《5G-ACIA:2023工业5G边缘计算白皮书-用例、架构和部署(英文版)(40页).pdf》由会员分享,可在线阅读,更多相关《5G-ACIA:2023工业5G边缘计算白皮书-用例、架构和部署(英文版)(40页).pdf(40页珍藏版)》请在三个皮匠报告上搜索。
1、Industrial 5G Edge Computing Use Cases,Architecture and Deployment5G Alliance for Connected Industries and Automation5G-ACIA White PaperWhite Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment2 Table of Contents1 Executive Summary 32 Introduction 33 Use Case Requirements and Ed
2、ge Computing Benefits 43.1 Use Cases:Introduction 43.1.1 Selected Edge Use Cases 53.1.2 New and Evolved Use Cases 53.2 Mapping Use Cases to Edge Deployment Location Options 84 Example Deployments for Use Cases 114.1 Factory Automation:Mobile Robots 114.2 Process Automation:Closed-Loop Control 134.3
3、Logistics and Warehouse Automation 144.4 Human-Machine Interfaces 144.5 Monitoring and Maintenance 165 Deployment Considerations for Standard Components of the Edge Computing Solution 175.1 Edge Runtime Infrastructure 175.2 Edge Deployment Locations and Data Plane Connectivity Path Considerations 18
4、5.3 Data Plane Connectivity Path Considerations 205.4 Choosing the Data Plane and Edge Application Server 245.5 Examples of Private and Public Network Integration Scenarios 256 Conclusions 267 Key Terms and Acronyms 278 Annex:General Method for Evaluating the Relevance of Edge Computation to Differe
5、nt Use Cases 298.1 Overview of Method 298.2 Examples of Edge Relevance Analysis Based on the Defined Method 318.2.1 Use Case Description and Assumptions(Step 1)318.2.2 Requirements for the Mobile Robot Use Case and Edge Capabilities for Meeting Them(Steps 2 and 3)318.2.3 Benefits of Edge Computing f
6、or the Mobile Robot Use Case(Step 4)328.2.4 Summary:the Relevance of Edge Computing to Mobile Robot Use Cases 339 References 3410 5G-ACIA Members 36White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 3 1 Executive SummaryThis document addresses manufacturing use cases that
7、 may require or benefit from edge computing,while proposing non-public deployment options for a number of cases and making recommendations on how appropriate network architectures could be built.We examine how the edge computing features of 5G net-works can improve industrial use cases,concluding th
8、at they deliver many other benefits besides latency.Use cases can also profit from locality and therefore greater data privacy as well as the ability to scalably run compute-intensive appli-cations such as media processing while consuming less net-work bandwidth.Solutions are also endowed with enhan
9、ced data aggregation capabilities as well as greater resiliency and reliability.We present examples of use cases that we have examined and show that all of them could be deployed using standard 5G network functions and configuration capabilities.We also discuss the details of various options for con
10、figuring the components of an edge computing solution(such as edge runtime infrastructure,network functions,and application functions)to illustrate how the flexibility provided by the standards makes it possible to meet the requirements of use cases in different setups.About 5G-ACIAThe 5G Alliance f
11、or Connected Industries and Automation(5G-ACIA)was established to serve as the main global forum for addressing,discussing,and evaluating relevant techni-cal,regulatory,and business aspects of 5G for the industrial domain.It embraces the entire ecosystem and all relevant stakeholders,which include b
12、ut arent limited to the opera-tional technology industry(industrial automation companies,engineering companies,production system manufacturers,end users,etc.),the information and communication tech-nology industry(chip manufacturers,network infrastructure vendors,mobile network operators,etc.),unive
13、rsities,gov-ernment agencies,research facilities,and industry associa-tions.5G-ACIAs overarching goal is to promote the best pos-sible use of industrial 5G while maximizing the usefulness of 5G technology and 5G networks in the industrial domain.This includes ensuring that ongoing 5G standardization
14、 and reg-ulatory activities adequately consider relevant interests and requirements and that new developments in 5G are effec-tively communicated to and understood by manufacturers.2 IntroductionEdge computing is a distributed information technology ar-chitecture in which client data is processed ph
15、ysically close to where it originates,namely at or near a networks periph-ery.Together with novel network functions,this approach enables new deployment options to help 5G and future net-works achieve both very low latency and very high reliability.Several 5G-ACIA publications have already discussed
16、 edge computing in industrial networks.The second edition of the white paper“5G for Connected Industries and Automation”1,for example,talks about the generic concept of industri-al edge computing and introduces the idea of an edge data center(a cloud located close to mobile devices where data is gen
17、erated).It also describes generic use of network slicing in industrial contexts.The primary objective of this white paper,which is explained in greater detail in chapter 3,is to analyze factory floor use cases that already rely on or could benefit from edge com-puting in various non-public network c
18、onfigurations(see 2 for details).There we also discuss the concept of“edge”by describing different edge locations and mapping use cases to them.In chapter 4,we then go further and discuss possible architectures,uses for existing standard functionalities,and White Paper Industrial 5G Edge Computing U
19、se Cases,Architecture and Deployment4 possible gaps.Chapter 5 provides recommendations on de-ployment and configuration options for solution components and analyzes the drawbacks and benefits of each one.Scope of This White PaperThis white paper addresses use cases described in various 5G-ACIA white
20、 papers and 3GPPs Technical Report 22.804 3,Technical Specification 22.104 4,and Technical Report 22.832 5(for example,on gap analysis and its applicabili-ty to use cases and related edge computing requirements)as well as newer edge computing use cases that may even be applicable to 5G-ACIA itself(l
21、ike for testbeds,proofs of concept,ecosystem trials,etc.).It describes the requirements for various edge locations in different factory domains,iden-tifies the existing use cases that would benefit the most from edge computing for meeting basic requirements,and discusses the advantages of edge compu
22、ting for the iden-tified use cases.It analyzes implementation,feasibility,and deployment options for the addressed use cases and points out possible gaps in existing specifications and especially those of 3GPP.Finally,it proposes best deployment practices and makes recommendations.3 Use Case Require
23、ments and Edge Computing Benefits3.1 Use Cases:IntroductionIn this section,we present industrial edge computing use cases that could be implemented with cellular wireless tech-nology(especially 5G)and examine the potential benefits of doing so.The primary advantages of edge computing for 5G networks
24、 are improved latency,greater reliability,local handling of data and therefore greater data privacy,reduced bandwidth requirements as a result of local processing,and facilitated data collection from devices.The scalable processing capac-ity of local edge clouds makes them ideal for use cases that r
25、equire very low latency.As an edge sites distance from the communicating devices increases,the area served(e.g.for data aggregation purposes)expands and latency increases accordingly.The reliability of a network service diminishes with increasing distance from the operational technology(OT)premises.
26、It is often required by law for data to be kept at a geographical location that is close enough to keep these problems in check.Other inherent advantages of edge com-puting are discussed in detail in connection with individual use cases.In this paper,we examine use cases in which wireless edge devic
27、es are used inside or near a factory,executing selected operations locally while sending other data to the nearest edge cloud for speedy processing.Exactly which types of op-erations can be processed in the cloud depends on the select-ed deployment option.For identifying the benefits of edge computi
28、ng and its appli-cability to the considered industrial use cases,we focus on the following:Computing and aggregating functions that either receive input from various data sources to save band-width or carry out analyses locally Offloading of heavy computational and coordinating tasks from individual
29、 enddevices to distributed infra-structure Low-latency responses and closed-loop feedback Security and data privacy considerations Some of the presented use cases are based on 3GPPs Tech-nical Report 22.804 3,Technical Specification 22.104 4,and Technical Report 22.832 5 and the 5G-ACIA white paper“
30、Key 5G Use Cases and Requirements”6 while others are White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 5 new industrial use cases.They are evaluated to determine their suitability taking advantage of edge computing.The annex of this white paper contains a lengthier examp
31、le ex-plaining in detail how to identify potential edge computing benefits for use cases in 5G wireless networks.The following examples are therefore kept concise.3.1.1 Selected Edge Use CasesProcess automation with closed PI/PID control loopsThis use case is based on a 3GPP closed-loop control use
32、case for process automation 4.Several sensors installed in a plant carry out continuous measurements.The resulting data is sent to a controller that decides whether or not to set actuators.A plant can have one or more such controllers.Here the use of a proportional integral(PI)or proportional integr
33、al derivative(PI/PID)controller is considered.This use case has very strict requirements in terms of latency and ser-vice availability.For the required service performance values for closed-loop control in process automation,see 4,Ta-bleA.2.3.1-1.Local data processing,aggregation,and storage are all
34、 crucial for the success of this automation use case.Benefits of edge computing:reduced latency,improved reli-ability,efficient use of local computing capabilities and avail-able bandwidth,additional aggregated data analytics in the local cloudRemote access,monitoring,and maintenance based on aggreg
35、ated sensor data and data aggregation functionsThis is an existing 3GPP use case(TS 22.104)4.The logic for triggering remote access can be moved from the device to the edge.Data is collected from multiple sensors installed on devices and aggregated to the edge infrastructure where it is processed fo
36、r instructing a remote control center or diag-nostic system to perform the required maintenance.Periodic access to the resulting huge volume of sensor data and rapid processing of the aggregated data are the prerequisites for taking advantage of edge computing capabilities.Benefits of edge computing
37、:greater efficiency as a result of locally aggregating sensor data,less bandwidth required for device maintenance,remote software upgrades,offloading of heavy local computations,more reliable data due to reduced distance.3.1.2 New and Evolved Use CasesMany new and evolved use cases involve factory a
38、utomation and human-machine interfaces.They are based on existing use cases discussed in the cited sources but have evolved fur-ther as a result of our analyses.Factory automation:mobile robots and automated guided vehicles(AGVs),robot tooling(see 3,section 4.5)AGVs can be deployed to move products,
39、product compo-nents,tools,or raw materials around a factory between stor-age areas and production lines in accordance with logistical requirements.To execute these complex tasks,AGVs have to be mobile robots with the ability to process information flows on inventory and other things,handle materials
40、,monitor and control activities,recognize images and more.This scenario involves both indoor and outdoor communication.There are two options in this use case:a)AGVs are operated by a centralized automatic controller based on input received by video streaming and or other methods such LIDAR sensor da
41、ta,or b)they are controlled by a human operator.Modules for video processing and detection and/or tracking as well as simultaneous localization and mapping(SLAM)functions process incoming video streams and provide information on the robots environment.The main prerequisites for AGV control use cases
42、 are a de-vice-to-network endpoint latency of about 5 ms,high service reliability with an uplink speed of 3 to 8 Mbit/s for 1080 pixel images,and a remote control bit rate of 100 kbps.The following additional requirements apply to the robot tooling use case:A cyclic data communication service with a
43、 minimum cycle time of one millisecond for precise cooperative robotic motion control and between one and 10 milli-seconds for machine control Jitter 50%of cycle timeWhite Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment6 As the number of mobile robots increases,hardware acce
44、l-eration in the video processing function becomes essential for using LIDAR sensor data and video streaming to control them.Benefits of edge computing:very low and ultra-low latency for robot control,high reliability(99.999%),real-time video pro-cessing for SLAM,on-premises data for meeting privacy
45、 and security requirements,and bandwidth savings from locally processing the video dataFactory automation:routing of a fleet of robots and collab orrative robot management This use case is an enhanced version of one described in sec-tion 5.11 of 5.In a smart factory,multiple mobile robots or AGVs ca
46、rry large and/or heavy workpieces from one place to another.They need to collaborate for safely and smoothly moving large items.This is achieved with an application that controls the drives and motions of the individual mobile ro-bots/AGVs.Usually one of them assumes the role of leader and controls
47、the others.The edge computing application is like a director who has the task of coordinating the motions of all of the robots.It in-volves a)capturing the momentary pose,position,and end-of-arm tool angle and position of each robot,b)comparing their status with interpolations to prevent collisions,
48、c)cal-culating new setpoints,andd)sending these to each robot.Low-latency(10 ms)communication is required among the director and mobile robots.The robots can communicate with one other via a sidelink mechanism and choose a leader to communicate with the 5G network on their behalf.Source:5G-ACIA/ZVEI
49、 e.V.Figure 1:Simplified functional architecture for ARVisualizationSensors/camerasMap managementRenderingSLAMClosed loop 1Closed loop 2Object detectionObject trackingWhite Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 7 Benefits of edge computing:real-time communication b
50、etween robots and a centralized control function running on the premisesHumanmachine interfacesAR offloading options for industrial use casesDifferent AR-enabled use cases can have different require-ments and characteristics depending how the AR device user and object move,which of them“owns”the aug
51、mentation,the type of AR device,etc.Figure1 shows a simplified AR functional architecture with the main types of AR function-ality,the relationships among them,and the main control loops.Table1 shows the typical offloading options and corre-sponding network requirements.ARbased product quality and a
52、ugmented field procedure with digital twinsThis use case for optimizing product quality involves develop-ing and merging complex digital models to create an idealized digital twin corresponding to the design intention.Streaming data from actual physical systems is then analyzed to as-certain how clo
53、sely it corresponds.Machine learning is har-nessed to detect and compensate for negative factory-spe-cific anomalies in the production environment.AR-based field procedures of this type can increase both safety and operating efficiency by augmenting field proce-dures and making them mobile.Such a sy
54、stem is able to:guide and assist the user step-by-step through the workflow and record all details of its execution,overlay field data as augmented context in an image of the physical area,generate up-to-date,accessible information on pro-cesses and other important aspects,guide manual inputs and tr
55、igger actions,take pictures and store comments,and support both manual and automatic data capture.The general prerequisites for this use case are low latency (99.999%).Greater safety for manufacturing workers with AI and XR Collaboration between humans and machines is essential for achieving flexibl
56、e production.Specifically,it involves combin-ing humans ability to learn quickly with that of machines to precisely perform repetitive tasks.In the context of their in-teractions,its essential to take steps to ensure the safety of human workers since there is a risk of serious injuries in the event
57、of failures.The environment should therefore be con-stantly monitored by sensors to analyze and assess safety risks posed by objects that are close to humans.This use case comprises various approaches for ensuring the safety of humans who interact with machines in the context of an industrial proces
58、s:1)Closed-loop safety control 2)Ultralow-latency communication 3)Connection of AGVs,XR devices,and machines via 5G wireless 4)Real-time assessment and mitigation of risksSource:5G-ACIA/ZVEI e.V.Offloading optionEndtoend latency budget(communication+computation)Communication latency(uplink and downl
59、ink)Uplink data rateDownlink data rateLow offloading50ms110ms5Mbit/sModerate (augmentation coordinates only)Moderate/high offloading20/30ms110ms560Mbit/sModerate (augmentation coordinates only)Table 1:Benefits of edge computing:offloading optionsWhite Paper Industrial 5G Edge Computing Use Cases,Arc
60、hitecture and Deployment8 Benefits of edge computing:real-time control and monitoring in wireless scenarios,ultralow-latency communication,and greater operating efficiency,reliability,and availabilityLogistics and warehouse automation:mobile positioning and asset trackingProducts,tools,and assets va
61、ry greatly and can be difficult to locate in large,complex deployments.Mobile positioning can help improve this situation,both indoors and outdoors.This use case involves monitoring a large number of items,is calculation-intensive(especially for moving targets),and is typically prone to measurement
62、errors.5G radio access technology delivers various enhanced pa-rameters for improving positioning accuracy,especially when using time-and angle-based methods.Beamforming may increase measurement accuracy,although it depends on the power received.In these use cases,the edge computing re-quirements ar
63、e local processing capabilities,data storage,and application infrastructure.Benefits of edge computing:low latency for wireless real-time control,greater network reliability,and improved privacy due to local control of data.Monitoring and maintenance:video surveillance service In this use case,multi
64、ple video surveillance cameras are de-ployed to monitor some of the most critical process steps in an industrial plant.This adds value by letting users watch the actual production process if its disrupted.The cameras are in turn monitored by operators and optionally also recorded,and the data they g
65、enerate is locally stored in the edge.As a rough estimate,each camera will require an uplink data rate of more than five megabits per second,but the required bandwidth will depend to a large extent on the encoding method used as well as the desired quality.The primary requirements for this use case
66、are local storage and local processing capabilities.Benefits of edge computing:proximity of the edge computing application environment to the use case deployment site re-sults in greater reliability,reduces the required network band-width,opens up additional low-latency automation options,and can al
67、so provide data privacy.3.2 Mapping Use Cases to Edge Deployment Location OptionsTelecoms operators have different options for supporting an edge cloud within their own facilities,on corporate premises,or elsewhere.From a telecommunications network perspec-tive,its possible to distinguish operationa
68、l technology(OT)premises,Near Edge(between about one to 100 km from the data source),Far Edge(up to 500 or 1000 km from the data source),and centralized options(usually from the op-erators perspective,but“edge”can also be interpreted from a global viewpoint).The distance between the OT premises and
69、the Far or Near Edge isnt fixed,since it depends on the vertical applications,customer needs,quality of service,and service agreements in each case.Whats more,an industrial 5G network can be implemented in different ways:it can be a completely isolated private network or integrated in a public netwo
70、rk and deployed as either an on-premises private edge or a public edge.5G-ACIA has also defined an industrial automation commu-nication structure consisting of an OT production domain,an IT enterprise domain,and a public network domain as deployment options for supporting industrial use cases(see se
71、ction2 in 8).In later sections,we will examine feasible deployment op-tions in greater detail.We will also discuss the capabilities of each deployment option and the kinds of edge computing benefits that they can provide for use cases,as well as the factors that determine the relative simplicity or
72、complexity of a given model.In this section,we map use case requirements to edge node location options(presented in figure 2).This is done while keeping in mind that the edge nodes can be located in any of the three industrial automation domains and that,from a telecommunications operator perspectiv
73、e,the“edge”can be on the OT premises or at a Far Edge,Near Edge,or central White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 9 location.This results in clearly defined edge deployment op-tions,which we will refer to from here on.Table2 shows the mappings.Different use ca
74、se requirements(figure2)are divided into three groups.The first group involves use cases that have very stringent requirements in terms of latency,service reliability,and avail-ability.They are also characterized by the need to store sen-sitive private information locally within an enterprise.All of
75、 their devices and entities are distributed locally within the OT domain.Either there is no need for mobility at all,or else the mobility of devices and entities is limited to that one domain.The second group contains use cases that are basically similar to those of the first group but involve a sho
76、p floor infrastruc-ture with two or more interlinked OT production domains.Figure 2:Approach for mapping use case requirements to edge location optionsSource:5G-ACIA/ZVEI e.V.Latency-sensitiveHigh reliability and availabilityData must be stored locallyLocations not geographically distributed(e.g.dev
77、ices inside a single OT domain)Examples of group 1 use cases:1)Closed-loop control2)Mobile robots(no mobility among OT production domains)3)Evolved collaborative robots(no mobility among OT production domains)Edge in single-OT production domain(on premises)Examples of group 2 use cases:1)Mobile robo
78、ts(with mobility among OT production domains)2)Evolved collaborative robots(with mobility among OT production domains)3)Remote access and maintenance(data needs to be stored locally)4)Video surveillance service(data needs to be stored locally)Edge in multi-OT production domains(on premises)Examples
79、of Group 3 use cases:1)Remote access and maintenance2)Video surveillance service 3)Mobile positioning and asset trackingEdge outside the enterprises premisesLatency-sensitiveHigh reliability and availabilityData must be stored locallyLocations not geographically distributed(e.g.devices inside multip
80、le OT domains)Latency-tolerantNo need to store data locallyData dont need to be stored locallyMobility among remote locationsGroup 1 requirementsRequirementsUse CasesEdge locationsGroup 2 requirementsGroup 3 requirementsWhite Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment10
81、 In other words,devices and entities are distributed locally across multiple domains within the enterprise.This makes it important to support mobility between domains(like,say,a mobile robot traveling from one OT domain to another).The third group comprises latency-tolerant use cases with devices an
82、d entities at geographically distinct locations.Its essential to support mobility among these remote sites to the required extent,but there are no data privacy obligations and no necessity to store information locally within the en-terprise.By way of example,the closed-loop control use case describe
83、d in 4 falls into the first group.When robots are stationary or their mobility is constrained to a single OT production do-main,they also belong to the first group.If mobile robots need to move between OT production domains,however,the use case belongs to the second group.Scenarios involving more ev
84、olved,collaborative robots can belong to either the first or the second group depending on the degree of mobil-ity.Remote access and maintenance,video surveillance ser-vices,mobile positioning,and asset tracking are all delay-tol-erant use cases and therefore normally belong to the third group.If th
85、ey involve private,OT-sensitive information that is stored locally within the enterprise,however,they fall into the second group instead.In the first group defined above,the edge is on the premises and contained within a single OT production domain.In the second group the edge is also on the premise
86、s,but it serves multi-OT production domains and can be situated(for exam-ple)within a larger enterprise-wide IT domain.In the third group,the edge is outside the company premises in the pub-lic network domain.The proposed edge location options correlate to user plane function(UPF)deployment options
87、and the correspond-ing advantages and drawbacks,which are described in sec-tion5.2.Table 2:Mapping of edge location options to proposed edge node locations from a telecommunications operators perspective,based on analyses of use case requirements and applying the terminology of the 5G-ACIA white pap
88、er “Integration of Industrial Ethernet Networks with 5G Networks”4Deployment options for edge computingCommunication service providers birdseye viewOn OT premisesFar edge (approx.1 to 100kmNear edge (regional or within city)Central (from CSPs perspective)Definition from 5GACIA white paperOT producti
89、on domainIT enterprise domainPublic network domainEdge node locations proposed in this paper(based on authors analyses of requirements)Single OT production domainMultiple OT production domainsOutside enter-prise premise (but nearby)Central locationSource:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Ed
90、ge Computing Use Cases,Architecture and Deployment 11 4 Example Deployments for Use Cases In this chapter,we analyze various use cases and propose de-ployment options for different edge node locations(as de-scribed in chapter 3)with edge computing architecture com-ponents specified by 3GPP,ETSI MEC,
91、and GSMA OPG.For the sake of simplicity,here we use various high-level terms to describe the components of an edge computing system.Figure3 shows an example:a full high-level stack solution comprising an edge runtime infrastructure,net-work functions,application functions,and management of all these
92、 elements.The point of this exercise is to highlight the functionality and components that are most relevant to actually deploying a use case.The edge location(dark turquoise)can be an“edge in a single OT production domain”,an“edge in multiple OT production domains,”or an“edge outside the enter-pris
93、e premises.”Despite this,we treat the“central site”and“mobile device”as locations for inserting functions in use cases that require them.The edge runtime infrastructure(light turquoise)can be dedicated hardware,the cloud,or any other infra-structure that is capable of running the network func-tions
94、and application functions.The mobile devices have a special device operating system as the runtime infrastructure for their NF and AF components.This infrastructure may also include required firewalls and network routing functions.The network functions(white)are based on a similar concept,namely the
95、 virtual network function defined by ETSI NFV.They comprise all functions of the 3GPP network that are required in order for it to work as intended.They dont include the underlying infrastruc-ture or the functions needed to manage and operate it.The application functions(white)are typically inde-pen
96、dent of the communication network infrastructure but required for the use case.The management layer(black)isnt discussed in detail in this document.It may comprise multiple compo-nents that are presumably able to manage and oper-ate all or parts of an edge computing system.4.1 Factory Automation:Mob
97、ile RobotsFactory automation involves automated control,monitoring,and optimization processes and workflows for a factorys pro-duction systems.It typically comprises discrete applications with specific requirements in terms of low,bounded latency,high availability and reliability,cybersecurity,and f
98、unctional safety.In the following,we examine a specific use subcase,namely that of mobile robots.Source:5G-ACIA/ZVEI e.V.Figure 3:A high-level diagram showing the components of an edge computing systemManagementAFDevice OSMobile deviceNFNetwork functionEdge runtime infrastructure 1Edge locationNetwo
99、rk functionEdge runtime infrastructure 2Application functionApplication functionEdge runtime infrastructure 3White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment12 Mobile robots deliver a whole new level of flexibility in manu-facturing by handling production assets in ways
100、 that include but arent limited to transportation,storage,commissioning,placement,and picking.The related use cases(see 4 for more examples)are typical of factory automation scenarios that can benefit from the edge.In edge-enabled robotics,control intelligence is uncoupled from the robot platform,wi
101、th complex computing tasks and functions being virtual-ized and offloaded to the edge ecosystem.This approach sig-nificantly reduces the computational load for operating mo-bile robots.Centralized control also makes it more efficient to operate both individual robots and entire fleets.Depending on w
102、hich control function is offloaded,cloud-en-abled mobile robot control can involve multiple data streams with differing characteristics and requirements.Servocon-trol calls for reliably constant,deterministic communication with a time lag of between only one and 10 milliseconds,LiDAR traffic is spor
103、adic with a periodicity of 25 to 100 ms,and camera streams need a large bandwidth.Servocontrol requires ultralow-latency communication,which is enabled by the 5G URLLC feature.If a 5G domain is integrated in a legacy time-sensitive networking(TSN)domain in a facto-ry,the 5G-TSN support feature shoul
104、d be applied.The user plane(UP)of the 5G Non-Public Network must be deployed on the factory premises,although the control plane(CP)can be shared with the public network domain.Figure4 shows different function placement options for im-plementing use cases.Instead of dealing with infrastructural and m
105、anagement aspects,it focuses on critical latency lev-els(aspects with less-than-critical latency are surrounded by green borders).To minimize latency,servocontrol and trajec-tory planning components should be deployed in one or more OT edge domains.Since fleet control poses more relaxed latency requ
106、irements,it can be deployed off the enterprise premises in the edge.Capabilities can be exposed and the 5G network configured by invoking procedures such as network resource manage-Figure 4:Different deployment options for placing telecommunications and application functions for a specific factory a
107、utomation use case involving mobile robots.Telecommunication network functions and application components are also shown.Functions in green boxes may tolerate higher latency.Source:5G-ACIA/ZVEI e.V.ExposureServo controlExposureExposureServo controlFleet controlFleet controlTrajectory planningTraject
108、ory planningFleet controlRANRANCPCPCPUP (URLLC)UPUP (URLLC)Dedicated infrastructureDedicated infrastructureAny cloudAny cloudAny cloudEdge with single OT production domainEdge with multiple OT production domainsEdge or central,outside enterprise premisesWhite Paper Industrial 5G Edge Computing Use C
109、ases,Architecture and Deployment 13 ment,e2e QoS management,TSN stream configuration,etc.To minimize latency,the servocontrol and trajectory planning components should be deployed in the edge within one or more OT domains.Because fleet control has more relaxed latency requirements,it can be deployed
110、 in the edge off the enterprises premises.The network control plane and associ-ated exposure capabilities can be located almost anywhere owing to their more relaxed latency requirements.A variety of redundant transmission solutions exist for ensuring ro-bust communication between mobile robots and t
111、he control application in the edge(see 9).4.2 Process Automation:Closed-Loop ControlIndustrial process automation applications typically pose very specific requirements in terms of determinism,reliabili-ty,redundancy,cybersecurity,and functional safety.They ap-ply to controlling the production and h
112、andling of substances such as chemicals,foods and beverages,pulp,etc.in order to boost efficiency,reduce energy consumption,and enhance the safety of facilities.Closed-loop control use cases(see 4)are a typical example of process automation for which the edge can be taken advantage of.Here we consid
113、er a use case involving multiple sensors and actuators installed in a plant,with each sensor performing continuous measurements.The values obtained can be sent to a closed-loop controller application in the edge that per-forms calculations with them for sending control signals to actuators.This use
114、case has quite stringent requirements in terms of la-tency and service availability,but doesnt require interaction with the public network(which would necessitate,for exam-ple,continuous service and roaming capabilities)4.Because user data is only handled locally,the user plane also needs to be with
115、in the non-public network and OT production domains as shown in figure 5.Measurement data received from sensors and processed data sent to actuators can be stored locally in the edge and Source:5G-ACIA/ZVEI e.V.Edge appEdge appRANUPCPRANUPCPCPAny type of cloud or dedicated infrastructureAny type of
116、cloud or dedicated infrastructureAny type of cloud or dedicated infrastructureEdge with single OT production domainEdge with multiple OT production domainsEdge or central,outside enterprise premisesPerimeter of enterprise premisesFigure 5:Function deployment options for closed-loop control.The green
117、 bordered-items are more relaxed in terms of requirements like data privacy,latency,and service availability.Other aspects,such as the edge runtime infrastructure type and management,arent shown here.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment14 analyzed either lat
118、er or in real time,for instance for statistical analysis,optimization and so on.Dedicated data analysis ap-plications for efficiently determining how to optimally set ac-tuators can be moved to an OT production domain in the edge computing platform.The on-premises edge used on this case makes it pos
119、sible to control production more efficiently while keeping the data within the OT production domain in order to maximize security,privacy,bandwidth,availability,and re-liability.4.3 Logistics and Warehouse AutomationA particular localization use case may require a high level of data confidentiality
120、and locality,as well as close to real-time operation in some cases.This may call for a local edge com-puting setup right on the premises,in which case its essen-tial to locally deploy the 3GPP-defined location management function(LMF)as well as all other functions that are needed for operation,such
121、as radio functions,databases,and the access and mobility management function(AMF).Network exposure capabilities are essential for enabling interactions with the application(to provide it with location information or reply to location queries).The currently provided support for precise localization a
122、nd positioning doesnt depend on the user plane function(UPF),only on core network control functions and the exposure ser-vice.Real-time positioning(figure 7)also requires the related 5G core control functions to be located at the network edge,and its recommended that they be close together to preven
123、t network components from introducing any delays.As shown in the non-real-time case in figure 6,its basically possible to deploy each function anywhere,even at a central loca-tion.However,for data security reasons the solutions in light green boxes may only be used if data is allowed to leave the pr
124、emises,for example for remote assistance and monitoring purposes.4.4 Human-Machine InterfacesClose interaction between humans and machines is an essen-tial prerequisite for achieving flexible production.In a given manufacturing environment,each class of machinery should be associated with a certain
125、way of interacting with humans.To achieve this,a closed-loop safety management function monitors the data generated by all equipment(such as mo-bile robots and AGVs)in order to assess and mitigate risks.This function relies on an edge computing setup,since highly complex machine learning models and
126、large data volumes are Figure 6:An example of precise non-real-time localization showing various components of the system besides RAN deployed at different edge and central sites.RANEdge with single OT production domainEnterprise edge outside enterprise premisesCentral locations (outside enterprise
127、premises)5G core and location management functionNetwork exposureEnterprise applicationSource:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 15 RANEdge in single or multiple OT production domainsEdge outside enterprise premisesCentral locations5G core
128、 and location management functionExposure functionsEnterprise applicationFigure 7:An example of precise real-time localization showing the need for core network control,location,and exposure functions to be located close to the device.Datasets,alert mgmtDatasets,alert mgmtDatasets,alert mgmtMachiner
129、y controlMachinery controlComputer visionComputer visionRisk mgmtRisk mgmtRANUPCPRANUPCPCPUPDedicated infrastructureAny cloudDedicated infrastructureAny cloudAny cloudEdge at single OT production domainEdge with multiple OT production domainsEdge outside enterprise premisesFigure 8:Deployment option
130、s for a human-machine interoperation use case.Multiple instances represent different op-tions for deploying use case elements,showing dedicated infrastructure for components that are probably time-critical.Source:5G-ACIA/ZVEI e.V.Source:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use C
131、ases,Architecture and Deployment16 involved.This is particularly important when the machines only have limited processing resources at their disposal.An-other benefit of shifting processing to the edge is that this makes it possible to access data from anywhere in the facto-ry for predicting safety
132、risks and optimizing production.Edge processing is also needed to ensure that equipment required for safe human-machine interactions responds promptly.The applications of this closed loop that rely on edge comput-ing are those that tightly interact to ensure safe human-ma-chine interactions.They inc
133、lude risk management(identify-ing the safest movements and trajectories for machines and equipment),computer vision(detecting and distinguishing different objects in the area),machine control(actual opera-tion of equipment),and alert management(which sends spe-cial safety warnings to workers).Figure
134、8 shows deployment options while making it clear that the user plane and appli-cation functions can be situated in one or multiple OT pro-duction domains.Data asset management and related data channels,including the core network control plane functions themselves,can be relegated to edge sites with
135、more relaxed latency requirements.4.5 Monitoring and MaintenanceMonitoring and maintenance are about tracking certain pro-cesses and/or assets in the context of industrial production without directly influencing the processes themselves.(This is in contrast to the closed-loop control systems typical
136、ly used in factory automation.)They mainly involve applications such as condition monitoring and predictive sensor data-based maintenance,as well as big data analytics for optimizing future parameter sets for a particular process.The remote access and maintenance use case(see 4)is a typical exam-ple
137、 of process automation that can benefit from the edge.In this use case,maintenance-related data from numerous sensors installed on devices is collected and aggregated,then sent to the edge where it is processed to determine whether to trigger maintenance alerts at remote control centers or in diagno
138、stic systems.The logic for activating remote access is also shifted from the device to the edge.The devices and entities for this use case can be installed at geographically distributed locations,and data acquisition and maintenance arent latency-critical processes in any case.Consequently,the edge
139、for this use case could be situated in the public network domain.If sensitive maintenance data is involved,it can be kept both on the premises and in the edge for serving multiple OT production domains.These options are shown in figure 9 below.Edge appRANUPCPEdge appRANUPCPEdge at single OT producti
140、on domainEdge with multiple OT production domainsEdge outside enterprise premisesFigure 9:Function deployment options for remote access and maintenance.Source:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 17 5 Deployment Considerations for Standard C
141、omponents of the Edge Computing SolutionThis section describes options for implementing and deploy-ing the edge computing solution components using standard toolset and functional components.5.1 Edge Runtime InfrastructureSome kind of runtime infrastructure is required to deploy network functions an
142、d applications.Different types may be needed depending on the use case and nontechnical con-siderations,for example when implementing brownfield deployments or leveraging cloud solutions to increase flexi-bility or reduce the total cost of ownership(TCO).Virtual ma-chine-based solutions are still co
143、mmonly employed for 3GPP functions in networks,but cloud-native design is becoming the new standard.To achieve both very low latency and high reliability,however,it may be necessary to implement some network functions in bare-metal systems or proprietary HW infrastructure(this is discussed in 4).App
144、ropriate runtime infrastructure for edge application func-tions can involve proprietary hardware,bare-metal installa-tions,local IT infrastructure,infrastructure extensions for network functions,and/or third-party cloud runtime plat-forms.In many of these cases,special hardware acceleration RAN func
145、tionsIndustrial applicationsOther 5GC functionsUPF for URLLCDedicated HW infrastructure networking performanceDedicated HW infrastructure computing performanceNFVI,cloud-native infrastructureEdge location I.:single OT production domainEdge location IIFigure 10a:Example deployment of use cases requir
146、ing high reliability and ultralow latencyFigure 10b:Example deployment of use cases with low(but not ultralow)latency Industrial applicationsNetwork functions including UPF and 5GCOther IT applicationsCloud solutionEnterprise edge location(covering a complete site)Source:5G-ACIA/ZVEI e.V.Source:5G-A
147、CIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment18 may be required to achieve efficient processing,for example using GPUs,smart network interface cards,P4 switches,per-sistent memory,or real-time operating systems.The benefits and drawbacks of using differe
148、nt infrastructure types in various deployment setups greatly depends on the individual scenario.Flexibility is therefore very important.Fig-ure10a shows a typical deployment architecture for a high-re-liability,low-latency solution in which proprietary infrastruc-ture is needed in order to highly ef
149、ficiently and resiliently handle a networking and computational workload in an in-dustrial or other application.However,it is also possible for a runtime infrastructure environment to be shared by telecom-munications and use case application components,as shown in figure 10b.This is an option for ap
150、plications that have less stringent requirements in terms of latency and reliability.5.2 Edge Deployment Locations and Data Plane Connectivity Path ConsiderationsUPFs play a key role in a 5G network for accessing edge com-puting systems and applications(see 9).In particular,a local UPF steers user p
151、lane traffic toward a data network contain-ing targeted edge applications.This UPF may be part of the edge implementation,as mentioned in 5.In this case the UPF deployment correlates to the edge location.For inte-grating industrial infrastructure with 5G networks,an earlier 5G-ACIA white paper 5 pre
152、sented three possible options for deploying UPF physical locations:in an OT production do-main,in an IT enterprise domain,and in the public network domain.There may also be others.We based the three edge location options presented in section 4.2 on these UPF de-ployment models,especially for cases i
153、n which the edge is within a single-OT or multi-OT production domain on or off the enterprises premises.In the following analysis,we examine possible locations for the UPF.Figure11 shows three different possible locations for the UPF.In this example,the edge application and its runtime environment a
154、re both situated at the same location with the UPF.Although this configuration isnt mandatory,it is a very likely choice for many practical deployments.The figure also shows the data path running from the devices to the appli-cation servers.Here we wont discuss either the type of un-derlying edge ru
155、ntime infrastructure or whether it is some-how shared between the telecommunications and enterprise applications;instead we will consider the advantages and disadvantages of different data path options and locations based on the previously identified requirements.UPF deployed in a single OT producti
156、on domainIn this deployment option,the UPF is located on the prem-ises and integrated in the enterprises shop floor infrastruc-ture.This is a very likely option for a standalone non-public network(SNPN)(scenario 1 in 2).It can also support Public Network Integrated Non-Public Network(PNI-NPN)scenari
157、os 2 and 3 of 2,with either only RAN or both RAN and control plane shared with the public network.One advantage of this approach is that,owing to the fact that the wireless and wired connections with the edge plat-form pass via the local UPF deployed in the OT production domain,it provides the short
158、est data path to the edge ap-plication without traversing interdomain gateways with FW/NAT functionality and therefore supports low latency require-ments.Another is that,because the edge infrastructure and edge application are both situated within a closed enterprise environment(in a private network
159、),privacy-sensitive and se-curity-relevant data can be locally processed and stored.De-ploying them in the same isolated OT production domain also yields benefits in terms of reliability and availability by pre-venting the external environment from causing any failures.The drawback of this option is
160、 that it requires the edge com-puting solution components to be deployed in every existing L2/L3 OT infrastructure,resulting in greater complexity and costlier implementation.If an enterprise has multiple segre-gated OT production domains,each of which has its own edge site,seamless mobility of a wi
161、reless device between domains may cause issues whenever its IP anchor point(i.e.,the UPF)changes.Owing to these advantages and drawbacks,this edge location option is appropriate for delay-critical,highly reliable,secure use cases that dont require seamless mobil-ity outside OT production domains.A t
162、ypical use case that works well here is closed-loop control.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 19 Figure 11:Three different edge locations with a possible data path for each one.In all three cases,the UPF corresponds to the edge site.UPFUPFUPFUPFN6N6N6N3N3
163、N3Local DNLocal DNLocal DNInter-domain GWInter-domain GWInter-domain GWgNBgNBgNBOT production domain 1Nonpublic networkPublic networkIT enterprise domainMNO core network local siteMNO core network central siteOT production domain 2Edge platformEdge app X Edge platformEdge app Y Edge platformEdge app
164、 Z N9N6External DNEdge app X connectivity path Wireless device/entityWired device/entityL2/L3 router/switchEdge app Y connectivity path Edge app Z connectivity path UPF deployed in a multiOT production domainIn this deployment option,the edge runtime infrastructure and associated UPF are both locate
165、d on the premises.Al-though it resembles the previous option,its edge runtime infrastructure is shared by multiple OT production domains.The edge site can be deployed either within the enterprise shop-floor infrastructure or within the IT enterprise domain.This option corresponds to the same deploym
166、ent scenarios as option 2 in 2:SNPN with shared RAN or PNI-NPN with shared RAN and control plane.Source:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment20 This solution provides data privacy and low latency.The edge site is on the premises,its service a
167、rea contains multiple OT production domains,and the wireless and wired data connec-tivity paths are slightly longer than in the first option.How-ever,they are still short enough to support many use cases that require low latency.Data that is sensitive in terms of privacy and security can be safely s
168、tored locally within the enterprise.Since all user plane traffic takes place on the premises and nothing in the external environment can cause any failures,this option is also good in terms of reliability and availability.Another ad-vantage compared to the single OT domain option is that it enables
169、seamless mobility for wireless devices between OT production domains,since all of the latter share a common IP anchor point(the UPF).The drawback of this option is that it requires 5G network elements to be integrated into the existing IT or OT infra-structure,which increases the complexity and cost
170、 of imple-mentation.If there is a FW/NAT gateway between different OT production and IT domains,and if it goes via the wireless domain data plane,latency may be slightly greater and re-liability/availability somewhat lower than in the single-OT production domain deployment option.It might not be app
171、li-cable to use cases that require extremely low latency or very high reliability/availability.A typical use case that fits well here is mobile robots that need to move between OT production domains.UPF deployed outside the enterprise premisesIn this deployment option,the edge and associated UPF are
172、 located outside the enterprise premises in third-party cloud or mobile network operator data centers,which can be either local or centralized.This deployment option is most likely in the PNI-NPN deploy-ment scenario,namely scenario 4 in 2,in which an NPN is host-ed by the public network.In this sce
173、nario,all network traffic is routed via an external network to or from the NPN subscribers.The advantage of this option is that it doesnt require the edge and associated UPF to be integrated in an enterprise IT/OT infrastructure.It therefore incurs low costs and can be quickly deployed.Since this op
174、tion addresses the ordinary MNO network,it can support typical public network features such as mobility within and outside the enterprise.Because the edge data center may be in close geographical proximity,it delivers low-latency communication and computing capa-bilities for data aggregation and ben
175、efits from significant savings in terms of the bandwidth of the transport network.The downside of this option is that any traffic flowing to or from devices within the OT production domain has to pass through two FW/NAT gateways to reach the edge,one be-tween an OT production and an IT enterprise do
176、main and the other between the IT enterprise and the public domain.This results in additional delays and more complex gateway configurations.Whats more,if the edge and the associated UPF deployment are far from the enterprise premises,in oth-er words deeply inside the core network hosted by the MNO,
177、this may result in latency and availability levels that only sup-port use cases that require neither real-time communication nor high reliability.In addition,because the edge infrastruc-ture is located in the public realm,due to privacy and security concerns it may not be the best choice for storing
178、 sensitive private information.Weighing these advantages and disadvantages,this edge lo-cation option is appropriate for less latency-critical use cases and especially for use cases spanning geographically distrib-uted locations for which it isnt necessary to keep sensitive data safely stored inside
179、 the enterprise.A typical use case here is asset tracking.5.3 Data Plane Connectivity Path ConsiderationsLegacy installations can pose challenges when introducing 5G wireless networks and edge computing capabilities for shop floor use cases.In many cases,wired devices will continue to be used,for se
180、veral reasons.Some wired devices wont nec-essarily benefit from wireless connectivity,or else it would be too costly to build wireless versions.Some wired devices,like industrial computers,also host edge application serv-White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment
181、21 ers.In this section,we present different approaches for the single-OT production domain edge location option,some of which can be used instead of other edge location options.For background information on the scenarios,we recommend the 5G-ACIA white paper“Integration of 5G with Time-Sensitive Netw
182、orking for Industrial Communications”4.Data paths that traverse the 5G systemOne approach for connecting to the edge is for the data paths from both wired and wireless devices to pass through the 5G system(5GS).This option makes it possible to take advan-tage of UPF with layer 2 and 3 automation con
183、trol systems as described in Annex A of 5.Since 3GPP treats wired OT devices as trusted but non-5G-capable(N5GC)devices,ac-cording to TS 23.501 and 23.316 a Trusted Non-3GPP Gateway Function(TNGF)should be involved in the connectivity paths.Figure12 shows an example in which both wireless and wired
184、data plane connectivity paths traverse the UPF to the edge,which is deployed within in a single-OT production domain(OT Production Domain 1).The data path for the wireless de-vices runs from the gNB to the UPF there,while the data path for wired devices reaches the UPF via the TNGF.The UPF off-loads
185、 data traffic from wired and wireless devices to the local data network via N6 toward the edge platform and associat-ed edge application.If the edge platform is shared by multi-OT production domains(in other words,deployed in the IT enterprise domain),the UPF plays the role of common IP an-chor and
186、traffic aggregation point,in addition to supporting seamless mobility for devices moving between OT production domains.If the edge and associated UPF are located within the mobile network operator(MNO)network,in other words in the public domain,a UPF near the enterprise is chosen for Figure 12:Wirel
187、ess and wired data plane connectivity paths via the UPF to the edge deployed in an OT production domainUPFN6N3Local DNInter-domain GWInter-domain GWgNBgNBOT production domain 1Nonpublic networkIT enterprise domainOT production domain 2Edge platformEdge app X TNGFWireless device/entityWired device/en
188、tityL2/L3 router/switchWireless data plane connectivity path Wired data plane connectivity path Source:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment22 steering data traffic to the local data network that contains the edge runtime infrastructure and a
189、pplications.The advantage of this connectivity approach is that the edge application and platform communicate with both wired and wireless devices via the 5GS.The edge application and plat-form can therefore be access-agnostic,which makes them easier to design and implement.The drawback is that it i
190、s necessary to integrate wired access into the 5GS,thus in-creasing the deployments overall complexity.Data paths that bypass the 5G systemAnother connectivity approach is to have wireless devices pass through the UPF as in the previous option,but with the wired access to the edge designed to bypass
191、 the 5GS.Then the data path from a wired device passes through the L2/L3 infrastructure and is linked to the edge platform via a gate-way.Figure13 shows an example of a single-OT production domain.This connectivity scenario is also straightforward to use when the edge is deployed in an IT enterprise
192、 or the pub-lic domain.The advantage of this scenario is that it is easy to implement.The downside is that the edge application and platform need to be orchestrated to support communication both with wired devices via LAN and in parallel with wireless devices via a 3GPP network.In addition,if wired
193、and wireless devic-es(such as controllers)need to exchange information,they should do so via the edge application and platform.UPFN6N3Local DNInter-domain GWInter-domain GWgNBgNBOT production domainNonpublic networkIT enterprise domainOT production domain 2Edge platformEdge app X GWWireless device/e
194、ntityWired device/entityL2/L3 router/switchWireless data plane connectivity path Wired data plane connectivity path Figure 13:Wireless data plane connectivity paths via the UPF and wired connectivity paths that bypass the UPF to the edge,which is deployed in an OT production domainSource:5G-ACIA/ZVE
195、I e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 23 Data paths for timesensitive communicationsIn Release 16,3GPP expanded the 5G system architecture to support IEEE 802.1 Time-Sensitive Networking(TSN)as specified by TS23.501.This 3GPP-enabled 5G-TSN integration
196、can be taken advantage of when TSN technology is used to connect devices to the industrial application deployed in the edge.In this case,the 5G network domain is treated by TSN control plane entities as one or more virtual 5G TSN bridges on a User Plane Function base,which ensures seamless inte-grat
197、ion of the 5GS and wired TSN domains.Figure14 shows an on-premises deployment case for this connectivity scenar-io,in which the UPF and edge platform are deployed within the OT production domain.To reach industrial applications deployed in the edge,wireless devices are connected to a 5G virtual brid
198、ge while the N6 in-terfaces of the UPFs are linked to a legacy wired TSN bridge.Figure14 shows a scenario in which the UPF is deployed in the OT production domain,but the solution also works if the edge is in the IT enterprise domain with multiple UPFs deployed in the OT production and enterprise do
199、mains.This connectivity setup can support scenarios with multiple on-premises edge locations.An advantage of this solution is that wired devices can use the legacy TSN network.This eliminates the need for the 5GS to provide wireline access support for wired devices.Beyond this,no other specific feat
200、ures are required because 3GPP has built TSN support into the 5GS.TSN technology can therefore be used as a converged communication solution.One con-straint does exist,however:in practical terms TSN can only be deployed on the premises,so this connectivity option cant be used if the edge is somewher
201、e else.UPFN6N3Local DNInter-domain GWInter-domain GWgNBgNBOT production domainNonpublic networkIT enterprise domainOT production domain 2Edge platformEdge app X Wireless device/entityWired device/entityTSN bridgeWireless data plane connectivity path Wired data plane connectivity path Figure 14:Seaml
202、ess integration of wireless device in wired TSN segment 3GPP 5G-TSNSource:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment24 5.4 Choosing the Data Plane and Edge Application ServerUPF selection and deployment have been discussed in detail in the precedi
203、ng sections.The presented options for placing the UPF and steering data plane connectivity paths show clearly that where a UPF is deployed affects not only latency and availability,but also how easy or hard it is to implement and the associated costs.The network configuration be-tween the UPFs and e
204、dge application servers(EASs)also af-fects latency,redundancy,privacy,and complexity,especially considering that in most cases these correspond to different entities.The method used to select the data plane and thus also the serving EAS(to which the device will actually con-nect)depends on the appli
205、cable standards and the types of applications involved,among other things.The 3GPP specifications describe features for enabling the 5G Core network to select,based on the UEs location,a UPF that is able to steer traffic to the data networks most appropriate local access point,which can be an OT pro
206、duction domain,an enterprise IT domain,or an edge platform off the premises,and supporting application clients(AC)in the UE for iden-tifying the address of the appropriate(e.g.,topologi-cally closest)edge application server(EAS)in the data network.Some deployment options support coordinated UPF sele
207、ction and EAS discovery.The 3GPP specifications describe features for the following:IPtype traffic:When the application layer relies on IP communication,the EAS is selected by the DNS server in response to a query by the UE.The DNS query must be handled by an authoritative DNS server within the data
208、 network.A DNS resolver within the 5G Core network can be configured to receive the DNS query,possibly add more information(like the ECS option defined by 3GPP technical specification#23.548 12),and forward it to the corresponding authoritative DNS server in the data network.When a DNS query is sent
209、,the UPF forwarding the DNS can be:a local UPF(distributed anchor,see 3GPP technical specification#23.548 12)that was previously selected when the UE initiated the PDU session for a particular data network and/or network slice or a local UPF that is dynamically selected based on information derived
210、from the DNS query.This can be done by:establishing a new PDU session and selecting a UPF for a specific data network and/or particular network slice based on the destination address of the DNS query or reconfiguring an existing PDU session by select-ing a local UPF based on a session breakout or di
211、stributed anchor model.The local UPF can be chosen by the 5G Core network(see EASDF in 3GPP technical specification#23.548 12)based on information included in the DNS query or the subsequent DNS response(for example,an EAS IP address).Ethernettype traffic:When the application uses Eth-ernet communic
212、ation,traffic is tunneled between the UE and a UPF chosen by the 5G Core network when establishing the PDU session.UPF selection can be performed for a specific data network and/or network slice requested by the UE.If IEEE TSN features are used for Ethernet-type traffic,the 5G Core network exposes t
213、he API that the TSN Central Network Con-troller(CNC)invokes to configure the 5G system like a TSN bridge.In this case,the CNC steers traffic to the appropriate local access point of the data network(the 5G systems“bridge port”).Applicationlayer methods:An Edge Enabler Client(EEC)in the UE(possibly i
214、mplemented by the device OS)supports the application clients for discovering the address(such as the URI or IP address)of the appro-priate EAS.If the EAS address is a URI,an additional DNS query is required to discover the EAS,reverting to what was described above for the IP traffic case.White Paper
215、 Industrial 5G Edge Computing Use Cases,Architecture and Deployment 25 Figure 15:Private network(PNI-NPN)deployment scenarios according to use-case-specific needsUDRBSS publicBSS privateOSSCentral cloudRAN Single PLMN IDPrivate localPrivate local UPFPrivate centralUDMS-NSSAI of the individual UE is
216、configured in UDRNSSFPCFSMFUPFUPFNRFNSSFSMFSMFPCFNRFAMFAMFUDMAUSFAUSFPCFPrivate 5GC(CP)AMFSMFUPFNSSF must be aware of all network slices(in both public and private 5GC)RAN shared between campus and public networkUPFPublic 5GC5.5 Examples of Private and Public Network Integration ScenariosPNI-NPN dep
217、loyments offer a variety of choices for deploy-ing different network functions to meet specific customer requirements,such as low complexity,local handling of data,URLLC,a positioning service,etc.5G network elements can be also deployed in an edge-cloud environment to achieve greater flexibility,sca
218、lability,and efficacy.When a simple solution is wished,a central private network is an option(see figure 15).In this scenario,all network func-tions are deployed at a single central location.The custom-ers applications and edge can be deployed either centrally or locally.The resulting greater simpli
219、city and absence of local network functions can save resources while also supporting use cases that dont require very great latency.It is possible to deploy user plane functionality(UPF)on a campus(see figure 15 “Private local UPF”).In this case,cus-tomers applications and associated data traffic ca
220、n benefit from not having to traverse the transport network to the cloud.This approach achieves ultra-low latency for traffic flowing between an enterprises devices and the correspond-ing applications.Sometimes there is a greater need for latency in connection with additional services(such as positi
221、oning).This can be the case if more network functions have to be on the factory premises(such as control plane functionality like AMF,SMF,and LMF-see figure 15 Private local).To meet the require-ments for use cases in scenarios of this kind,control plane entities could also be located in the local e
222、dge.They could also handle control plane data on the premises,with only subscription data remaining in the main cloud.Customers applications/EDGEDedicated network function for single campus networkDedicated network function for public network(Network)function shared by public and private networksNFN
223、FNFNFNFSource:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment26 Figure15 shows all of the scenarios in which a separate 5G cloud(devoted exclusively to managing private networks)could be used to provide dedicated services to private networks(such as po
224、sitioning,control plane functionality e.g.AMF,SMF,and LMF,URLLC,etc.,that would not be available to public sub-scribers),thus distributing the control plane.Public networks may be required by law to meet additional requirements(such as support for emergency calls and lawful interception capabil-itie
225、s)that dont apply to private networks and therefore dont need to be enabled in a private 5G cloud.A PNI-NPN can have more than one slice.For example,a pos-sible two-slice solution might have one for URLLC and anoth-er for all other communications within the private network.One of them could include
226、a radio access network(RAN)scheduler with functionality for ensuring the availability of radio resources.Private networks often have access to a dedicated spectrum(like the hotly debated so-called“industry spectrum”in Ger-many),which they may only use for a specific purpose and not for public mass m
227、arket services.Built-in functionality in the RAN associates users of a network slice with the corre-sponding spectrum.This can be accomplished with dedicated cells within the dedicated spectrum that are only available to UEs that belong to the corresponding campus network(iden-tified by a network sl
228、ice ID,the so-called single-network slice selection assistance information or S-NSSAI).These dedicat-ed cells can also be protected using the closed access group(CAG)feature.Alternatively,there can be a single cell encom-passing both mass market and dedicated spectrum.Within such a cell,only UEs wit
229、h radio resources within the spectrum corresponding to UEs S-NSSAI are made available.6 ConclusionsWe have reviewed a number of industrial automation use cas-es and assessed whether they can benefit from or even require edge computing capabilities.The analyzed capabilities as well as the use cases a
230、re based on specifications defined by indus-trial and standardization forums.It has been observed that the capabilities,and therefore also the benefits,of the use cases include latency and locality as well as the capabilities of scal-able computing resources,which can improve data aggrega-tion and s
231、ave bandwidth with local processing.After identifying the relevant use cases,their principal com-munication and computation requirements,and the benefits of using the edge,we have proposed alternative locations for edge implementation and deployment options.3GPP addresses the general requirements fo
232、r implementing and deploying use cases in terms of networking and communi-cation.We have investigated several deployment options in greater detail,especially for the data plane(the User Plane Function(UPF)of the 3GPP 5G Core network)and analyzed the drawbacks and benefits of each option.For cloud an
233、d runtime environments,we have identified several possible implementation options involving multiple edge site instances that host different bare-metal and cloud platforms,without going into detail on the management and orchestration aspects.We have also outlined the interactions between runtime app
234、lication instances and the network and how they affect the deployment options for 3GPP exposure interfaces,and concluded that local exposure of capabilities is required for some use cases.In general,the standard feature set and flexible deployability of 5G enable a wide variety of edge computing con
235、figurations for implementing all currently anticipated use cases.We have also observed that the deployment options excel in terms of flexible use in a variety of runtime environments.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 27 7 Key Terms and Acronyms3GPPThe 3rd
236、 Generation Partnership Project(3GPP)is an umbrel-la term for a consortium embracing a number of standards organizations worldwide that are collaborating to develop globally accepted specifications for mobile telecommunica-tions.As its name implies,it was originally created to es-tablish specificati
237、ons for the third generation(3G)of mobile communication systems.It has continued working on subse-quent generations,including the fifth generation(5G),which is considered in this white paper.5GACIAThe 5G Alliance for Connected Industries and Automation is the globally leading organization for shapin
238、g and promoting Industrial 5G.5GSMART5G-SMART is an EU-funded research project devoted to demonstrating,validating,and evaluating the potential of 5G in actual manufacturing environments.AGVAutomated guided vehicle.AMFAccess and Mobility Management Function.EASEdge application server.Edge computingE
239、dge computing takes storage and computation closer to where data is sourced.Traditionally,cloud computing has taken place in remote data centers.Edge computing moves part of this activity right onto or very close to the premises to achieve data privacy,an actionable feedback loop,bandwidth savings,l
240、ocal management etc.ETSI MECETSI multi-access edge computing.ETSI NFVETSI Network Functions Virtualization.Far EdgeFar Edge is the edge computing infrastructure that is de-ployed farthest from the cloud data center(s)and closest to the users.It can be deployed at enterprises and factories.MEC infras
241、tructure typically gets deployed as a Far Edge.The applications that run there are characterized by very low la-tency,high scalability,and high throughput;typical examples are AR/VR and gaming apps.GCSGuidance control system.GSMA OPGGSMA Operator Platform Group.The GSM Association rep-resents the in
242、terests of mobile network operators worldwide.IPInternet protocol.LMFLocation management function.MNOMobile network operator.NAT Network address translation.Near EdgeNear Edge is an edge computing infrastructure deployed be-tween the Far Edge and the cloud data centers.While Far Edge computing infra
243、structure hosts applications specific to the location where it is deployed,a Near Edge hosts gener-ic services.A mobile network operator service aggregation point can constitute a Near Edge.NEFNetwork Exposure Function.Network functionBased on the Virtual Network Function defined by ETSI NFV,this te
244、rm refers to all functions that are needed for the 3GPP network to operate.It does not include either the underlying White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment28 infrastructure or the functions needed for management and operation.NICNetwork interface card.NPNNon-p
245、ublic network.OTOperational technology.P4Programming Protocol-independent Packet Processors,an open-source,domain-specific programming language for network devices.PDUProtocol data unit.PIDProportional integral derivative(PI/PID).PNINPNPublic Network Integrated Non-Public Network,a 5G network for pr
246、ivate enterprise use.Private edgeThis is when the storage and computing infrastructure is lo-cated on the premises and devoted exclusively to private use.It may be run and managed by a factory operator or service provider or jointly with the owner of the premises or factory.It provides data localiza
247、tion and low latency and supports ap-plications that require time-sensitive data.The usual latency requirement is 15ms.Public Edge or MNO EdgeThis describes a shared infrastructure with storage and com-putation capabilities that can accommodate multiple tenants and is located close to where data is
248、generated.A Public Edge is managed by a service provider,with users receiving access to virtual space and a computing platform for launching ap-plications that communicate with a local data source.Usually the latency is greater than 15 ms.The bandwidth capacity and data security(device to edge)vary
249、depending on the service providers offering.Public cloudThis comprises data centers that are typically operated by cloud service providers such as AWS,Google,Microsoft,etc.A public cloud can be located in the Internet or distributed.Far Edge and Near Edge are examples of public clouds in which a clo
250、ud infrastructure is shared by multiple tenants via secure connections.Nonpublic network(NPN)3GPP has broadened the scope of private networks by in-troducing non-public networks.Previously,a private net-work meant an isolated network with certain device access.Non-public networks have two deployment
251、 options:SNPN and PNI-NPN.An SNPN is a standalone non-public network,while a PNI-NPN is a public network integrated non-public network connected to the operators network.5G-ACIA has defined four different non-public network configurations in the 5G ACIA white paper“5G Non-Public Networks for Indus-t
252、rial Scenarios”2.OPCFThe OPC Foundation,an industry consortium that creates and maintains standards for open connectivity of industrial automation devices and systems.It has established interop-erability standards for securely and reliably exchanging data in the industrial automation space and other
253、 industries.The standards are platform-independent and ensure a seamless flow of information among devices of multiple vendors.Private networkUsually a standalone deployment of a non-public network.The term is also commonly used to designate a network that is under a property owners control,although
254、 it may be man-aged by a third party.RAN Radio access network.SLAMSimultaneous localization and mapping.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 29 SMFSession management function.SNPNStandalone non-public network.TNGFTrusted Non-3GPP Gateway Function.TSNTime-Sen
255、sitive Networking,a set of standards under devel-opment by the Time-Sensitive Networking task group of the IEEE 802.1 working group.Time synchronicityClock synchronicity,or time synchronization precision,is de-fined between a sync master and a sync device.The require-ment on the synchronicity budget
256、 for the 5G system is the time error contribution between ingress and egress of the 5G system on the path taken by clock synchronization messages.UE User equipment(wireless 5G transceiver).UPFUser plane function.URLLC5G ultra-reliable low-latency communications.8 Annex:General Method for Evaluating
257、the Relevance of Edge Computation to Different Use CasesA wide variety of use cases exists,each of which has different requirements in terms of,for example,end-to-end laten-cy,data rate,service continuity,security and privacy,service availability,and reliability.Latency can be important for one use
258、case,while another may require privacy or reliability or both.Edge computing offers excellent capabilities and features that can help meet the requirements of specific use cases.These can include,for example,local data collection,data processing,data storage(including localized content),data analyti
259、cs,data aggregation,and firmware and software updates.For evaluating the“edge computing relevance”of a given use case,we apply the principle that a use cases eligibility for edge computing increases with the number of requirements that an edge systems capabilities can help meet.8.1 Overview of Metho
260、dThe method comprises the four main steps shown in fig-ureA1.Step 1 involves briefly describing the use case and assump-tions about it.The description mentions any related func-tions or operations that could potentially be accomplished with edge computing.The assumptions capture and compare possible
261、 types of computations(e.g.,edge versus remote cloud computing or edge versus device computing)and the type(s)of data communicated(e.g.,control commands,man-agement data,measurement data,video transmissions etc.).Step 2 consists of describing requirements that are critical for or important to a use
262、case.These can be service performance requirements like end-to-end latency or service availability,or other requirements unrelated to performance such as off-White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment30 loading of heavy computational loads to another device,cen-tr
263、alized coordination,etc.Step 3 is identifying any“edge computing”capabilities and features that may help support the specific use case require-ments defined in the second step.Examples of these edge computing capabilities and features are:Local data capture Local data processing Local data storage(i
264、ncl.localized content)Local data analytics Local data aggregation from a large number of sources Local firmware/software updates(from edge devices to end devices)Bandwidth savings(localization)Offloading of heavy computations(from end device(s)to edge computing)Coordination of end devices and entiti
265、es Security PrivacyFigure 16:The main steps of the method used to evaluate the relevance of edge use casesReduced E2E latencyReduced transport network traffic loadImproved data privacyImproved securityImproved reliability and availabilityLocal data processingLocal data storageLocal data aggregation/
266、collectionLocal data analyticsLocal firmware/software updatesEdge benefitsEdge capabilities43Short UC descriptionUC assumptionsData must be stored locallyInitial use case setting1E2E latencyUser experienced data rateData privacySecurityReliability and availabilityData rate experienced by user2Source
267、:5G-ACIA/ZVEI e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 31 Step 4 involves mapping edge computing capabilities to rel-evant use case requirements to identify achievable benefits:for example,that local data processing in the edge cloud can help meet a use case
268、s low latency requirement.The next sections explain how the four steps of this method work in detail,taking the case of a mobile robot as an example.8.2 Examples of Edge Relevance Analysis Based on the Defined Method8.2.1 Use Case Description and Assumptions(Step 1)The mobile robot use case is descr
269、ibed in detail in 3GPP TS 22.104 and TR 22.804.A mobile robot is a programmable machine thats able to perform a wide variety of tasks while following programmed paths.Automated guided vehicles(AGVs)are a subgroup of mobile robots.Mobile robots and AGVs are monitored and controlled by a guidance cont
270、rol system(GCS),which assigns jobs to them,sends them up-to-date process information,and manages traffic while preventing collisions TS 22.104.Edge computing infrastructure(or a remote cloud)can be used to host remote control system functionality of this kind(for example,GCS).In particular,edge comp
271、uting capacities can store and process video and images received from camer-as of the AGVs.After processing the data,control and man-agement commands are sent from the edge to the AGVs.For instance,if the edge application recognizes an obstacle in an AGVs path or detects a malfunction,it sends a com
272、mand in-structing the AGV to perform an emergency stop.The generic assumptions for this use case are as follows:Remote control system(guidance system)functional-ity is hosted in the edge infrastructure.Only communications between mobile robots and the edge computing infrastructure are considered(ind
273、ividual mobile robots dont communicate with one another directly).In other words,the edge controls a number of mobile robots and communicates bidirec-tionally with each one.Uplink communications(from a mobile robot to the edge)include(1)process-related measurement data and(2)video or image data.Down
274、link communications(from the edge to a mobile robot)include(1)process data for controlling and managing mobile robots and(2)emergency stop com-mands(such as the alarm command).The following analysis sheds light on the relevance of edge computing by considering a case in which GCS functionality can r
275、eside in the remote cloud.8.2.2 Requirements for the Mobile Robot Use Case and Edge Capabilities for Meeting Them(Steps 2 and 3)The service performance requirements for this use case are presented in TS 22.104 4,A 2.2.3.The most important KPI values are an end-to-end latency of between one and 10 ms
276、 for periodic machine control communication,communication service availability greater than 99.9999%,and service reli-ability equivalent to a mean time between failures of about 10 years with periodic machine control communications.These values should be achievable for a large number of mobile ro-bo
277、ts(up to 100 per square kilometer).The video streaming uplink requires a data rate above 10 Mbit/s.High bandwidth is essential for this use case,since the video streams from the cameras mounted on all the robots can generate heavy network traffic.Two additional requirements for this use case are dat
278、a pri-vacy and security,which are covered in the 5G-ACIA White Paper“Key 5G Use Cases and Requirements”6 and 3GPP TS 22.261.Summing up,the requirements for this UC(the second step of the method)include end-to-end latency,reliability,avail-ability,data privacy,security,and high bandwidth.White Paper
279、Industrial 5G Edge Computing Use Cases,Architecture and Deployment32 From the list of edge computing capabilities and features(see section 8.1 above),we have selected the following three capabilities(the third step of the method)that may poten-tially help meet the requirements of this use case:local
280、 data processing,local data storage,and local data aggregation/collection.Local data processing:The edge application locally processes video data or images received from cameras on mobile robots and control data received from sensors on mobile robots.Local data storage:Video data or images received
281、from the cameras on mobile robots and control data received from sensors on mobile robots can be stored locally in the edge.The stored data can be used for statistical purposes,caching,analyses,etc.Local data collection and aggregation:When there are many mobile robots,the images,video data,and meas
282、urement data are locally collected and aggregated for processing in the edge.8.2.3 Benefits of Edge Computing for the Mobile Robot Use Case(Step 4)To illustrate the achievable benefits,the fourth step of the method maps edge computing capabilities onto relevant use case requirements.1)Local data pro
283、cessing helps meet the following requirements:Reduced E2E latency(benefit#1):There is no need to send images,video data,or control information deeply into the core network for processing.This results in a shorter end-to-end roundtrip latency for communications between the edge and mobile robots.Note
284、 that low latency is very important for control commands in general and especially for alert commands such as emer-gency stop.Improved reliability and availability(benefit#2):The edge is hosted within one or more reliable enterprise environments.This precludes failures induced by external events.The
285、 communication data path(TS 22.104)to the edge infrastructure is shorter than that to the remote central cloud,which reduces the risk of failures.Improved data privacy(benefit#3):The edge infrastructure is located inside a closed enterprise environment(a private network).This ensures the safety of s
286、ensitive private information flowing between the edge infrastructure and mo-bile robots.Improved security(benefit#4):The communication link between the edge cloud and mobile robots runs within an enterprise envi-ronment(a private network controlled by OT)and is therefore secure.In contrast to this,a
287、 communica-tion link between mobile robots and a remote cen-tral cloud includes an external(third-party)network that may not always have guaranteed security.2)Local data storage helps improve privacy similarly to the local data processing feature(benefit#4):Edge computing data is stored within a clo
288、sed enterprise environment(a private network)for later use for statistical purposes,caching,analysis,etc.This ensures that sensitive private information is kept locally where external third parties cant access it.3)Local data collection and aggregation help meet the need to save bandwidth(benefit#5)
289、:The cameras of mobile robots produce a considerable volume of video streaming data traffic(a mobile robot like that in use case 4 in A 2.2.3 of TS 22.104 can generate more than 10 Mb/s).Collecting and aggregating these data streams from multiple mobile robots and localizing them to the edge infrast
290、ructure instead of remotely routing them to the central cloud uses less bandwidth and significantly reduces the network load.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 33 8.2.4 Summary:the Relevance of Edge Computing to Mobile Robot Use CasesEdge computing capabil
291、ities can help meet five requirements that have been defined for mobile robot use cases by 3GPP TS 22.104,TS 22.261,and the 5G-ACIA white paper“Key 5G Use Cases and Requirements”6:end-to-end latency,reliability and availability,reduced network traffic load(bandwidth sav-ings),security,and data priva
292、cy.Figure17 shows the benefits that taking advantage of differ-ent edge capabilities can yield in connection with mobile ro-bots/AGVs.This use case can therefore be considered a prime candidate for deriving benefits from edge computing.Figure 17:Relationship of edge capabilities and benefits for mob
293、ile robots use caseImproved reliability and availabilityReduced E2E latencyImproved data privacyImproved securityReduced transport network traffic loadRequirements that the edge can help meetLocal data processingLocal data storageLocal data aggregation/collectionEdge capabilitiesSource:5G-ACIA/ZVEI
294、e.V.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment34 9 References1 5G-ACIA White Paper,“5G for Connected Industries and Automation,”2nd edition,published in March 2019.2 5G-ACIA White Paper,“5G Non-Public Networks for Industrial Scenarios,”published in July 2019.3 3GP
295、P Technical Report 22.804,“Study on Communication for Automation in Vertical Domains,”Rel.16.4 3GPP Technical Specification 22.104,“Service requirements for cyber-physical control applications in vertical domains”,Rel.18.5 3GPP Technical Report 22.832,“Study on enhancements for cyber-physical contro
296、l applications in vertical domains,”Rel.17.6 5G-ACIA White Paper,“Key Use Cases and Requirements,”published in May 2020.7 5G-ACIA White Paper,“Integration of 5G with Time-Sensitive Networking for Industrial Communications”,published in June 2021.8 5G-ACIA White Paper,“Integration of Industrial Ether
297、net Networks with 5G Networks,”published in November 2019.9 3GPP Technical Specification 23.501,“System architecture for the 5G System(5GS),”Rel.17.10 3GPP Technical Specification 23.401,“General Packet Radio Service(GPRS)enhancements for Evolved Universal Terrestrial Radio Access Network(E-UTRAN)ac
298、cess,”Rel.15.11 ETSI White Paper no.28,“MEC in 5G networks,”published in June 2018,ISBN no.979-10-92620-22-1.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 35 5GACIA White PaperIndustrial 5G Edge Computing Use Cases,Architecture and DeploymentContact5G Alliance for Co
299、nnected Industries and Automation(5G-ACIA),a Working Party of ZVEI e.V.Lyoner Strasse 960528 Frankfurt am Main GermanyPhone:+49 69 6302-292Email:info5g-acia.orgwww.5g-acia.org Published byZVEI-German Electro and Digital Industry Association5G Alliance for Connected Industries and Automation(5G-ACIA)
300、,a Working Party of ZVEIwww.zvei.orgPublished in February 2023Last updated in March 2023Design:COBRAND ZVEI e.V.This work,including all of its parts,is protected by copyright.Any use outside the strict limits of copyright law without the consent of the publisher is prohibited.This applies in particu
301、lar to reproduction,translation,microfilming,storage,and processing in electronic systems.Although ZVEI has taken the greatest possible care in preparing this document,it accepts no liability for the content.White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment36 As of March 202310 5G-ACIA MembersWhite Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 37 White Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment38 NotesWhite Paper Industrial 5G Edge Computing Use Cases,Architecture and Deployment 39 Notes5GACIA.org