您的浏览器禁用了JavaScript(一种计算机语言,用以实现您与网页的交互),请解除该禁用,或者联系我们。[未来移动通信论坛]:边缘原生智能白皮书 - 发现报告

边缘原生智能白皮书

AI智能总结
查看更多
边缘原生智能白皮书

1/123AbstractEdgecomputing,asakeytechnologyofthenextgenerationofradioaccessnetworks(RAN),hasdriventhedecentralizationofnetworksandcomputingfacilities.Edgeserversclosertouserterminalscansignificantlyreduceservicelatencyandcopewithemergingnewscenarios.Simultaneously,therapiddevelopmentofartificialintelligence(AI)playsasignificantroleinenhancingtheperformanceofedgecomputing,aidingedgedevicesincopingwiththerapidlyincreasingdataontheedge.Therefore,combiningthelocalcomputingcapabilityofedgedatawiththestrongcomputingcapabilitiesofAI,knownasedgeintelligence,canenhancethedataprocessingcapabilitiesontheedge,improvetheoverallperformanceofwirelesscommunicationsystems,andenhanceuserserviceexperiences.Edgeintelligenceisahotandrapidlydevelopingfieldinrecentyears,andthiswhitepaperaimstoanalyzethecurrentresearchprogressinedgeintelligence.Itmainlyincludes:(1)6GEdgeIntelligenceNetworksandInfrastructure:Firstly,theedge-nativeintelligentarchitecturefor6Gnetworksisanalyzed.Then,theedgeintelligencecomputinginfrastructureisintroduced,includingedgeintelligenthardwareandcloudplatforms.Finally,theedgeintelligencenetworkinfrastructureisdescribed,includingtheedgeintelligenceaccessnetworkandcorenetwork.(2)KeyTechnologiesofEdge-NativeIntelligence:Itisintroducedfromtheaspectsofmodellightweight,edge-cloudcollaborativeintelligence,edgeintelligentdeployment,anddeepedgenodes.Edgeintelligenceinwirelessfederatedlearningisalsoexplainedindetail,includingmodelsparsificationandmodelquantizationin(3)ApplicationsofEdge-NativeIntelligence:Typicalapplicationsofedge-nativeintelligenceareanalyzed,suchassmarttransportation,smartmanufacturing,and federatedlearning.smartenergysaving. 2/123Contents1.Introduction.....................................................................................................................................31.1Background...........................................................................................................................31.2OverviewofEdgeComputingandEdge-NativeIntelligence.............................................41.3ImportanceofEdge-NativeIntelligence..............................................................................52.6GEdgeIntelligenceNetworksandInfrastructure........................................................................72.1Edge-NativeIntelligenceArchitecturefor6G.....................................................................72.2EdgeIntelligenceComputingInfrastructure......................................................................122.3EdgeIntelligenceNetworkInfrastructure..........................................................................373.KeyTechnologiesofEdge-NativeIntelligence............................................................................593.1ModelLightweighting........................................................................................................593.2Edge-CloudCollaborativeIntelligence..............................................................................683.3WirelessFederatedLearninginEdgeIntelligence.............................................................753.4EdgeIntelligenceDeployment...........................................................................................853.5DeepEdgeNodes...............................................................................................................914.Edge-NativeIntelligenceApplications.......................................................................................1014.1SmartTransportation........................................................................................................1014.2SmartManufacturing........................................................................................................1104.3IntelligentEnergySaving.................................................................................................1185.DevelopmentandChallengesofEdge-NativeIntelligence.......................................................1206.Acknowledgment........................................................................................................................123 1.Introduction1.1BackgroundFrom1Gto5G,communicationtechnologyhasundergonemultipleupgradesandtransformations,significantlyimprovingdatatransferrates,reducinglatency,andexpandingnetworkcoverage.However,withtherapiddevelopmentoftechnologiessuchas the Internet of Things(IoT)and AI,the Internet of Everything andincreasinglycomplex application scenarios pose challenges that existing networkarchitecturescannot meet.Therefore,as the next generation of communicationtechnology,6G must possess higher performance and more powerful intelligentcapabilities,driving the transition of edge-side networks from"Internet ofEverything"to"IntelligenceofEverything."Tobetteradapttofuturediverseandcomplexuser requests and application scenarios,the concept of edge-nativeintelligencecameintobeingbyintegratingintelligenttechnologyintothedesignandimplementationofcommunicationsystems.Inrecentye