1、AbstractDigital twin,with emerging potential 6th generation(6G)technology,can be a new driving force for 6G.It can benefitwidely in the scenarios of networks,industry,agricultural andhuman bodies.Digital twin technology is an organic combinationof many supporting technologies.Moreover,some solutions

2、 to theadoption of digital twin are presented,such as mobile network,intelligent transportation,and Internet of things(IoT).Intent awaredigital twin 6G networks can be a full life-cycle solution,drivenby the knowledge graph.Keywords:digital twin，6G，Digital twin 6G networks.Table of Contents1.Overvie

3、w.11.1.History of Digital Twin.11.2.Current Researches on Digital Twin.21.3.Digital Twin+6G Fusion and Development.42.Application Scenarios of Digital Twin in 6G.52.1.Twin Network.52.2.Twin Industry.62.3.TwinAgriculture.62.4.Twin City.72.5.Twin Body.83.Key Technologies of Digital Twin.93.1.Technical

4、 Framework of Digital Twin.93.2.Main Supporting Technologies for Digital Twin.114.Digital Twin Solution for 6G.134.1.Digital Twin Network for 6G.134.1.1.Lifecycle Twin Network.134.1.2.Knowledge-driven Twin Network Control.184.1.3.Intent-Driven Twin Network Management.264.1.4.Terminals data-related s

5、upporting technologies.324.2.Other Applications of the Digital Twin.334.2.1.Smart Transportation.334.2.2.IoT.424.2.3.UE Twin.475.Development Prospect of Digital Twin.52Acknowledgements.53References.53Abbreviations.5511.Overview1.1.History of Digital TwinThe twin concept was originated from NASA in i

6、ts Apollo program in 1970.NASA attempted to create a mirror system to monitor physical spaces that are notaccessible to people,that is,to build two identical spaceflight vehicles,one launchedto the space for missions and one left on the Earth(mirror system)to reflect theworking state of the vehicle

7、in space so that engineers could analyze and handleemergencies occurred in space 1.For example,by simulation of the mirror system,astronauts were instructed to rebuild oxygen tanks that exploded in outer space 2.These two spacecraft were real physical entities.In 2002,Professor Grieves of the Univer

8、sity of Michigan for the first timeproposed the idea of Digital Twin(DT)in the course of Product LifecycleManagement.Later he elaborated in a whitepaper 3 that the digital twin was mainlycomposed of three parts:physical objects,virtual objects,and information flowbetween physical objects and virtual

9、 objects,as shown in Figure 1.In 2003,Frmlingproposed an agent-based system structure where each product was associated with avirtual counterpart or agent,and the agent kept pace with the physical counterpart viathe Internet for a one-to-one correspondence between the digital twin and the physicaltw

10、in(bijection)4.Figure 1 Digital Twin Model Proposed by GrievesIn 2010,the term Digital Twin was officially used by NASA in its technicalreportand definedasan integratedmulti-physics,multi-scale,probabilistic2simulation of a vehicle or system 5.In 2012,NASA and the U.S.Air Force jointlypublished a pa

11、per on the digital twin,pointing out that the digital twin would be oneof the key technologies driving the development of future spaceflight vehicles.In thesame year,Frmling put forward a conceptual model showing how to use the digitaltwin as a virtual sensor to predict the life of a spaceflight veh

12、icle and ensure itsstructural integrity 6.In recent years,with the development of new generations of informationtechnologies such as the Internet of Things(IoT),big data,cloud computing,andartificial intelligence(AI),the implementation of digital twins has gradually beenimplemented in place.At this

13、stage,the digital twin application is relatively mature inaerospace,intelligent manufacturing,smart city,and some other fields,and remains atthe growing phase in healthcare,agricultural development,etc.Governments,enterprises,and organizations in different countries value the digital twin highly tha

14、t itis becoming a new driver to national digital transformation,a new direction for thebusiness layout of multinational enterprises,and a new focus of global ITdevelopment 7.1.2.Current Researches on Digital TwinCurrently,there is no consensus among the academic and industrial circlesregarding the d

15、efinition and connotation of the digital twin.Their respectiveunderstandings of the digital twin connotations are given inTable 1.Table 1 Comparison of Understandings of Digital Twin Connotations by the Academic andIndustrial CirclesSourceInstitutionConnotation of Digital TwinASurveyonDigitalTwin:De

16、finitions,Characteristics,Applications,andDesignImplicationsBresciaUniversity,ItalyDTs can be defined as(physical and/or virtual)machines or computer-based models that aresimulating,emulating,mirroring,or“twinning”the life of a physical entity,which may be anobject,a process,a human,or a human-relat

17、edfeature2.3DigitalTwinandItsPotentialApplicationExplorationBeihangUniversity,etc.Digitaltwin,asa technologyofintegratingmulti-physics,multi-scale and multidisciplinaryattributes,characterizedbyreal-timesynchronization,faithfulmapping,andhighfidelity,could realize interaction and integrationbetween

18、physical space and virtual world 7.DigitalTwinWhitepaperChinaCenterfor InformationIndustryDevelopmentThedigitaltwinreferstoacomprehensiveapplication of information technologies such asperception,computing,and modeling to describe,diagnose,predict,and make decisions on physicalspacesthroughsoftwarede

19、finitions,thusachieving the interactive mapping between thephysical space and cyberspace 8.White Paper ofDigitalTwinApplicationChinaElectronicsStandardizationInstitute,RootcloudTechnology Co.Ltd.The digital twin is a digital expression of a specialphysical entity or process with data connectionsthat

20、 ensure the convergence between the physicalstate and the virtual state at the same rate andprovide an integrated view of the entire lifecycleof the physical entity or process,helping tooptimize the overall performance 1.DigitalTwinComputingDTCInnovationForumDigital twin computing is a new computing

21、 modelthat,byexecutingvariousoperationsandcombining digital twins freely,reproduces the realworld in an unprecedented,new,large-scale,andhigh-precision way and makes new interactionstrue in cyberspace,including interactions withinhuman beings,surpassing the physical copy in thereal world 9.White Pap

22、er forDigital Twin of5G CitiesAsiaInfo,Migu,DigitalTwinConsortiumThe digital twin is an integrated multi-disciplinary,multi-physicalquantity,multi-scale,andmulti-probability simulation process for mappingin digital space upon the full use of physical4models,sensor updates,operation history,andother

23、data,to reflect the lifecycle process of thephysical counterpart.The digital twin is a conceptbeyond reality and can be regarded as a digitalmapping system for one or more important andinterdependent device systems 10.According to the definitions listed inTable1,the digital twin has the following ty

24、picalcharacteristics:1)Bidirectional precise mapping(bi-mapping)means that the data flow isbidirectional between physical objects and twins.Physical objects input data totwins for modeling;while twins give feedback to physical objects for prediction,control,and decision-making.Precise mapping is to

25、fully present,accuratelyexpress,and dynamically monitor physical objects on twins.2)Real-time means that a real-time relation between physical objects and twins canbe established so that twins characterize physical objects as the time axis variesand enable the real-time mapping of physical objects.3

26、)Lifecycle refers to the whole process of a product in which the digital twin canrunthrough,includingdesign,development,manufacturing,service,maintenance,scraping,and recycling.1.3.Digital Twin+6G Fusion and DevelopmentThe digital twin will be closely integrated and mutually promoted with 6Gtechnolo

27、gies.On the one hand,6G technologies enable the data and feedback transmissionwith ultra-large capacity and ultra-low latency for the digital twin at the interactionlayer,promoting better applications of the digital twin technology.On the other hand,the digital twin gives new ideas and solutions for

28、 the researchof 6G key technologies.For example,in the digital twin network,the digital twintechnology is applied to the network field,through the virtual expression of thephysical network,to analyze,diagnose,simulate,and control the physical networkbased on data,models,and interfaces.This yields lo

29、w-cost network optimization,intelligent network decision-making,efficient network innovation,and closed-loop5management of the network throughout its lifecycle 11.The digital twin technologyis employed to model the indoor environment,by virtue of tunable metasurfaces ofgraphene,to control the propag

30、ation paths of indoor THz signals and reduce theprobability of blocking THz signals 12.The edge network-based digital twins reducethe average offloading latency,offloading failure rate,and service migration rate ofthe edge network 13.This white paper addresses the application scenarios of the digita

31、l twin inChapter 2,key technologies of the digital twin in Chapter 3,the digital twin-enabled6G network and solutions for some vertical industries in Chapter 4,and the outlook tothe future development of the digital twin in Chapter 5.2.Application Scenarios of Digital Twin in 6G2.1.Twin NetworkIn th

32、e 6G era,the digital twin technology will be widely applied tocommunication networks.With up-and-up perception+modeling technologies,virtual digital twins of real physical networks will be built to offer capabilities in thequery in the real world+prediction in the virtual world+interaction between r

33、eal andvirtual worlds.The network digital twin will be implemented by virtue of many technologiessuch as perception,modeling,data processing,and control,including the constantlydeveloping technologies for network measurement and data collection;more refinednetwork awareness technology to perceive th

34、e network state;unified data platformtechnologies providing underlying data supports for the internal and external bybuilding unified and reliable data platforms through cloud-based technologies;network model technologies for NE&topology modeling and simulation of networkoperationthroughdigitalappro

35、aches;andnetworkmanagement&controltechnologies connecting the interactive channel for management&control operationsbetween virtual digital twins and real physical entities based on standard andautomatic interfaces.The implementation of the digital twin technology in the future network willenhance 6G

36、 network capabilities in multiple aspects.The powerful reality restoration6capability will provide more comprehensive network states and more accurate faultlocating.The flexible simulation capability will offer easier strategy simulation,safersolution pre-evaluation,and more intuitive result visuali

37、zation relying on accurate,virtual,and efficient mechanism modeling.The handy management&controlcapability will enable simplified,automatic,and visual operations,greatly reducinglabor costs.2.2.Twin IndustryThere are some relatively mature application cases of the digital twin and 5G inthe manufactu

38、ring industry,for example,workshop status information display andanalysis management,M&E product design optimization,machine tool faultprediction&health management,etc.14.The digital twin technology is still in thebud,and it will take two or three decades to achieve the digital twin fusion andintera

39、ction across platforms in various fields.It is expected that the digital twin fusionmay become active in the 6G era 10 years later 15.The twin industry in the 6G era will not be limited to the concept of intelligentplant,but develop a new form of twin industry specific for the future society.Strateg

40、ically,based on real-time dynamic analysis of market data,the industrialsolutions for production,storage,and sales will be developed and updated tomaximize the industrial benefits,while achieving highly industrial integration,andeffectively coordinating and optimizing all business activities of the

41、whole industry.From the aspect of technologies,based on data and models,technologies such as AI,big data,6G,cloud computing,and edge computing will be applied to forming asmart manufacturing mode coordinated with labors,machines,and materials 16.2.3.TwinAgricultureThe digital twin technology may sim

42、ulate and deduce the agricultural productionprocess so that some adverse factors can be eliminated in advance,further improvingagriculturalproductivityandutilizationefficiency.Moreover,byintegratingblockchain technology,the digital twin can include information about enterprises,certification bodies,

43、sales enterprises,and logistics and storage enterprises into aunified and shared chain to ensure that sources are identifiable,products are trackable,7and persons in charge are held accountable.Meanwhile,the digital twin can closelyfollow the urban consumption demands and the supply of agricultural

44、products,largely energizing the agricultural product flow and promoting the construction ofsmart agriculture ecology.Big data,IoT,cloud computing,and some othertechnologies will support larger-scale UAVs,robots,environment detectors,and otherintelligent devices,realizing the full connection between

45、things and between humanbeings and things,and making great differences in crop farming,forestry,husbandry,and fishery 16.2.4.Twin CityThe digital twin city concept was incubated in 2017 and 2018,its technicalarchitecture was conceived in 2019,and the digital twin city was officially launchedin 2020.

46、As national strategies are issued,and local planning put in place,enterprisessolutions are formulated,academic researches are arranged proactively,marketspurted,industrial ecology is built,application scenarios are gradually improved,andthe global consensus is reached.In short,the digital twin city

47、is the only way and thefutures choice for the construction and development of new smart cities.Specifically,the digital twin city consists of three horizontal layers:newinfrastructures,intelligent operation center,and intelligent application systems,andtwo vertical layers:city safety line and standa

48、rd specifications.The digital twin cityhas nine core capabilities:IoT perception control,all-element digital expression,visualization presentation,data fusion supply,spatial analysis computing,simulation&deduction,virtual-real fusion&interaction,self-learning&self-optimization,andcrowd innovation ex

49、pansion.However,as the digital twin city is implemented in place,a number of problemsare revealed,such as insufficient depth of typical application scenarios,repeatedconstruction of city information model(CIM)platforms,difficulty in coordinatingtime-space data standards,constraints of critical techn

50、ologies,etc.The CIM-basedcoordination,interconnection of data specifications and standards,development oftypical application scenarios and market demands,and ecological cooperationmechanisms,all are decisive to the development of the digital twin city in the nextstage 17.82.5.Twin BodyThe digital tw

51、in-based body area network(BAN)technology will be one of theimportant features of the next generation mobile networks(NGMN),and the medicaltwin on this basis will be the main direction of future medical businesses.Unlike thedigital twin in industrial manufacturing,the digital twin integrating person

52、al wirelesscommunications is people-oriented and focuses on the services for human beings.Thedigital twin will combine with the BAN to diversify the NGMN features and make itone of the infrastructures for critical technologies of other communications.In the 6G era,the BAN consisting of wireless sens

53、ors intensively deployed inand outside human bodies will collect,analyze,and model human body information inreal time for the digital twin of human beings,i.e.,personalized human digital twin.The human digital twin will facilitate efficient research on virus mechanisms andorgans and help doctors to

54、make accurate surgical predictions.Imaging when doctorsare performing an operation,the human digital twin will simulate the conditionchanges of the patient after being operated on at different positions,so as to assistdoctors doing best in the operation.Even after the patient is discharged,the hospi

55、talcan still provide the patent with follow-up health management based on the change inthe human digital twin of the patient.The human digital twin will also play asignificant role in medical research.For example,the extremely complicated humanbrain makes it more difficult to track and study brain a

56、ctivities.Researchers focusesand difficulties are always the way the brain thinks and the function of motionperception.The application of the digital twin in brain researches can ease theexperimental simulation and help experimenters to discover the secrets in the brain.Similarly,the attack of virus

57、es and bacteria can be simulated by some control over thehuman digital twin to provide a reference for the study of viral mechanisms.The four key technical links of the human digital twin are data collection,transmission convergence distribution,collaborative computing&digital twin,andlarge network

58、communication interaction.Data collection is to collect the physicalinformation of human beings with different sizes of sensors,cameras,and otherinternal and external data collectors.Data convergence is to transfer the collected datato the data center through molecular communication or traditional e

59、lectromagneticcommunication.The computing is to compute and analyze the converged data with9the technologies such as collaborative computing,digital twin,and holographicpresentation.The communication and interaction with large networks are to transmitdata to large networks for storage or further scr

60、eening&analysis.For the digital twinand real-time interaction of all human information,higher requirements are imposedfor indicators such as network bandwidth,latency,reliability,and security.3.Key Technologies of Digital Twin3.1.Technical Framework of Digital TwinThe current technical frameworks of

61、 the digital twin mainly include thefollowing:1)In the Digital Twin Whitepaper 8,the digital twin technology architecture isdivided into the physical layer,data layer,model layer,and function layer.Thephysical layer consists of physical entities.The data layer involves data collection,data processin

62、g,and data transmission.The model layer includes mechanismmodels and data-driven models.The function layer covers description,diagnosis,predictions,decision making,etc.2)In the White Paper of Digital Twin Application 1,the digital twin ecosystemconsists of the basic support layer,data interaction la

63、yer,model building andsimulation analysis layer,common application layer,and industrial applicationlayer.The basic support layer contains specific devices,including industrialequipment,urban construction equipment,transportation means,medical devices,etc.The data interaction layer involves data coll

64、ection,data transmission,dataprocessing,etc.The model building and simulation analysis include datamodeling,data simulation,and control.The common application layer coversfour aspects:description,diagnosis,prediction,and decision-making.Theindustrial application layer includes a variety of applicati

65、ons including intelligentmanufacturing and smart city.3)In the publication Digital Twin Computing 10,digital twin computing combinesvarious digital twins freely to implement a new world in an unprecedented,large-scale,and high-precision way,making new interactions true in cyberspace,including more c

66、omplicated interactions between humans,and surpassing the10entities in reality.The twin computing technology architecture consists of realspace,information/physical interaction layer,digital twin layer,digital worldpresentation layer,and application layer.The real space contains physical entities.Th

67、e information/physical interaction layer collects data and gives feedback on thecontrol information.The digital twin layer generates and maintains digital twins.The digital world presentation layer produces the digital twin derivatives to builda virtual world.The application layer uses the digital w

68、orld presentation layer todeploy and execute applications.To sum up,the digital twin architecture is summarized as shown in Table 2.Although the definitions for each layer are different in 1 and 7,the contentsexpressed are substantially consistent.The architecture in 10 is for differentpurposes and

69、emphasizes more on the importance of the interaction layer andpresentation layer,that is,the digital twin can produce virtual social functions throughreplication,fusion,exchange,etc.Table 2 Comparison of Existing Digital TwinArchitecturesSourceInstitutionContainPhysicalLayerContainDataLayerContainMo

70、delLayerContainFunctionLayerContainApplicationLayerContainPresentationLayerContainInteractionLayerDigital TwinWhitepaperChina CenterforInformationIndustryDevelopmentWhite PaperofDigitalTwinApplicationChinaElectronicsStandardizationInstitute,RootcloudTechnologyCo.Ltd.Digital TwinComputingDTCInnovatio

71、nForum11From the above,it can be seen that the digital twin is a complex technicalecosystem consisting of at least three layers of abstract architectures:physical entitylayer,digital twin layer,and application layer.3.2.Main Supporting Technologies for Digital TwinFrom the bottom to the top of the h

72、ierarchical digital twin architecture,the mainsupporting technologies are the IoT,5G/6G,big data,modeling,simulation analysis,cloud computing,edge computing,AI,API,VR/AR,etc.Specifically,at the data layer,data collection needs IoT,data transmission requires 5G/6G,and data processing usesbig data.At

73、the model building and simulation analysis layer,modeling andsimulation analysis technologies are required.At the function layer(commonapplication layer),implementing the functions,such as description,diagnosis,prediction,and decision making,needs AI,cloud computing,edge computing,andsome other tech

74、nologies.At the function and industrial application layer,somevisualization technologies such as API and VR/AR are needed.Besides,specialattention needs to be paid to security issues in the application of the digital twin,andblockchain technology is one of the methods to solve such security issues.M

75、ain supporting technologies for the digital twin are related to applicationscenarios.The digital twin application in different scenarios generally requires thefollowing technical supports.1.Data collection,transmission,and storageData is the basic element of the digital twin.It comes from target loc

76、ations,physical entities,control,or service digital systems.Data collection,transmission,andstorage are cornerstones of the digital twin.Twin data integrates physical perceptiondata of all elements,all businesses or all processes,and massive data generated bymodels.It is characterized by multiple so

77、urces,multiple types,and multiplestructures.The physical network technology implements data perception and acquisitionfrom the control or service systems of different hardware devices.After parsing thedata format,it cleans and sorts out massive raw data,and initially screens outreasonable and reliab

78、le data output to subsequent digital twin systems.With unified orcustom interfaces,it enables large-capacity,high-reliability,high-rate,and stable datatransmission in 5G and 6G networks.For example,the twin city requires a larger12number of connections,i.e.,107/km2,making the uplink regional traffic

79、 density up toTbps/km2level.The medical twin needs shorter latency and higher reliability,i.e.,0.1-1ms and 99.99999%,as well as lower energy consumption 18.Selecting appropriate big data storage solutions in different application scenariosimproves the reliability of massive storage and the speed of

80、data reading&writingwhile reducing the costs.2.Twin modelingThe digital twin can solve such non-linear and uncertain issues that are difficultto be solved or cannot be solved by traditional models.The core element of the digitaltwin is to build a matching model.Based on the constantly generated real

81、-time data,with traditional models and AI/ML,the digital twin can not only accurately analyze,train,and predict the properties and states of physical entities,but also pay moreattention to dynamic changes in twin data for iterative updates of models,making thedigital twin more valuable to be continu

82、ously improved.Digital twin models are now divided into two categories:general models andspecial models.General models are built to mainly study the possible unified modelconcept,model development methods,modeling languages,and specific tools,anddescribe the lifecycle control,general system behavior

83、s,and workflow of physicalentities with unified methods.Special models are built to focus on the implementationof digital twin projects,and the possible use of different model development methodsand tools in different niche markets,for example,the manufacturing and qualitysupervision in traditional

84、manufacturing,biopharmaceutical R&D,etc.The digitaltwin network,for example,models functions,NEs,and networks through big dataprocessing based on data collection,which converts issues difficult to solve at eachstage of the network into the digital world for solutions,enabling network autonomy.The AI

85、/ML-based network knowledge graphs and user knowledge graphs enable theverification and optimization of network allocation solutions based on the twinenvironment.They even directly translate network demands to apply the deeperintelligence and intent state insight into the network.Complex virtual ent

86、ities built through the fusion of different models requirecalibrations to realize and constantly update the accurate mapping between twinmodels and physical entities.3.Simulation,AI/MLAI/ML has developed rapidly over the past several years.The use of AI/ML for13simulation,training,prediction,and dec

87、ision-making of twin data and twin models issuperior to the direct application to physical entities.By setting different conditions,AI/ML even can test the parameters that cannot be set in reality,to avoid possibleerrors.AI automatically executes data preparation,analysis,and fusion for in-depthknow

88、ledge mining,to explain and predict the causes,processes,and results of realevents/incidents.This will generate various types of services,and greatly improve thedata value and response capabilities&accuracy of different services.AI/ML can givefull play to the role of the digital twin.In order to imp

89、rove the availability of thedigital twin,upon the quick and effective analysis of massive data,strategies orparameters optimized by AI/ML training and inference are fed back to twin physicalentities in real time.4.Interaction and securityBased on twin data,technologies such as 3D GIS,AR,VR,and even

90、XR can beused to reproduce the real world in the digital world,for digital virtual consistency inaspects of geometric dimensions,physical structures,and motion characteristics,visualization,as well as information mapping and feedback of physical entities anddigital models.The trust mechanism establi

91、shed through the blockchain can ensure the securityof service transactions to a certain extent.The blockchain makes twin datatamper-proof,trackable,and traceable while preventing errors and deviations due todata tampering,and improving the security of the digital twin.4.Digital Twin Solution for 6G4

92、.1.Digital Twin Network for 6G4.1.1.Lifecycle Twin NetworkWhen looking forward to the 6G service,we believe that it will extend toall-scenario on-demand services in all industries.The future differentiated demands indifferent vertical industries will show exponential growth,but the network bearerres

93、ources can only develop linearly,indicating a huge gap between demand and14supply.At the same time,the 6G will further integrate a wider range of cloud,edge,network,terminal,and fog resources.These resources now are fragmented anddiscrete,in lack of cross-domain management,control,and coordination,p

94、lus thepoor perception of lifecycle states of end-to-end resources,making it difficult to buildthenetworkbearercapacitiesfeaturedbyall-domainperceptionandquickcoordination.The wireless access network is characterized by a tremendous numberof base station cells,a large number of device models,wide di

95、stribution of stationsites,complex networking,and high energy consumption.It is the highest part of theCAPEX in mobile communication networks and the highest one in the OPEX.Currently,all links of the networks from planning,construction,operation,tooptimization,require massive manpower as well.With

96、the iterative development,themobile communication networks become more and more complex,the businessscenarios get more and more diversified,designs for user experience go more andmore profound,and the network O&M and optimization complexity showsexponential growth,to the extent that cannot be handle

97、d by human beings.The 6G network shall fulfill the demands that services are offered as expected,networks changed as needed,and resources shared as desired,realizing a good visionof 0 O&M,visualization,self-optimization,self-orchestration,and self-evolution inall scenarios in all domains.Figure 2 6G

98、 Network Autonomy Based on Digital Twin NetworkThe above visions challenge the architecture and capabilities of the 6G networkitself.On the data plane,to implement the real-time perception of network statuses,15the system can obtain relevant data in near real time.Through unified data models andstan

99、dard interfaces,supplemented by self-correction and self-generation capabilities,the system guarantees data quality.On the intelligent plane,the network enablesaccurate modeling and simulation verification,quick iterative optimization anddecision-making,and centralized or distributed intelligent gen

100、eration modes asrequired.The digital twin network is an important technical means and path to achieve the6G network visions and address challenges in networks and capabilities.Figure 3 Lifecycle Digital Twin NetworkDigital twin network technologies include functional modeling,NE modeling,network mod

101、eling,network simulation,parameter&performance models,automationtesting,data collection,big data processing,data analysis,AI/ML,failure prediction,and topology and router optimization.They convert issues difficult to solve at eachstage of the network into the digital world for solutions,enabling net

102、work autonomythrough monitoring,prediction,optimization,and simulation.The 6G network based on digital twin and AI technologies is an autonomousnetwork with self-optimization,self-evolution,and self-growth capabilities.Theself-optimizing network predicts the trend of future network statuses in advan

103、ce forearly intervention of possible performance gradation.It continuously identifies theoptimal status and verifies the simulation of physical networks in the digital domain,while issuing proper O&M operations in advance to automatically calibrate thephysical networks.The AI-based self-evolving net

104、work analyzes and makes decisionson evolution paths of network functions,including optimization&enhancement ofexiting network functions,and design,implementation,verification,and execution ofnew functions.The self-growing network identifies and predicts the demands ofdifferent services.Upon automati

105、c orchestration and deployment of network functions16in different domains,it generates end-to-end service flows that meet the servicerequirements.The network automatically expands the stations with insufficientcapacities while planning the areas not covered by the network,starting hardware,andloadin

106、g software automatically.As a new concept applied to the network field,the digital twin technologyrequires more consensus in the industry.This may take a long time in the industry andother sectors.At the same time,the digital twin technology relies on massive datacollection,which will increase the d

107、evice costs,and the data collection also needsbreakthrough innovations.The 5G network automation and intelligence is a kind of automatic analysis forspecific scenarios aiming to assist in manual decision-making.It is problem-orientedand independently implemented for specific domains.The autonomous a

108、pproach isbased on vendors private implementation.Its typical application scenario is thepatch-based SON standard.It solves simple problems with complex methods.Differentautomationmodesbetweendifferentsystemsmakeitdifficulttointerconnect,so human intervention is required.The 6G network autonomy requ

109、iressystems to automatically process all scenarios and achieve cross-domain autonomy.Itis necessary to lead the entire ecosystem to implement unified principles for networkautonomy in their development,deployment,and operation:Mobile networks are complex systems but members follow unified simplerule

110、sClosed-loop control based on perception,analysis,decision-making,andexecutionMobile network resource abstraction and schema abstractionFunction decoupling,managed by their respective orchestration systems bycategoryThe control ring forms a hierarchical architecture for autonomy of the entirenetwork

111、 from bottom to topReusablesoftwarecomponents,uniforminterfaces,forflexibleinserting/splicingNative supporting digital twin networkNative support online simulation testDesign concept from manual management to machine managementNative supporting self-evolution17NativeAIThe 6G network autonomy,on the

112、basis of traditional 5G NFVI and NFVMANO,forms the AI/ML layer,simulation test layer,twin network layer,and closed-loopcontrol layer,for hierarchical orchestration management,and supporting the buildingof a cross-domain hierarchical structure for network autonomy on this basis.Figure 4 Autonomous Ar

113、chitecture of Twin NetworkThe network autonomous architecture based on the digital twin contains thefollowing potential key technologies:1.Efficient and intelligent network measurement technology:Telemetry is aremote high-speed data collection technology from physical devices to virtual devicesand n

114、ow it has been widely applied in cloud computing,microservices,and someother fields.Devices actively transmit information about traffic statistics,CPU,ormemory data periodically to collectors at push mode.Compared with the traditionalpull mode(in which,the interaction is established through requests

115、 and replies),thenew push mode enables more real-time and higher-speed data collection.2.Unified data modeling for different network applications:Data modeling is toabstractly organize various types of data in the real world.The data modeling ofwireless networks relies on multiple different data mod

116、eling technologies.Wireless18access devices and network topology can obtain effective solutions in the digital twinspace only when a unified data model is established for them.3.Network visualization technology:The big data-based visualization appliesadvanced visual effects to the presentation of wi

117、reless access networks and deeplyintegrates with high-performance manipulations.This helps decision-makers todiscover rules behind data and improves their decision-making efficiency andcapabilities.4.Closed-loop network automation management and orchestration technology:Smart orchestration is a new

118、type of network brain to implement the unified controland allocation of network functions and resources.Its core technology is theclosed-loop control of network functions.5.High-performance AIOps technology,especially in prediction,cause analysis,exception detection,and intent translation:The AIOps

119、is essentially a series of O&Mfunctions implemented by applying the native AI,focusing on the O&M data analysis,including monitoring,log analysis,security,etc.The AIOps platform enables O&Mautomation,O&M improvement,and continuous insight into business performance.Operations that may take several ho

120、urs in the past now can be completed in a fewseconds with higher accuracy on theAIOps platform.Network automation technologies applicable to service-based and virtualizednetworks:including service registration,service discovery,lifecycle management,andsome other technologies,as well as the automatio

121、n framework under cloud-basednative services/microservices,such as Service Mesh,FaaS/BaaS,Serverless,etc.4.1.2.Knowledge-driven Twin Network ControlTo build a 6G all-scenario all-domain network intelligent control architecture,technical breakthroughs must be made in the following three key technolog

122、ies fromthe analysis and research of the current industry:(1)Service custom:The 6G network independently explores the implicitrelationship model of all-domain resource behaviors,such as services,users,computing,storage,connections,and data,while effectively and autonomouslydiscovering and automatica

123、lly implementing the user,data,and network customservice issues.19(2)Smart self-networking:The 6G network autonomously infers and generatesprogrammable deployment paradigms for on-demand services and efficient resourcesharing in all scenarios.(3)Network self-control:The self-planning,self-configurat

124、ion,self-assessment,self-healing,self-evolution,and some other service issues for control and allocation ofthe 6G network can be implemented in on-demand,agile,and closed-loop ways incomplex scenarios where human beings and machines,virtuality and reality,andtwins coexist.The 6G all-scenario service

125、s are based on the wireless network autonomouscontrol engine driven by all-domain knowledge graphs.In this control system,thebusiness QoE evaluation system is established based on user perception models(human vital sign model,as well as human visual,auditory,tactile sense,gestation,andemotionmodels)

126、specificfordifferentapplicationscenarios.Businesscharacteristics are extracted and classified,including interaction characteristics,popularity,latency,throughput,and packet loss rate requirements of users businesses,terminal mobility,object location,3D model characteristics of objects,user behaviorm

127、ode,etc.The network knowledge graphs are generated based on machine learning,includingslicingknowledgegraphs,energyconsumptionknowledgegraphs,automatic data labeling knowledge graphs,communication semantic knowledgegraphs,wireless environment knowledge graphs,etc.The ML-based user behaviormode and e

128、xception prediction help to form user knowledge graphs,providing userswith the perception QoE module,orchestrating network behavior predictions,offeringnetwork deployment policy generation module,and implementing the verification andoptimization of network deployment schemes based on the twin enviro

129、nment.Byreproducing the above knowledge graphs,a twin environment of the native network isbuilt to support the quick iteration,development,and testing of network deploymentschemes,for the purpose of all-domain,intelligent,and real-time network control.20Figure 5 Autonomous ControlArchitecture and Te

130、chnologies of6G Network Based onDigital Twin1.On-demand service control system of the 6G network based on the digitaltwinThe 6G control system mainly aims to implement the all-scenario,hierarchicalmodel,lifecycle,programmable,and intelligent network control.Multi-layer modelsfor control include end-

131、to-end network,single-domain network(radio access network,transmission network,and core network),and various wireless devices.The lifecycleof the control covers the planning,construction,maintenance,and optimization of 6Gwirelessnetworks.Thereprogrammablefeatureistheobjectprogrammingmanagement of al

132、l wirelessdomains and is the basis for automation andintelligentization.Intelligentization is to implement intelligent control with theknowledge intelligent agent(knowledge graph and reasoning engine)and networkcontrol agent(reprogrammable paradigm and control engine)by following theclosed-loop conc

133、ept of monitoring,analysis,reasoning,and execution.The core of the control engine is model-driven control automation.The controlmodel provides the control configuration through the reprogrammable paradigm.Thecontrol engine automatically executes control actions of each layer based onreprogrammable p

134、aradigm configurations.In addition,the objects managed by themodel and engine are expandable.The reprogrammable paradigm refers to the digital modeling for constructing the21all-domain network management model,including the end-to-end model and domainmodel(wireless network,core network,and transmiss

135、ion network).The digitalizedmodeling information,exchanged between the network intelligent agent and thenetwork control agent,must be recognized and executed by both of them.Thereprogrammable paradigm modeling includes the information model and data model.The information model represents the data be

136、havior logic of the model,and the datamodel represents the specific language for the information model.Thecontrolengineisresponsibleforparsingthedatamodelofthereprogrammable paradigm,translating the control actions in the wireless domain forthe model,and automatically executing the wireless control

137、actions for each layer,including being driven by the knowledge intelligent agent and coordinating with thecorresponding network management.In addition,the network management is layered,including the all-domain management,wireless network management,transmissionnetwork management,and core network man

138、agement.The wireless closed-loop control covers end-to-end network control in aclosed-loop manner.Because the all-network devices are distributed,the loop closingprocess can be classified into device-level intelligent loop closing,single-domainintelligent loop closing,and all-domain intelligent loop

139、 closing based on the real-timeand autonomy of intelligent data.The device-level intelligence is embedded andintegrated into a single device.From the perspective of the control system,thesingle-domain-level and all-domain-level control should be focused on.The knowledge intelligent agent and network

140、 slicing control agent are introducedto all domains for all-network intelligent loop closing.Through all-domain knowledgeintelligent agents,actions including modeling,analysis,training,and reasoning needto be implemented for all-network data.Then,the knowledge intelligent agentprovides the execution

141、 solution to the network control agent,and the network controlagent comprehensively makes decisions and implements execution.Ineachsubdomain,thesingle-domainknowledgeintelligentagentandsingle-domain network control agent are introduced for single-domain intelligent loop22closing for the wireless net

142、work,transmission network,and core network.Theknowledge intelligent agent implements analysis and reasoning for the single-domaindata,predicts the trend for the service and resource KPIs,analyzes the root cause,andoffers the candidate reprogrammable paradigm for single-domain network controlagents t

143、o decide and execute,implementing single-domain intelligent loop closing.2.6G Network Knowledge GraphIt requires the complete characteristics of all objects and elements to construct theknowledge graph of the space-air-ground integrated wireless network for 6G usersensing.The wireless knowledge grap

144、h can be described in a hierarchical structure.The knowledge includes sub-graphs which are wireless environment knowledge,wireless NE knowledge,and user knowledge.The sub-graph can be furtherdecomposed into characteristic subsets containing relevant characteristics.In the knowledge-graph-driven 6G a

145、utonomous control structure,based on therequirements of the NE intelligentization use case,it is considered to create profilesfor base stations and UEs,to describe the attributes,statuses,and behavioralcharacteristics of UEs and each service object in the base station.The profile iscreated based on

146、objectification modeling.In the base station,each service object has its own attribute,status,andbehavioral characteristics,which are changing constantly with time.To describe allinformation of an object at a time point,the service at this time point needs to besliced.The slice should contain values

147、 of all attributes,status,and behavioralcharacteristics of the service object.The NE knowledge graph is the general name for all service object profiles,including the attributes,status,and behavioral characteristics under differentperspectives.The relationship between service objects cannot be compl

148、etely andaccurately presented through a sole perspective,and thus the model is constructedbased on different perspectives in service relationships.The UE profile describes the attributes,statuses,and behavioral characteristics ofterminals.Since the UE profile involves multiple protocol layers and mu

149、ltiple types of23services,the profile for one UE needs to be disassembled due to the difference in thechanging frequency of its attributes,status,and behavioral characteristics.Figure 6 Strategy Learning and Reasoning System Driven by the Wireless Knowledge GraphFigure 6 shows that the preceding gen

150、erated wireless network knowledge graphcan provide knowledge graph model data for training the twin network.The twinnetwork performance knowledge obtained through training can be used to generatethe network allocating policy template through machine learning approaches(such asdeep learning).Through

151、the wireless network knowledge graph,the network changesare identified or predicted,and the policy template applicable to the network changesis derived.The generated network allocating solution is sent to the wireless networkcontrol engine for decision making and execution.The twin network needs to

152、offer the following network performance knowledgein order to assess the user requirement satisfaction and experiences1)Network KPI:The latency,throughput,connection success rate,coverage,packet loss rate,handover success rate,delay jitter,and so on for each network unit;2)Knowledge related to user s

153、ensing:QoS/QoE-related parameters,users fivesenses,emotional model parameters,and so on;3)User/object status knowledge:Human vital sign model parameters,3D objectmodel parameters,and so on.Various algorithms of the original network need to be copied in order to verify the24network allocating solutio

154、n.The algorithms include the following:algorithms andstrategies on each protocol layer of the core network,transmission network,andaccess network,the routing and handover algorithm strategies of the space-air-groundintegrated networks,federated self-learning cross-domain alliance control algorithmst

155、rategies,resource-spectrum-driven wireless network autonomous control algorithmstrategies,intelligent edge buffer technology,smart avoidance control algorithm fordeterministic communication faults,and so on.3.Knowledge-driven digital twin network controlThe decision making based on the knowledge-dri

156、ven digital twin network has thefollowing advantages:Virtual modeling based on knowledge and scenario information:Theontology base is used to unify the multi-source heterogeneous data;Theontology-based modeling can better map the actual physical world to thevirtual model;The requirement input and re

157、asoning provided by theknowledge graph are applied in addition to the ontology-based model,togenerate more accurate virtual model corresponding to the real physicalnetwork.Smart strategy template generation:The experience and knowledgevisualized based on the knowledge graph can be used in networkmai

158、ntenance measures such as fault diagnosis for the initially establishedvirtual model,to import the knowledge graph for decision making and obtainthe decision making results;The decision making process is a cyclic process.That is,the formed decision is applied in the virtual model for repeatedexecuti

159、on,to obtain the network maintenance measures decisions,and thestrategy template is repeatedly updated based on the strategy performance.The updated strategy template is saved in the knowledge graph.The template,which is called again when the same scenario and service occur on thenetwork,no longer n

160、eeds to be generated repeatedly.Precise on-demand management of network resources:In network25resource management,knowledge can be used for template selection as wellas a supplement for background or user preference used for network decisionmaking,to improve the machine learning algorithm performanc

161、e andimplement the accurate on-demand allocation of resources.To implement personalized resource allocation,proper estimation of servicesmust be performed for each user,so that the service resources can be accuratelyreserved for each user.This requires not only the collection of the current status o

162、fusers but also the utilization of user history behavior knowledge to accurately predictusers intent and allocate network resources.Figure 7 Knowledge-Driven DT Service Classification FrameworkTo achieve this goal,knowledge-driven 6G technology represented by theknowledgegraphcanofferuserbehaviorrea

163、soningandexploreinter-userassociations.As shown in Figure 7,a single user can be regarded as a node in themulti-dimensional knowledge graph to analyze the association between different usersin different requirement dimensions.Alternatively,the user twin can be regarded as aknowledge graph to explore

164、 the association between the users own attributes toimplement the attribute-based knowledge reasoning.The knowledge-driven digital twin resource management also has the followingdistinct characteristics:Knowledge attribute associations:The knowledge graph can be used toreason out new attributes of t

165、he current user based on the attribute or information of26its associated node.In the reasoning process,the knowledge graphs map each twin toa node in the knowledge graph.In this case,if each twin is a knowledge graph,thetwin is a data resource that can implement communication.When a user needs acert

166、ain type of data in a certain knowledge graph,the real network can transmit thedata sub-graph of related information in the graph to the requesting party.Thisenables the user to obtain the association between information while the user obtainsinformation.Highly efficient on-demand resource allocatio

167、n:Collecting user behavioralinformation under different statuses based on the digital twin technology helps createthe user preference and knowledge graph.Then different QoS requirements can beobtained through analysis,to implement on-demand resource allocation based on userpreference.Sensors are use

168、d to collect the environmental characteristics,and theenvironmental characteristics and user preferences are used to analyze the impact ofdifferentenvironmentsonuserrequirements,toaccuratelyobtaintheuserrequirements in different environments.In addition,knowledge,as effectiveinformationobtainedthrou

169、ghanalysisandsummarization,canserveasasupplementaryinputformachinelearningalgorithms,improvingalgorithmperformance such as efficiency and accuracy.4.1.3.Intent-Driven Twin Network ManagementVarious new services and applications emerge,with the rapid development of bigdata,cloud computing,Internet of

170、 Things and other technologies.Therefore,thenetwork operation and maintenance become more complex.In addition,it isincreasingly essential to guarantee 6Gs requirements for embedded security.In thisscenario,it becomes an important research of network to ensure highly reliableoperation of network serv

171、ices and create low-cost self-healing of network failures.DTnetworks are composed by both the real physical network and virtual twin network,and they are mapped to each other.The virtual twin network has the same27characteristics,information and attributes as the real physical network.On the onehand

172、,the DT network enables real-time modeling of the communication network andhelps uniformly manage the operating model and data of physical network elements.In this way,the DT network can promote dynamic operation and maintenance andsmart decision-making for the network.On the other hand,the DT netwo

173、rk canprovide real-time digital display interfaces to the network administrator,in order toimplement end-to-end network monitoring and network policy verification.Intent-driven network(IDN)is a new network management form,aiming atnetwork autonomy.IDN combines artificial intelligence,machine learnin

174、g,andnetwork orchestration technology,so as to apply the deeper intelligence and intentstatus insight into the network.To be specific,IDN is designed to reduce thecomplications in creating,managing,and implementing network strategies,andreduce the manual operations related to traditional configurati

175、on management.This isconsistent with the vision that the 6G network intelligently discovers usersrequirements and provides on-demand services to users.In IDN,the networkadministrators declare the expected results or goals to describe the intent.And thenetwork software determines how to implement the

176、 goal through artificial intelligenceand machine learning.Therefore,IDN can automatically execute policy and providereal-time visualization of the network operations to verify the intent.In addition,predicting the potential deviation to the intent and formulating the correction policycan ensure the

177、effective execution of the intent and realize the autonomous monitoringand correction of the network.From the perspective of network communication,IDN initially realizes thedigitalization of user intent.Intent-driven DT network can directly translate thenetwork requirements,collect the network statu

178、s in real time,and dynamicallyoptimize the network policy.The benefits to 6G networks are as follows.Reducedmanualparticipation:Intent-drivenDTnetworkautomaticallyconverts intent to configurations,and network administrators no longer need tomanually configure the network.The network translates the i

179、ntent and verifies its28validity,and then the options are reported to the network administrator to changethe configuration.Quick troubleshooting:Intent-driven DT network continuously monitors thenetwork status and instantly discover network running problems.Machinelearning can be used to verify the

180、feasible solutions in the twin network,anddetermine the optimal solution,and send the solution to the physical NE forexecution.Improved network security:Generally,Intent-driven DT network proactivelydetects threats during the monitoring,especially for encrypted communications.Once a security vulnera

181、bility is detected,the network can instantly identify thevulnerability and take control measures.In other words,IDN can save massivetime for planning,tests,verification,and manual configuration.Optimized strategy analysis:Intent-driven DT network continuously collectsthe network operation and mainte

182、nance data.It then analyzes the data usingmultiple methods to provide valuable information related to network performance,security threats,and so on.After fully understanding the network operation status,the network administrator can make better decisions to generate the optimalnetwork configuration

183、s.1.Intent-driven twin network architectureThe DT network is an effective way to implement lifecycle management ofintent.It integrates the twin network and IDN,dynamically generates optimalsolutions that can be executed,and produces good results based on the real-timenetwork status.Figure 8shows the

184、 intent-driven twin network management and control,includingthree layers:Physicalnetworklayer:Thephysicalnetworklayerincludesvariousinfrastructures of the 6G network.Technologies such as in-band networktelemetry can be used to collect the physical network element operation status.Moreover,the networ

185、k data and network control information can be interacted29with the digital twin entities through the southbound interface of the twinnetwork.Twin network layer:The twin network layer is the sign of the DT network,including three key subsystems of the database,service mapping model,anddigital twin ma

186、nagement.The database collects and stores various types ofnetwork data,including network operating status and history configuration dataof the physical network layer.Meanwhile,the data service,unified interface anddata support are provided to service mapping model and digital twin management.The ser

187、vice mapping model executes key operations such as schedulingoptimization,fault diagnosis,traffic analysis,topology model,and simulationverifications,and then the network strategy is verified in the twin network andexecuted on the physical network layer.The digital twin management completesthe inten

188、t translation,configuration verification,automatic troubleshooting,andintent assurance.The users intent is translated to effective and feasible networkstrategy,and automatically completes troubleshooting.Network application layer:The network application layer provides openinterfaces to users,and use

189、rs can directly manage the network.Users can enterthe intent in natural languages through text,audio,or graphical interfaces,without mastering the professional knowledge and expertise of networkmanagement.30Figure 8 Intent-Driven DT Network ArchitectureFrom a technological standpoint,the physical ne

190、twork forms massive networkconfigurations after user intent is translated.If the configurations are sent directly tothe physical network,they may affect other services,resulting in unanticipatedconsequences.The DT network service mapping mode validates and simulatesconfiguration delivery in advance,

191、allowing for the detection of configurationanomalies.Numerousservicemappingmodelsmayimplementconfigurationverification,intent assurance,and automatic troubleshooting in intent-driven twinnetwork control systems to verify user intent from the network application layer inreal-time.Additionally,theinte

192、ntassuranceandautomatedtroubleshootingcapabilities based on the service mapping model convey the physical networksworking status to the twin network layers database via data collection.The servicemapping model continually checks users.Suppose it determines that the network isdeviating from the servi

193、ce intent.In that case,the DT network can utilize intelligenttechnologies such as AI to do root cause analysis and produce a troubleshootingapproach.The DT networks service mapping model validates the troubleshootingstrategy in advance to guarantee its accuracy and then delivers the strategy to thep

194、hysical network via an automated configuration module.2.Key procedure of intent-driven twin network31The DT network in the communication domain can accurately show the networkoperating status in real time.The real-time interaction between the twin network andthe physical network can predictably main

195、tain and optimize network strategy.Figure9 demonstrates the major steps required in intent-driven twin network control loop.From a physical 6G network to a virtual 6G network,the procedure traversesnetworkintent,configurationdata,real-timestatus,dataaggregation,andperformance design to establish a v

196、irtual entity relationship in the DT network.Thevirtual 6G network can validate and optimize the strategy in advance to ensureefficient network deployment and minimize the impact on the real network,therebyboosting the capability for intelligent network simplification.Figure 9 Key Procedure of the I

197、ntent-Driven Twin NetworkThe intent-driven twin network may be used to ensure the end-to-endperformance of a network.The DT network technology is capable of simulating thenetwork operation state and the implementation impact of network strategies toenhance network lifecycle management performance an

198、d assure the correctness ofend-to-end closed-loop management.Assume the system identifies a decline in theend-to-end service performance.In this situation,the DT network may do intelligentanalysis using real-time data monitoring and historical operation and maintenancedata to precisely detect the ne

199、twork issue and give a matching network recoverysolution to the physical network following the twin networks verification.324.1.4.Terminals data-related supporting technologiesData collection is essential for constructing 6G digital twin networks.As one ofthe main ways of data collection,the termina

200、ls first sense and collect data and thenextract,process,and transmit data to meet the needs of the digital twin networks.Thedata-related supporting technologies of the terminals to 6G digital twin networks canbe listed as follows:1)Comprehensive and effective data acquisition.Comprehensive coverage

201、of the acquired data is required to meet the needs of thedigital twin network.With the development of information technology,the data can beobtained through a variety of terminals,including mobile phones,wearable devices,vehicles,IoT sensors,and so on.More and more data can be collected and stored i

202、nreal time.At the same time,it is necessary to deal with the tradeoff betweencomprehensiveness and effectiveness of the data.Irrespective of the validity of thedata,it may result in a large amount of data obtained by the terminals,such asirrelevant data,abnormal data,and redundant data.Targeted data

203、 acquisition wouldincrease the effectiveness of the data,thereby conducive to reducing the burden on theterminal.2)Data mining and processing.Even if the data volume can be reduced owing to targeted data acquisition,thedata acquired by the terminal is still massive for transmission.It is necessary t

204、oconduct data mining,knowledge extraction and generalization.Partial data can beprocessed within the terminals.3)Iteration and optimization of data.Based on the acquired new data along with other stored data，it can meet true timeoptimization of data.Various kinds of data require different iteration

205、periods.To fullyrealize an immersive remote experience,the acquired data require real-timeperformance.Iterative optimization for different data may drop the outdated data andinvalid ones.It can also update twin model parameters,and improve 6G services33adaptability to the practical wireless environm

206、ent.The updated data and previous datacan rectify each other，in case some data is missing or changes dramatically.Thus，itcan ensure the accuracy,consistency and comprehensiveness of the data.Besides,itcan reduce the burden of data storage and processing on the terminals.4)Exchange and integration of

207、 data.In a 6G wireless network,multiple types of terminals with different statuses aredistributed at various locations.The data from multiple terminals can be exchangedand integrated.Mutual complementation and enhancement would be performed tobuild an overall model based on valid data acquired.Besid

208、es,through data fusion,data from different sources can be used for one specific terminal in user-centricnetworks.4.2.OtherApplications of the Digital Twin4.2.1.Smart TransportationThe digital twin can provide new technological support and developmentdirection for the intelligent transportation syste

209、m(ITS).The physical objects in theITS are mirrored,and the digital twin can implement full data sensing,real-timeinformation sharing,and accurate collaborative decision making,to promote theoriginal ITS into a revolutionary transformation and upgrade toward a scientific,accurate,and ecological integ

210、rated traffic management system.As shown in Figure 10,the DT-enabled ITS has the following seven typicalapplication scenarios.34Figure 10 Typical applications of DT-enabled ITS1.Intelligent traffic controlIn the ITS,there are significant problems of the high cost of interaction betweenvehicles and t

211、he MEC,long latency for decision making on the cloud server,and poorexecution of scheduling commands by vehicles.In the DT-enabled ITS,a twin of thephysical traffic is created in the virtual world to create a high-definition mirror of theITS.Through road network layout,infrastructures,twin data of t

212、he vehicular users,the cloud server can implement simulation optimization,properly arrange publicvehicle scale,and implement traffic guidance based on the vehicles preferences.Onthe edge layer,rapid interaction between the vehicular users digital twin and the MECcan avoid frequent information transm

213、ission between the vehicle entity and the MEC.In the meantime,the driving path of higher efficiency is planned in advance based onthe users personalized driving requirements.This improves transportation efficiencyand provides vehicular users with high-quality driving experiences.2.Collaborative unma

214、nned driving technologyCollaborative driving for autonomous vehicles(AVs)can significantly increasethe capacity for the intelligence bottleneck of individual AVs.However,AVs withdifferent sensing,calculation,and communication capabilities need to frequentlyshare information and decisions to determin

215、e the collaboration group scale and work35distribution of different AVs.In the DT-enabled ITS,the resource status of AVs andthe users requirements and preferences are synchronized to the digital twin connectedwith the MEC.Therefore,the digital twins of AVs in the virtual world compose anall-new virt

216、ual interaction network in the MEC,where the DTs can implementinformation communication and formulate strategies for the AVs.This saves thecalculation and transmission resources and implements the smart group collaborativedriving of theAVs.3.Heterogeneous resource allocationIn the future,the ITS app

217、lication scenarios are broadened,the performanceindicators are diversified,and the service needs are more detailed.However,there areinsufficient considerations of the device difference and service personality for thecurrent resource allocation solution,which cannot provide high-quality services forv

218、ehicles in different application scenarios.In the DT-enabled ITS,the resource statusof the physical entity is synchronized to its digital twin in real time,and the cloudserver can flexibly implement integration and scheduling for the digital twin,toconduct transportation resource management in small

219、 granularity.On the basis ofreal-time monitoring and configuration of the resource status,the cloud server andMEC can customize the knowledge-based resource slice through big data analysis andartificial intelligence,to satisfy the personalized resource need of diversified services,significantly impr

220、oving the service experience quality for users.4.Infrastructure maintenanceThe ITS,with incomplete sensing ability,low accuracy,and insufficientintelligence,can hardly implement highly accurate and all-around comprehensivemonitoring and predictive maintenance for vehicles and infrastructures.In theD

221、T-enabled ITS,the sensors deployed in the vehicles and on the road andinfrastructures can implement status monitoring and regular update to the twins on thecloud.The cloud server comprehensively considers the infrastructure status andenvironmental factors based on the information provided by the dig

222、ital twins toformulate optimal decisions.In addition,the digital twin will save the history status36information and create the all-element lifecycle digital profile and the predictivemanagement and maintenance system for the vehicles,roads,and infrastructures.Thisimproves driving safety as well as r

223、educes the maintenance overhead for vehicles,roads,and infrastructures.5.Road emergency rescueFaced with the explosive growth of the traffic and driving needs of different users,the ITS can hardly manage the execution of scheduling commands by normal vehicleswhen planning the optimal path for emerge

224、ncy vehicles as the green passage.In theDT-enabled ITS,the cloud server can assign different rewards for different trafficflow statuses and comprehensively consider the driving time,rewards,personalizeddriving needs of the digital twins of vehicles to plan the optimal driving path,enabling users wit

225、h different driving needs to comply with the scheduling commands.On this basis,the cloud server and MEC can quickly adjust the traffic lights and theemergency incidents corresponding to the road rewards,to significantly shorten thetime for the emergency vehicle to pass and reduce the loss of the eme

226、rgency incidentand the impact on normal vehicles.6.Digital asset managementDigital assets of great utilization value will be generated after the collection,integration,analysis,and deep learning are implemented on traffic data.In theDT-enabled ITS,the twins can replace the physical entities to compl

227、ete theclassification,screening,storage,authorization,and transaction,and create theall-element lifecycle digital profile.In this way,a new model for commercializing theITS data assets can be created for the digital account of the twin.In addition,theadditional benefits also encourage physical entit

228、ies in the ITS to proactivelyparticipate in the data sensing and maintenance.This is beneficial for the deploymentof various applications and services in the ITS,promoting the construction of theecological industrial chain for the ITS industry.7.Test solution verificationThe traditional ITS has high

229、 cost and redundant cycles,making it difficult to test37the algorithms,solutions,and architectures on large scales.In the physical world,massive intelligent sensors dynamically map the sensed data into the twins in the twinworld in real time,creating a virtual world test platform for the deduction a

230、nd test ofnew algorithms,new solutions,and new architectures.The twin of the data requestercan negotiate and make decisions with the digital twins of the data owners in thevirtual world,to develop the transaction plans for data.In the meantime,tests of newalgorithms,new solutions,and new architectur

231、es can be implemented repeatedly onthe digital twins in the virtual world for prompt corrections.In the end,newalgorithms,new solutions,and new architectures can be deployed on the applicationsof the physical entities while the test cost and cycle are both reduced.DT-enabled Intelligent Traffic Cont

232、rolTrafficjamswillincreasethetimefordriving,energyconsumption,environmentalpollution,andaccidentrate,andwilldirectlyhamperurbandevelopment.In the existing intelligent traffic control system,the information can beshared among different physical entities through the vehicular networks to make thetraff

233、ic smart.Specifically,as shown in Figure 11,IoTs devices such as cameras,magnetic sensors,and millimeter-wave radars are deployed along the road to sensetraffic data in real time.The BS transmits collected traffic data to the cloud server.The cloud server comprehensively analyzes the historical data

234、 and the data collectedby the IoT devices to predict the traffic status and determine the traffic schedulingdecisions that are sent to the vehicle users through the BS.During the traffic flowcontrol,if a vehicle user needs to plan the optimal path,the user can send a pathplanning request to the conn

235、ected BS.After the BS forwards the users request to thecloud server,the cloud server conducts analysis based on the driving needs and trafficflow and determines the optimal driving path accordingly for the vehicle user,toimprove the driving experience for the vehicle user.38Figure 11 Intelligent Tra

236、fficArchitectureThough the traffic system and vehicle users,in theory,can benefit from thecurrent scheduling architecture,there are a series of new challenges in the applicationof the system.High cost:In the ITS,the test cost is high for new algorithms,new solutions,andnew architectures of traffic f

237、low scheduling and control,and consequentlyfrequent massive tests in the real world are impossible.Long latency:The number of networked vehicles drastically increases,causing along latency to determine and distribute the scheduling decisions for planningpaths for massive users.Thus,the traffic opera

238、tion system can hardly reach theexpected efficiency.Poor execution:The existing ITS is a path recommendation system,and canhardly ensure that all vehicle users will follow the control scheduling commandssent from the cloud.As a result,the vehicle traffic cannot be effectively managed.Poor experience

239、:When the cloud server is planning traffic flow solutions,theprimary reference indicators are usually the shortest distance or shortest drivingtime,and the server does not consider personal needs.Therefore,it cannotprovide high-quality driving experiences for vehicle users.To deal with the preceding

240、 challenges,it is urgent to create ecological,39digitalized,intelligent,and controllable ITS,improving the comprehensive command,rapid response,and scheduling and control capability for road traffic management andeffectively supporting green,intelligent,and immersive traffic.The new paradigm of the

241、DT-enabled intelligent traffic flow control combiningthe digital twin technology and the current traffic flow control architecture is theoptimal idea to deal with the preceding challenges.Specifically:1)The digital twincan implement real-time traffic data collection,accurate update of traffic entity

242、 status,and synchronized display of traffic operation.Thus,it provides a test platform for thededuction and tests of new algorithms,new solutions,and new architectures;2)Theinteraction between the digital twins of vehicles in the virtual world and the cloudserver can help plan driving paths for vehi

243、cle users in advance,effectively reducingthe path planning latency;3)In the DT-enabled traffic control system,road segmentscan be used as a resource to motivate users.The cloud server can give differentrewards to different segments based on the vehicle density.In addition,the drivingstatus of indivi

244、dual vehicles will be synchronized to the twins in real time,andtherefore the cloud server can get to know the scheduling command execution byvehicles and adjust the scheduling decisions promptly;4)The digital twins can reflectthe personalized driving preferences of the vehicles,and provide the opti

245、mal pathbased on the personalized needs of the driving time and driving reward during thecontrol and scheduling,to improve the service experience of vehicle users.1.Virtual-real mapping of the traffic control systemThe virtual traffic DT platform is primarily composed of the digital twin of thetraff

246、ic environment,digital twins of the vehicles,and cloud server.The digital twin ofthe environment is divided into different areas based on the base station coverage,anda digital twin is deployed to represent the real traffic environment in each area.Withhighly precise maps as basic data,the digital t

247、win can adopt different types ofreal-environment data and 3D rebuild technology,to reconstruct the traffic scenariosin the physical world with high precision.In addition,IoT devices deployed along theroad,such as cameras,laser radars,and geomagnetic sensors,and multiple types of40sensors installed o

248、n the vehicles can collect traffic data in different road segments anddynamically sense the traffic environment,and the data is efficiently transmitted ondemand through the heterogeneous networks,to realize the data interconnectionbetween the digital twin world and the real physical world.The digita

249、l twins ofvehicles store the driving needs and preferences of vehicles.When the driving needsor preferences need to be updated for the vehicle,the on-board unit(OBU)forcommunication can be used to update the information via the BS to the digital twin onthe cloud.Based on the current traffic environm

250、ent and the driving needs andpreferences of individual digital twins,the cloud server develops the optimal trafficcontrol solution,dynamic multiplexing solution of road resources,and importantvehicle path planning solutions.2.DT-enabled traffic scheduling solutionDifferent from the traditional traff

251、ic scheduling idea of using the shortestdistance or shortest driving time as primary indicators,the DT-enabled intelligenttraffic scheduling further considers the personalized needs of vehicle users and usesthe road as resources and the price as the motivation method to guide vehicle users tocomply

252、with the control solution.Firstly,from the perspective of the traffic flowcondition,the IoTs devices deployed along the road and sensors on vehicles in the realtraffic will collect the original traffic flow data in real time.The collected data will bepreliminary processed by the MEC,and the result w

253、ill be uploaded to the cloud server.The cloud server can specify proper rewards based on the traffic flow of each roadsegment to motivate vehicle users to choose different road segments for driving.Forexample,the cloud server can specify negative rewards in congested road segments tocharge fees from

254、 vehicle users.Similarly,the cloud server can specify positiverewards in road segments with few vehicles.Secondly,from the perspective of vehicleusers,different vehicle users have different driving habits and different preferencesregarding driving time and driving reward.Therefore,time and reward ca

255、n be used asbasic elements to establish the personalized model for individual vehicle users.Forexample,vehicle users who are sensitive to driving rewards are more willing to41choose longer road segments and spend more time to get more rewards.Accordingly,vehicle users needs and preferences are accur

256、ately mapped by the digital twinsdeployed on the cloud,thus promoting the cloud server to configure optimal rewardsfor different segments by comprehensively considering the current traffic status andusers driving needs and preferences.The digital twin architecture can implementclosed-loop management

257、 in collecting the traffic environment and users needs andpreferences in real time,specifying road segment rewards,scheduling traffic flows,selecting paths for vehicle users,and updating the traffic status.Figure 12 DT-enabled Intelligent Traffic ControlIn the closed-loop scheduling process,the digi

258、tal twin maps the needs andpreferences of the vehicles.As shown in Figure 12,if there are travel needs for avehicle,its digital twin sends a driving path planning request to the cloud server,andthe cloud server adds the request into the service list based on its start time.Then,based on the scheduli

259、ng time interval provided by the digital twin,the cloud servercomprehensively considers the traffic flow status and reward,as well as the vehiclesstart point,destination,needs,and preferences and uses optimization algorithms suchas heuristic algorithms,game theory,and artificial intelligence to plan

260、 the properdriving path for each vehicle user.If the vehicle does not reach the destination after around of path planning,the cloud server will add the service request of the digital twininto the service list again and wait for the next scheduling.After a round of path42planning for vehicle users in

261、 the virtual world,the cloud server will send thescheduling command and path recommendations to the vehicle user via the BS,toimplement the DT-enabled traffic flow control.4.2.2.IoTThe vision of 6G is three dimensions and generalization,and intelligentconnection for the world.With the rapid developm

262、ent of information technology,6Gwill enable a series of intelligent IoT applications that requires ultra-low latency andultra-high reliability,such as the future intelligent transportation system and smart city19.Communication and computing are both indispensable to promote the Internet ofEverything

263、.In the IoT scenario,as shown in 13,MEC and communication are hottopics in recent years 20.It features decentralization,low-latency computing,datasecurity and privacy,and joint optimization for computing and communication.However,it is also faced with the challenges of general computing,task plannin

264、g anddistribution,distributed storage and collaborative computing,and public use andsecurity of edge nodes.Figure 13 MEC and Communication under IoT Scenarios43For the Internet of Everything,how to conveniently manage and control massivenodes is also a hot topic in addition to the construction of mo

265、dels.The commoncontrol method following the node-gateway-cloud platform-control-end patterncannot be used for intelligent operations,and its scalability and network robustness donot meet the requirements of new services.However,introducing digital twins willeffectively resolve the control problem.On

266、e of the major functions of the DT-driven Internet of Everything controlsystem model is to simulate,monitor,diagnose,predict,and control the formationprocess and behavior of physical or simulated nodes in actual environments.Simulation:Before the deployment of heterogeneous physical network nodes,th

267、ecommunication and computing processes are simulated in a virtual simulationenvironment(for example,NS3 or other network simulation platforms)to learn nodestatus in the actual operating environment as much as possible.Behavioral policies,theprobabilityofservicesuccess,parametersetting,andissuesnotco

268、nsidered/expected in the design stage will provide the foundation for subsequentservice planning,service parameter determination,and decision making in abnormalsituations.Node operation in different service environments can be simulated bychanging the parameter setting in the virtual environment.The

269、 influence of differentservice parameters on the probability of service success can be simulated by changingtask parameters.Monitoring and diagnosis:In the node service process,service data will bereflected in the digital twins of the nodes in real time.The digital twin system of thenodes enables dy

270、namic,visual real-time monitoring of the actual node service processand allows fault diagnosis and positioning on actual nodes based on the real-timemonitoring data and historical data obtained.Prediction:The building of a node digital twin mode makes it possible toperform integrated simulations and

271、 validation of the testing process of node functionand service performance in digital space to predict potential service,functional,andperformance defects of the nodes.For those defects,the corresponding parameters can44be modified in the digital twins of the nodes.On that basis,simulation of the se

272、rvice,functional,and performance testing process of the nodes is repeated until theproblems are solved.Control:In the node service process,the process data of real-time service isanalyzed to control the status and behavioral policies of the physical nodes,includingservice policies,parameter changes,

273、etc.Figure 14 shows the architecture of the DT-driven Internet of Everything controlsystem,in which the IoT sub-net is the physical entity layer,the twin convergencenetwork is the digital twin layer,and the management and control center on top ofthem is the application layer.This networking architec

274、ture has four characteristics:First,edge intelligence is used and digital twin technology is introduced into thearchitecture.The twin network is at the center of the system architecture,enablingquick response,intelligent computing,and delivery driven by computing tasks.Second,there is a large-scale

275、hierarchical structure.Data information istransmitted from the IoT subnet to the twin network and then to the backhaul network.Finally,the data is organized by the management and control center.The clearhierarchical structure greatly boosts functional stability and scalability.Third,heterogeneous in

276、tensive networking.Multiple communication systemsandheterogeneousnetworksaresupported.Forintensive data,time-sharingscheduling is enabled so that the number of nodes and communication constraints areno longer bottlenecks for networks.Fourth,efficient,automatic networking.The architecture employs the

277、 automaticconfiguration technology for dynamic restructuring and reconstruction,enhancingnetwork robustness.45Figure 14 Architecture of the DT-Driven Internet of Everything Control SystemMeanwhile,the twin convergence layer at the core of the new architecture isexpected to deliver five functions:Fir

278、st,support for flexible access.The twin convergence layer has multiplecommunication systems,supports heterogeneous networks and intelligence operations,and is capable of automatic configuration of network access and strong accessscalability.Second,destruction-resistant restructuring and reconstructi

279、on.In the event ofnetwork outages due to signal quality or other uncertainties,the twin convergencelayer allows cut-vertex detection to automatically locate the outage point and performdynamic network restructuring and mobile node scheduling to maintain networksystem stability.Third,intelligent judg

280、ment.For multi-source data,comprehensive judgment isperformed so that algorithm is used to determine whether to report data situations andtake countermeasures.Fourth,an application layer gateway to facilitate intelligent service dataprocessing and node collaboration.Fifth,intelligent command executi

281、on.Commands on the control layer can be46executed with a single click and intelligent analysis of ambiguous commands can beconducted for intelligent command delivery and scheduling.Taking the architecture in Figure 14 as an example,after the introduction ofdigital twin technology,large-scale twin co

282、nvergence of physical and simulationnodes is realized in the architecture of the Internet of Everything control system.Thearchitecture is also capable of self-networking,self-configuration,and self-access andsupports multi-hop connections and various wireless communication networks,suchas WiFi,ZigBe

283、e,LoRa,etc.During actual application,the twin convergence network layer first simulates theservices used by the IoT node to be deployed.For example,a twin layer similar toNS3 software is used to build twins for ZigBee nodes and the services to be run by thenodes(e.g.,temperature and humidity measure

284、ment and analysis)are deployed tosimulate the communication and computing processes between the actual nodes andthe twin convergence nodes(ZigBee protocol communication or network mapping).Parameters in multiple dimensions are set to replicate the status of the actual nodeoperating environment.Data

285、produced by the twins is uploaded to the managementand control center via the backhaul network and then analyzed and maintained byadministrators to find and correct issues emerging in the simulation process and thoseexpected.After that,the physical nodes are deployed and correspond to the digital tw

286、ins inthe twin convergence network.With data information about the physical nodesuploaded via the backhaul network visible on the management and control center,administrators can monitor nodes in a dynamic,visual,and real-time manner.Theycan also diagnose and locate faults on the real-time monitorin

287、g data and historicaldata obtained.With the above modeling,the twin layer will become the key to truly simulatingand predicting the actual operation.Therefore,the integrated simulations andvalidation of the testing process of node function and service performance arecontinuedto predictpotentialnodes

288、ervicedefects.Forthose defects,the47corresponding parameters can be modified in the digital twins.On that basis,simulation of the service,functional,and performance testing process of the nodes isrepeated until the problems are solved.As a result,the main purpose of employing thedigital twin technol

289、ogy is achieved.In addition,while the digital twin application described above focuses more onreading node data,the adoption of digital twin technology,in fact,will be a hugehelp in writing in node control and management.For target nodes of active control,control commands can be delivered to physica

290、l nodes via the twin layer for remotemanagement and control.For example,a physical node can run various services.Atsome point of time when it needs to handover the service type,the control terminalcan deliver a control command to the twin layer for simulations.If the indicatorsconform to the origina

291、l setting,the twin layer will synchronize the services to thephysical node so that the node can run those services.The twin layer will also collectdata from the physical node for further analysis and improvement.The above content describes the example architecture of the DT-driven Internetof Everyth

292、ing control system.In fact,the combination of digital twin technology andthe Internet of Everything will enable a large number of applications.For example,itwill be easier for mobile operators to authenticate and manage devices accessing theirmobile networks,exercise prioritized control on devices i

293、nfluencing mobile networkstability or occupying too many resources,and ease the resource usage threshold onspecially approved devices with high requirements.The introduction of digital twinwill bring convenience to both users and managers.As a result,users canconvenientlyandperfectlydeployservices,w

294、hereasmanagerscanquicklyauthenticate node users and perform coordination and planning.4.2.3.UE TwinThe past few years have seen tremendous changes in the network andcommunication industry.The growth of mobile users has become almost saturated,only playing a minor role in driving operator performance

295、.However,with people48growing increasingly reliant on the mobile Internet,a wide range of mobile serviceshave appeared and numerous emerging applications,such as 360 panoramic videos,virtual reality(VR),augmented reality(AR),are being commercialized at speed.Thedemand for data computing quantity pos

296、ed by those services continues to increase.Agrowing number of applications in vertical industries also require lots of computingandprocessingworkloads.Internet-poweredreal-timecomputingwithhightransmission reliability is necessary for autonomous driving,industrial control,andother applications.In th

297、is sense,future mobile services will put more emphasis on theneeds of users and services,offering a brand new user experience.Mobile interactivegaming with huge data volume,3D,AR/VR,hologram images,and other new mobileservice applications have been incorporated into the technical requirements on fut

298、uremobile communication systems.The network requirements of new services will become more differentiated,diverse,and highly dynamic,which means that the network in the future will be ahighly dynamic complex network compatible with various types of user requirements.This network will be a completely

299、heterogeneous,large-scale network able to supportthe huge connectivity necessary for connecting everything in the Internet of Thingsand the ubiquitous wireless big data applications.However,the UEs carrying those emerging services are not large-scalecomputers and therefore are not capable of process

300、ing a large volume of data.At thesame time,the traditional communication network alone cannot meet the bandwidthand latency requirements posed by the large data volume.Against this backdrop,theintroduction of computing,including popular research topics in recent years,such asedge computing,service c

301、ache,and network operator,can address these problemsbetter.Digital twins can ease the pressure on UEs due to emerging services.If appliedin such scenarios,it will substantially improve the situation faced by future mobileservices.Taking the example of C-RAN architecture19,this section discusses thea

302、pplication values of digital twins in addressing the above problems.49As noted,the technical framework of digital twin is an architecture containingthree layers of abstraction:physical entity layer,digital twin layer,and applicationlayer.Mobile phones and other UEs are on the physical entity layer.T

303、he applicationlayer will be faced with insufficient computing power on the entity layer whendeploying actual applications.This problem can be better addressed if the architectureis driven by a digital twin.As shown in Figure 15,after the introduction of digitaltwin,a digital twin layer will be put a

304、bove UEsa twin that may be deployed on themobile edge cloud.That is,each UE accessing the network will have a twin on theedge cloud.Data on physical UEs and their twins will affect each other.That means afaulty physical UE can be corrected by correcting the twin data on the edge cloud andthat valida

305、tion in the twin can be done by correcting data on the physical device.Forservices with higher requirements on bandwidth and latency,a twin can substantiallyease the computing pressure on the physical device.It relies on the powerfulcomputing power of the edge cloud to complete required computing ta

306、sks andtransmit them to the operator.This prevents the downlink bandwidth pressure due todata flowing to the UE and the higher latency due to insufficient UE computingpower.Figure 15 C-RAN-Based UE Digital Twin50Take a smart UE for commercial application for example.The mobile edge cloudin an operat

307、or network can generate a twin for the UE to synchronize the computingrequests with a large data volume in real time.When the user accesses VR or anothercomputing service with a large data volume,the twin will receive the data from theUE.With the powerful computing power of the server that hosts the

308、 twin,data iscomputed at speed and then returned to the UE to ensure service reliability and lowlatency of big data computing.In such a system framework,UEs do not need topossess strong computing power or influence the mobile,portable user experience likebulky computers.Instead,they satisfy the ever

309、-changing service requirements byhanding complicated computing over to UE twins and focusing on networkcommunication.In addition,applying UE twin technology enabled by digital twin in the futurenetwork architecture will not occupy excessive operator resources.A mobile edgecloud often has the twins o

310、f multiple mobile UEs using limited computing resourcesat the same time.However,once a UE is offline,the cloud recovers the informationabout the UE and stores it by encapsulation so that the twin will no longer occupysystem resources.When the UE registers to access the network again,the cloudrestore

311、s the information recovered previously.When multiple UEs are simultaneouslyonline,computing requests with large data volumes are uploaded in a time-sharingmanner to satisfy the needs of all UEs as far as possible.If multiple UEs are onlineand initiating computing requests with large volumes of data

312、simultaneously,thealgorithm should be further optimized.Using a proper communication and computingresource allocation algorithm on the cloud can address the allocation problem ofnetwork communication,computing,and caching resources,improving the serviceexperience for users.UE twin is an implementati

313、on form of edge computing and possesses someadvantages of digital twin technology.In edge computing technology,conventional task offloading is generallyimplemented by a cloud carrying and computing certain tasks on UEs.However,for51new services emerging in an endless stream,no convenient way can be

314、adopted toensure and validate the implementation result on UEs,leading to poor serviceexperience for users,despite the capability of the cloud to carry large-scale computing.Based on digital twin technology,UE twin technology builds the twins of physicalUEs on the mobile edge cloud.It emphasizes tha

315、t the system,software,and otherinformation on the twins are the same as those on the UEs(realized throughvirtualization and software-defined technologies)so that all types of data can be fullysynchronized between the two sides.UEs can upload various types of datainformation(without specifying any se

316、rvice)and model policies trained for certainservices can be validated and optimized on the twins and then delivered to UEs(without being requested by the UEs),ensuring service implementation results on theUEs.In addition,UE twins based on digital twin technology reach a balance betweenreal-time capa

317、bility and separability.Real-time capability means that a twin interactswith the physical UE in real time for real-time exchange of services and data.Separability means that a twin can even operate independently without the physicalUE.During the time separation,less communication bandwidth and other

318、 resourcesare consumed.It is also ensured that,within a period after separation,the consistencyof key indicators between the twin and the physical UE satisfies the degree ofprecision required for services(which is realized by deep learning and other AIapproaches).Later,the twin can interact with the

319、 physical UE to compare andsynchronize data.The advantage of the latter is what conventional task offloadingdoes not have.UE twins will provide technical support for emergency scenarioswhere service quality should be ensured despite the short supply of communicationresources.Compared with traditiona

320、l edge computing,UE twins can offer customizedcomputing services without scheduling by complex algorithms,substantially reducingthe latency of computing service provision.As noted,UE twins can simulate andpredict the transmission environment and requirements of actual users with its built-in52AI,the

321、reby optimizing and tuning the policies accordingly.While UE twin exploration will be meaningful,few studies have been conductedon the actual application of UE virtualization,in contrast to the extensive research onfuture network virtualization applications.In addition,for the edge cloud,the digital

322、 twin can also be converged with othertechnologies to bring more applications and benefits.For example,it can be combinedwith federated learning and blockchain technologies.Regarding bandwidth,latency,and other constraints,digital twin technology can be introduced for modelingoptimizationandalgorith

323、mstoreducetheedgeconnectionlatencyinthenext-generation operator networks.This will further optimize the positive influence ofthe introduction of twins on network architecture.Digital twin technology combinedwith deep reinforcement learning can be introduced to serve as a bridge between theentity lay

324、er and the application layer to seek algorithms to reduce the averageoffloading latency,the offloading failure rate,and the service migration rate in edgemeshes.In this way,the digital twin will help find the optimal solution for taskcommunication and computing and even act as a digital double to im

325、prove security bypreventing attacks on the actual network.5.Development Prospect of Digital TwinAs the 6G,AI/ML,and security technology advances and become mature,digitaltwin,as a technology relying on sensing and control,will help build moresophisticated and effective digital models,form a digital

326、twin society by connectinginformation silos,create a community with shared life for mankind18,and enablesustainable development.With the potential to be applied in a wide range of scenarios,the digital twin willgreatly empower 6G and various sectors,such as industry,agriculture,and city,bringingunli

327、mitedpossibilitiesandconveniencetoindustrialmanufacturing,agricultural production,urban governance,social services,and peoples life.Digitaltwin technology will continue to evolve to meet the new objectives and requirements53in building a community with shared life for mankind.There may be new techni

328、cal breakthroughs and potential standardization demandif digital twin is applied in the 6G network for purposes from providing technicalsupport,such as twin network modeling,data acquisition,and intelligent computing,to overcoming the adverse influences(higher cost and greater energy consumption)ofl

329、arge-scale data collection and transmission.AcknowledgementsGrateful thanks to the following contributors for their wonderful work on theWhitepaper:Editors:Chih-Lin I,Mengjun WangContributors:Cict Mobile:Mengjun Wang,Xianjun YangChina Mobile:Junshuai Sun,Juan Deng,Cheng Zhou,Quan Zhao,ZhiminZheng,Le

330、i Kong,Qingbi Zheng,Danyang ChenUniversity of Electronic Science and Technology of China:Jie Hu,Chang CheXidian University:Nan Cheng,ChungangYang,Hao Luan,Yilong Hui,ZhishengYin,Ying Ouyang,Lulu Zhang,Yanbo SongChina Unicom:Chang Liu,Zelin WangZTE:Dong ChenApple:Xiaoyu QiaoReferences1.White paper on

331、 digital twin applications,China Electronics StandardizationInstitute，2020.2.B.R.Barricelli,E.Casiraghi and D.Fogli,A Survey on Digital Twin:Definitions,Characteristics,Applications,and Design Implications,in IEEEAccess,vol.7,pp.,2019.543.Michael Grieves,Digital Twin:Manufacturing Excel

332、lence Through VirtualFactory Replication,Digital Twin Institute，2015.03.4.K.Framling,J.Holmstrm,T.Ala-Risku,and M.Karkkainen,Product agentsfor handling information about physical objects,Helsinki Univ.Technol.,Dept.Comput.Sci.Eng.Ser.B TKO-B,Espoo,Finland,2003.5.M.Shafto,M.Conroy,R.Doyle,E.Glaessgen

333、,C.Kemp,J.LeMoigne,and L.Wang.Modeling,Simulation,Information Technology&Processing Roadmap,National Aeronautics and Space Administration,Nov.2010.6.E.Tuegel,The airframe digital twin:Some challenges to realization,in Proc.53rd Struct.,Struct.Dyn.Mater.Conf.,2012,p.1812.7.Tao,Fei,et al.Digital Twin and Its Potential Application Exploration.J.Computer Integrated Manufacturing Systems,2018,24(01):1-