there has been discussion among a number of manufacturing companies about a use case where high‐end cameras are connected via 5G on the factory floor for quality control and inspection within the production line. Another use is for mobile robots to send back what the robot is seeing. By high‐end camera we are talking about high resolution with around 100 frames per second. It then becomes clear that the uplink data rate and coverage in a highly complex radio environment will be the key requirements. The density of theses cameras for production is of the order of 10–20 cameras in an area of 100 m by 50 m initially and it may grow from there.

10.5.1 5G Use Case and Requirements for Smart Factory

In order to develop the right 5G standard solution, clear 5G service requirements from smart factories based on the right use cases are needed. This is the work of 3GPP SA1 WG. In March 2017, Siemens successfully created a study item (FS‐CAV) in SA1 WG on communication for automation in vertical industries, which initiated the 3GPP effort to develop suitable 5G for Industrie 4.0. With the creation of 5G ACIA in April 2018, more voices from smart factory OT players are being heard. With strong collaboration between ICT and OT companies in SA1 WG as well as 5G ACIA, FS‐CAV was completed, with its study report (23.208) published in May 2019. This report became an extremely important specification for 3GPP Release 16 and had a significant influence on the other 3GPP working groups. It has led to many new technical solutions, such as URLLC enhancement, TSN integration, new private network, and so on. Later, based on this study report, SA1 developed a standard specification (TS22.104) which was completed in November 2018 to define the first set of smart factory specific normative 3GPP service requirements for release 16 including all the QoS key performance indicator (KPI) requirements, for which the 3GPP downstream working groups will develop technical solutions with the successful completion of release 16 TS22.104. Furthermore, with the strong support of the OT and ICT industries, SA1, on November 2019, completed its release 17 TS22.104 enhancement with new requirements.

In OT systems, service performance is measured based on end‐to‐end system performance which includes the application layer, while 3GPP can only address the performance in the network transport layer. Therefore, a clear boundary for 5G requirements is needed. The communication service interface (CSIF) is introduced as the border between application and 5G communication service. Source and destination CSIFs are used as two performance measurement points for 3GPP to define 5G network performance, as illustrated in Figure 10.5.

There are many different applications associated with the different traffic types which may have different QoS KPI requirements. 3GPP SA1 groups the different QoS requirements, and use cases into four traffic classes for factory and process automation applications. Each traffic class has a different set of performance characteristic parameters as well as influence parameters which are not essential for the performance of an item but do affect its performance, such as service area, number of users, and survival time:

• Periodic deterministic communication: This is periodic with stringent requirements on timeliness and availability of the communication service.
• Aperiodic deterministic communication: This is without a pre‐set sending time but still with stringent requirements on timeliness and availability of the communication service.
• Non‐deterministic communication: This includes periodic/aperiodic non‐real‐time traffic and subsumes traffic types other than periodic/aperiodic deterministic communication.
• Mixed traffic: This is traffic which cannot be assigned to one of the above communication patterns exclusively.

SA1 has defined new 5G requirements in some key areas for smart factory in Release 16:

• Network QoS performance requirements. It defined much more stringent KPI requirements, especially on latency and service reliability and availability, such as for some periodic deterministic traffic, such as motion control applications with one way communication latency as low as 500 μs, with communication service availability of 99.999–99.999 99%, and communication service reliability (mean time between failures) of about 10 years.

  However, some system performance measurements used by traditional industry OT systems are different to that of the mobile network. Therefore, how to translate the OT's QoS KPIs to ICT's network KPIs is sometimes difficult. One example is the communication service availability defined above, which cannot be mapped to existing 3GPP defined network KPIs straightforwardly. From the OT perspective, the communication service availability requirement is the combination of latency, survival time, and reliability requirement for the 5G system, and the system is considered unavailable to the cyber physical application when an expected message is not received (e.g. transfer time is bigger than the maximum transfer latency) by the application after the application's survival time expires. There is ongoing effort in 3GPP SA1 and 5G ACIA to define a relatively clear mapping of KPIs between the two systems, and this is expected to be achieved in the coming new 3GPP release.

• Integrate Industrial Ethernet, such as TSN, into the 5G system. In the OT system, the Industrial Ethernet currently is the dominant industrial communication technology with 46% market share (Figure 10.6). Even with the projected growth of 5G wireless communications, it will still remain one of the main communication technologies (especially TSN) to link components together in the factory in the near future.

  Therefore, the 5G system must support the seamless integration and interplay with the Industrial Ethernet. Since TSN, defined by 802.1Q, is considered as the main technology replacing existing Ethernet and Fieldbus technologies deployed in factories for many motion control applications, 3GPP Release 16 is focused on supporting TSN as defined in IEEE 802.1Q in the 5G system. The TSN 5G integration requirements are not only about how to support TSN traffic as one type of PDU session but also support clock synchronization defined by IEEE 802.1AS across 5G‐based Ethernet links and other Ethernet transport such as wired Ethernet networks and Ethernet passive optical networks (EPONs).

  

• Support new private networks for smart factory. Because of the stringent QoS performance requirements, data privacy, security considerations, and other special characteristics of OT network operation, there is a strong need from smart factory players to deploy 5G factory networks in a private network format which are not influenced by other public users. The private network can be deployed by mobile operators with some shared resources with public networks, such as using network slicing or network sharing technologies, or be deployed as standalone networks managed by either the factory owner or mobile operators with their own security mechanisms and dedicated network resources. 3GPP SA1 defined the non‐public network (NPN) for this kind of vertical private network in order to distinguish it from the public network (PLMN) normally used in 3GPP. Because the NPN will be deployed differently to the public network, some new requirements on network selection, private user authentication and authorization and interworking between the NPN and the public network are defined.

3GPP SA1 has completed the first set of service requirements for smart factories in 3GPP Release 16, and just completed the additional new service requirements for release 17 on November 2019. The new service requirements which were introduced in release 17 for smart factories are:

• Industrial Ethernet integration, which includes time synchronization, different time domains, integration scenarios, and support for TSN.
• NPNs as private slices, and further implications on security for NPNs.
• Network operation and maintenance in 5G NPNs for cyber‐physical control applications in vertical domains.
• Enhanced QoS monitoring, communication service and network diagnostics.
• CSIF between application and 5G systems, e.g. information to network for setting up communication services for cyber‐physical control applications and the corresponding monitoring.
• Network performance requirements for cyber‐physical control applications in vertical domains.
• Positioning enhancements for Industrial IoT, including relative positioning information and vertical directions/dimension.
• Device‐to‐device/ProSe communication for cyber‐physical applications in vertical domains.

10.5.2 5G Standardized Solution Development for Smart Factory

After 3GPP SA1 WG has defined the 5G service requirements for smart factory for Release 16, the other 3GPP working groups have developed or are developing the technical solutions from different system component perspectives to fulfill those service requirements in Release 16, some working groups also start the work for release 17 since July 2019.

10.5.2.2 Other 3GPP Work for Smart Factory

One of most important requirements from the smart factory for 5G deployment is that the 5G radio must provide URLLC service. 5G‐NR is able to deliver 1 ms latency for Release 16, with further enhancement in capacity of URLLC support, further reduced latency and improved coverage expected for Release 17. The other important work in RAN WG for the smart factory is to provide high accuracy and flexibly deployed clock synchronization solutions for the factory devices, because the OT system, especially systems with motion control (Section 10.4.1), heavily rely on highly accurate clocks for synchronization within a working domain as well as crossing multiple working domains which may belong to one or different base stations.

Security is not only a key feature of 5G but is also a high priority requirement from smart factory for 5G deployment. SA3 WG (System Architecture WG 3) is the working group responsible for security and privacy in 3GPP systems. The working group determines the security and privacy requirements, and specifies the security architectures and protocols. They are looking into 5G security features to support vertical applications by also creating a study item on security for 5G's enhanced support of vertical and LAN services for Release 16. Most of their work focuses on security for verticals' private network. One of the challenges and a controversial issue in Security WG which is still open is whether to introduce other alternative security credential and mechanisms which have not been defined by 3GPP but are being used in existing OT systems. Because existing 3GPP security mechanisms, such as Extensible Authentication Protocol and Authentication and Key Agreement (EAP‐AKA) and Subscriber Identity Module (SIM), are the foundations of the existing 3GPP systems and global mobile operator's core operation, any changes can have not only significant technical impact on the current 5G architecture but significant impact on the business model being widely used today by ICT players. Therefore, any changes need to be fully studied in order to reach agreement within the industry.

3GPP SA5 WG is the working group responsible for network operation and management. Since the private network (NPN) can become one of the main deployment models for the verticals, especially for smart factory, and the private network is a much simpler and smaller scale network compared with the massive public network, the network Operation, Administration and Management (OAM) system for 5G needs to be enhanced and optimized for private network deployment. SA5 WG has created a new study item in this area, which will lead to normative 5G OAM system features for smart factory's private network in Release 16.

10.6.1 Spectrum

One of the keys to any smart factory deployment will be the spectrum. In the early days of this concept, there was a thinking that this could be deployed in unlicensed spectrum. As the industry gained more understanding of the use case, it became clear that the unlicensed spectrum cannot guarantee the reliability needed for OT networks (see the discussion in Section 10.4) and that a certain amount of licensed spectrum is needed. The location of this licensed spectrum, however, needs careful consideration. It must be considered and prioritized a globally harmonized 5G band for the mobile commercial service. Currently, the 3400–3600 MHz band, within the C band (3300–4200 and 4400–5000 MHz), is allocated to mobile services on a co‐primary basis in almost all countries throughout the world. The 5G NR specification will support 3300–3800 MHz from the beginning using a test‐driven development access scheme. This band is in line with the allocation plans from many countries and is the ideal spectrum for the harmonized 5G band. Thus, it is recommended that countries allocate at least 100 MHz of contiguous bandwidth from this band to each commercial 5G network [9]. This implies that countries should allocate the private network spectrum for the OT outside of this band or if they want to put some of the OT private spectrum in this band, care should be taken to leave sufficient spectrum for the usage of the commercial public network.

Perhaps the country that is furthest along in allocating spectrum for OT services is Germany. The current status is that the 3400–3800 MHz band will be divided; 300 MHz (3400–3700 MHz) for the public network and 100 MHz (3700–3800 MHz) for the NPN. The OT industry can bid for the non‐public part for the usage of the smart factory. The bidding for the public 5G spectrum, which also includes spectrum in the 2100 MHz band, started in March 2019 and concluded around June 2019. The regulations (including fees) for the non‐public part have been finalized currently. The application procedure has officially started since November 2019.

Although 100 MHz spectrum for OT applications seems like a large amount of spectrum, if the vision of all the use cases for the smart factory becomes a reality, it is likely that more than 100 MHz will be needed. Fortunately, for many of the use cases, the coverage is small, compared with the commercial macro cell, and that means it can take advantage of higher frequency bands. The World Radiocommunication Conference 2015 (WRC 15) identified several frequencies for study within the 24.25–86 GHz range (the higher frequency bands identified at WRC 15; Figure 10.10). Groups 30 and 40 are prioritized and all regions and countries are recommended to study and support these two bands for IMT, as well as aim to harmonize the technical conditions for the support of these bands at WRC 19. The OT industry should then study the usage of these bands for smart factories.

10.6.2 Early Trials

There have been a number of announced projects and demonstrations of the smart factory use case in the last few years. This section will give a brief summary of these activities. The list should not be considered comprehensive but is intended to give the reader a sense of the tremendous activity and enthusiasm within the manufacturing vertical for 5G.

Among the first cooperative efforts to speed up the development of smart factory is the cooperation between Huawei and Toshiba announced in 2017. This project is to collaborate on the integration of NB‐IoT for the development of smart factories. The goal is to accelerate the commercial availably of NB‐IoT in a diverse range of deployment use cases applied to real‐world manufacturing scenarios. Initial testing will be performed in Huawei's NB‐IoT Open Lab located in Shanghai. The second phase will be live field tests to develop a suite of smart factory solutions.

In the 2018 Hanover Messe, Bosch demonstrated an extended smart factory mock‐up that included a Nokia 5G network enabling real‐time data exchange and artificial intelligence fault pre‐emption.

In April 2018, Nokia Telia and Intel announced a smart factory test using a video application to monitor and analyze a process on the factory floor at the Nokia factory in Oulu, Finland. The application incudes a machine learning algorithm to alert the operator of inconsistencies in the process. A second test showed that real time data can be rendered and accessed at Telia's data center that is 600 km away in Helsinki.

Huawei and Beckhoff Automation GmbH & Co. demonstrated the wireless communication between two cooperating programmable logic controllers during the 2018 Hannover Messe. This proof of concept showed that with direct integration of 5G into the industrial controllers, industrial automation can be realized in a more flexible and economical manner. The two companies also verified predictive maintenance with a wireless sensor scenario.

Ericsson reported in 2018 that in a case study that applied cellular IoT technology to its Nanjing Ericsson Panda factory breakeven can be reach in less than two years with 50% return on investment in the first year [10]. The use cases investigated include asset monitoring, environmental sensors, and inbound logistics. The study showed both CAPEX and OPEX savings.

Qualcomm and Bosch announced in February 2019 the completion of a joint research effort to study the radio channel for industrial environment and characterize its features. Using this as a springboard, the two companies intend to launch an over‐the‐air trial using 5G NR industrial IoT. Among other things, the trial will investigate the usage of 5G Release 15 NR for current industrial usage.

In February 2019, the UK turned on its first ever 5G factory trial. The trial is being conducted by the Worcestershire 5G Consortium and used the Worcester Bosch factory as the test bed. The organizations involved in the trial include AWTG, Huawei, O₂, Malvern Hills Science Park, British Telecom, Worcestershire Local Enterprise Partnership, 5GIC at the University of Surrey, QinetiQ, Yamazaki Mazak, and Worcester Bosch. It will investigate methodologies to increase industrial productivity via “preventative and assisted maintenance using robotics, big data analytics and AR over 5G” [11]. Besides investigating the applicability of 5G for the manufacturing industry by taking the usual measurement of speed and latency, the Consortium will endeavor to make this trial network ultra‐secure with the help of 5G.