Data Driven and Intent Aware Smart Wireless Networ

5GWhitePaper:DataDrivenandIntentAwareSmartWirelessNetworkDataDrivenandIntentAwareSmartWirelessNetwork20195GWhitePaper:DataDrivenandIntentAwareSmartWirelessNetworkExecutiveSummaryToaccommodateawiderangeofscenariosandservicerequirements,the5Gnetworkisbecomingmoreagile,andinevitably,morecomplicated,thusposinggreatchallengestonetworkmanagementandradioresourceoptimization.Intelligenceisexpectedtobeintroducedinwirelessnetworkforallthedomainsandalllevels,fromlocal,edgetothecloud.Dataanalytics,machinelearningandartificialintelligenceareidentifiedasthekeydriversfortheintelligenceevolutionandrevolutioninthewirelessnetwork.Datadrivensmartwirelessnetworkusagescenariosandnetworkarchitecturearebeingheatedlydiscussedandinvestigatedrecentlybothinthewirelessnetworkindustryandacademia.“SmartandSimplicity”becometheindustryconsensusfor5Gandfuturenetwork.“Smart”requiresthenetworktodynamicallyadapttothediversifiedscenariosandservicestoefficientlyboostbothspectrumefficiency,energyefficiencyandoffertheconsistencytheuserexperience.“Simplicity”posestherequirementsofnetworkautomationtosignificantlyreducethenetworkoperationandmaintenancehumanlaborandcost.Thiswhitepaperinvestigatesthedatadrivenandintentawaresmartwirelessnetwork.Itisanupdateandevolutionofthepreviouswhitepapers“WirelessBigDataforSmart5G”and“WirelessBigDataandAIforSmart5G&Beyond”,wherethereferencearchitectureandusecasearecomprehensivelystudied.Inthiswhitepaper,datadrivenwirelessnetworkisfirstlyintroducedandreviewed.Differentcategoriesoftypicalusecasesandpotentialsolutionaredescribed,fromthenetworkmanagementandoperationtoMECoptimization,thenetworkresourceoptimizationandtheintelligenttransmissiontechnologies.TheMLanddataanalyticsalgorithmsutilizedinthealltheusecasesoftheseriesofwhitepapersarealsodiscussedandsummarizedtoprovidesomeinsightfortheresearcheronthefutureworks.Furthermore,intentawarewirelessnetworkisproposedandinvestigatedtoachievethenetwork“simplicity”leveragingthedatadrivennetworkintelligence.Theconceptandmotivationoftheintentawarenetworkisfirstlybrieflydescribed.Thereferencearchitectureofintentdrivenisinvestigatedwithdiscussionoftheintentexpression.Besides,thekeyissues,challengesandtheintentserviceevolutionroadmaparepresented.Lastbutnottheleast,thestandardizationprogressarediscussed.Wewishthiswhitepapercanprovidereaderswithsomevaluableinsightsforindustryonthedatadrivenandintentawaresmartwirelessnetworkandhelptoacceleratetheprocessofsmartwirelessnetwork.5GWhitePaper:DataDrivenandIntentAwareSmartWirelessNetwork1中文摘要为了适应多种多样的场景和服务需求，5G网络正变得越来越灵活，也不可避免地变得越来越复杂，给网络管理和无线资源优化带来了巨大挑战。为了应对挑战，智能有望被引入到从本地，边缘到云的无线网络的所有域和所有层级。“智能极简”成为5G和未来网络的行业共识。“智能”要求网络动态适应不同的场景和服务，以有效地提高频谱效率，能效并提供定制化的极致用户体验。“极简”提出了网络自动化的需求，以显着减少网络运维的人力和成本。本白皮书研究了数据驱动和意图感知的智能无线网络。这是对先前“无线大数据和智慧5G”和“无线大数据与人工智能使能的智慧5G+”系列白皮书的更新和发展，先前两本白皮书中对数据驱动的智能网络参考架构和用例均进行了研究和分享。在本白皮书中，首先回顾了数据驱动的无线网络，更新了网络运维、边缘计算优化、网络资源优化和智能传输技术的典型用例及潜在解决方案的进展；其次还进一步分析了智能无线网络用例所使用的机器学习、数据分析算法及数据需求，希望能为研究人员对未来的工作提供一些参考。本白皮书中还进一步提出了意图感知（或者意图驱动）的无线网络，研究如何简化无线网络管理接口，利用数据驱动的智能化手段将网络管理意图自动转换为复杂的网络管控操作，以更好的实现网络自治化，减少运维人员的人力和工作复杂度，并进一步面向垂直行业提供更简易的网络管理能力开放。文中首先简要描述了意图感知网络的动机和概念，并通过讨论意图感知的无线网络场景和用例，进一步研究了意图感知无线网络的参考架构。其次提出了意图感知无线网络的关键问题和挑战，并对意图感知无线网络的演进路线进行了讨论。最后，本白皮书总结了ITU-T/3GPP/ETSI等在数据采集、数据分析、机器学习、意图驱动与智能5G网络相关的标准化进展。我们希望本白皮书可以和读者分享有关数据驱动和意图感知的智能无线网络的一些有价值的技术探索以及学术、产业界的最新进展，为学术界及产业界在研究、规划和设计5G网络智能化相关技术、产品和解决方案时提供一些参考和指引，从而加快智能无线网络的技术发展与产业化落地。5GWhitePaper:DataDrivenandIntentAwareSmartWirelessNetworkTableofContents2SmartWirelessNetwork......................................................................................................................................12.1DataDrivenWirelessNetwork................................................................................................................12.2IntentAwareWirelessNetwork..............................................................................................................23DataDrivenWirelessNetwork............................................................................................................................43.1OverallIntroduction.................................................................................................................................43.2NetworkManagementOptimization........................................................................................................43.2.1Usecase1:AnomalyDetectionwithKPIs..................................................................................43.2.2Usecase2:AnomalyDiagnosiswithKPIs.................................................................................73.2.3Usecase3:SmartCarrierLicenceResourceScheduling............................................................93.2.4Usecase4:GeoDataEnhancedCoverageEstimationandNetworkPlanning.........................133.3MECOptimization.................................................................................................................................143.3.1Usecase5:IntelligentMECCachingOptimization............................................