面向6G時(shí)代的先進(jìn)技術(shù)的初步研究：QITs 2023（英文）

PreliminaryStudyofAdvanc2023ExecutiveSummaryOverthepastyear,thelatestdevelopmentsinthe6thgeneration(6G)communicationsystemsresearchhavebeenreportedallovertheworld.6Gisemergingasanimportantdirectionforresearchanddevelopmentinthefieldofcommunications.Asoneenablingtechnologyfor6G,quantuminformationtechnologies(QITs)continuetoattractinterestfromacademiaandindustryduetotheexpectedinformationprocessingcapabilitiesbeyondtheirclassicalcounterparts.Inthe6Gera,theimportanceofcybersecurityinmobilecommunicationsisexpectedtoriseexponentially.Chapter2focusesonquantumsecurecommunicationaimingatsafeguardingcriticalinformationbyapplyingquantummechanisms.Theintroductionstartswithkeytechnologiesincludingquantumkeydistribution(QKD)andquantumrandomnumbergenerator(QRNG),followedbystate-of-the-artstandardizationactivitiesforquantumkeydistributionnetworks(QKDN)allovertheworld.Regardingtheimplicationsfor6G,ChinaUnicomhasbuiltaquantumkeycloudplatforminXiong’anNewAreaandcarriedoutawiderangeofquantumencryptiontechnologyresearchandapplicationdemonstrations.Thus,twooftherepresentativeapplicationscenarios,namely,quantumencryptedcallandquantumpublicnetworkclusterintercomwillalsobeintroducedinthischapter.Tosatisfythedramaticallyincreasedcommunicationsystemperformanceandrichdiversityofinnovativeservicesexpectedby6G,theemergingquantummachinelearning(QML)hasattractedsignificantattentionduetoitsinformationprocessingparadigmbycombiningtheestablishedbenefitsofquantummechanismsandmachinelearning.Consideringquantum-enhancedreinforcementlearninghasthepotentialtorevolutionizethefieldofartificialintelligence(AI),chapter3getsinsightintotheresearchofquantum-enhancedmachinelearningbyanalyzingrepresentativeworksindetailfromtwoaspects.Oneistostudyhowtospeedupthereinforcementlearning(RL)byapplyingthequantummechanism.Theothershowsanexperimentperformedtoreconstructanunknownphotonicquantumstatewithalimitedamountofcopies,forwhichtheperformanceintermsoffidelitiescanbeimprovedwiththeassistanceofthesemi-quantumreinforcementlearningapproach.過(guò)去一年中，有關(guān)第六代通信系統(tǒng)(6G)研究的最新進(jìn)展在全球范圍內(nèi)被廣泛報(bào)道。6G正逐漸成為通信領(lǐng)域的重要研發(fā)方向。作為6G的使能技術(shù)之一，量子信息技術(shù)(QITs)因其超越經(jīng)典信息技術(shù)的信息處理能力預(yù)期，在學(xué)術(shù)界和工業(yè)界開(kāi)始受到青睞。在6G時(shí)代，網(wǎng)絡(luò)安全在移動(dòng)通信中的重要性預(yù)計(jì)將呈指數(shù)級(jí)增長(zhǎng)。本白皮書(shū)在第2章重點(diǎn)介紹旨在通過(guò)應(yīng)用量子機(jī)制保護(hù)關(guān)鍵信息的量子安全通信。該章節(jié)首先介紹了量子密鑰分發(fā)(QKD)和量子隨機(jī)數(shù)生成器(QRNG)等關(guān)鍵技術(shù)，接著全面回顧了全球量子密鑰分發(fā)網(wǎng)絡(luò)(QKDN)的最新標(biāo)準(zhǔn)化活動(dòng)。關(guān)于量術(shù)的6G應(yīng)用，中國(guó)聯(lián)通在雄安新區(qū)建設(shè)量子密鑰云平臺(tái)，并開(kāi)展了廣泛的量子加密技術(shù)研究和應(yīng)用示范。本章也將分享了其中兩個(gè)具有代表性的應(yīng)用場(chǎng)景，即，量子加密通話和量子公網(wǎng)集群對(duì)講。為滿足6G所期望的大幅提高的通信系統(tǒng)性能和豐富多樣的創(chuàng)新服務(wù)，新興的量子機(jī)器學(xué)習(xí)(QML)處理范式融合了量子機(jī)制和機(jī)器學(xué)習(xí)的技術(shù)優(yōu)勢(shì)而備受關(guān)注。考慮到量子增強(qiáng)強(qiáng)化學(xué)習(xí)具有徹底改變?nèi)斯て湟谎芯咳绾瓮ㄟ^(guò)應(yīng)用量子方法來(lái)加速?gòu)?qiáng)化學(xué)習(xí)(RL)。其二展示了用有限數(shù)量的副本重建未知光子量子態(tài)的實(shí)驗(yàn)，在半量子強(qiáng)化學(xué)習(xí)方法的幫助下，可以提高保真度方面的性能。TableofContentsTableofContentsExecutiveSummary01前言1Introduction042QuantumSecureCommunication052.1KeyTechnologies 2.1.1OverallPicture 2.1.2QuantumKeyDistribution 2.1.3QuantumRandomNumberGenerator 2.2StandardizationActivitiesforQKDN 2.2.1ITU-T ITU-TStudyGroup13 ITU-TStudyGroup17 ITU-TStudyGroup11 ITU-TFG-QIT4N 2.2.3ISO/IECJTC1/SC27 2.3Implicationsfor6G 2.3.1ApplicationScenario1:Quantumencryptedcall 2.3.2ApplicationScenario2:Quantumpublicnetworkclusterintercom 3QuantumMachineLearning(QML)193.1Quantum-EnhancedReinforcementLearning 3.2ReconstructionofaPhotonicQubitStatewithReinforcementLearning Reference24Acknowledgement25PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs20231.IntroductionThescopeofthisannuallyrevisedwhitepaperistointroducequantuminformationtechnologies(QITs)withtheaimoftakingadvantageoftheirpowerfulinformationprocessingcapabilitiestofulfillstringentdemandsofcommunicationandcomputingenvisagedby6Gsystems.Theversionof2023willfurtherintroducetwobenefitsexpectedfromQITstocommunicationandcomputingsystems,i.e.,quantumsecurecommunicationandquantummachinelearning.Chapter2.QuantumSecureCommunicationIn6Gera,theimportanceofcybersecurityinmobilecommunicationsisexpectedtoriseexponentially.Chapter2focusesonquantumsecurecommunicationaimingatsafeguardingcriticalinformationbyapplyingquantummechanisms.Chapter2startswithkeytechnologiesincludingquantumkeydistribution(QKD)andquantumrandomnumbergenerator(QRNG),followedbystate-of-the-artstandardizationactivitiesforquantumkeydistributionnetworks(QKDN)allovertheworld.Regardingtheimplicationsfor6G,twonovelapplicationscenariosareintroduced,namely,quantumencryptedcallandquantumpublicnetworkclusterintercom.Chapter3.QuantumMachineLearning(QML)Tosatisfythedramaticallyincreasedcommunicationsystemperformanceandrichdiversityofinnovativeservicesexpectedby6G,theemergingQMLhasattractedsignificantattentionduetoitsinformationprocessingparadigmbycombiningtheestablishedbenefitsofquantummechanismsandmachinelearning.Consideringquantum-enhancedreinforcementlearninghasthepotentialtorevolutionizethefieldofartificialintelligence(AI),chapter3getsinsightintotheresearchofquantum-enhancedmachinelearningbyanalyzingrepresentativeworksindetailfromtwoaspects.Oneistostudyhowtospeedupreinforcementlearningbyapplyingthequantumapproach.Theothershowsanexperimentperformedtoreconstructanunknownphotonicquantumstatewithalimitedamountofcopies,forwhichtheperformanceintermsoffidelitiescanbeimprovedwiththeassistanceofthesemi-quantumreinforcementlearningapproach.PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs20232.QuantumSecureCommunication2.1.1OverallPictureQuantumcommunicationisanewandrapidlydevelopingcommunicationtechnologythathasbecomeahottopicinfrontierscienceandtechnology,thesecurityofwhichisguaranteedbyquantummechanics.Quantumkeydistribution(QKD)isthemostmaturelydevelopedquantumcommunicationtechnology,usingquantumsuperpositionstatesorentanglementtodistributequbits,withunconditionalsecurityatthetheoreticallevel.TheQuantumRandomNumberGenerator(QRNG)isknowntothegeneralpublicasarelativelymatureproductwiththehelpofQKD.QRNGisasystemforgeneratingtruerandomnumbersbasedontheprinciplesofquantumphysicsorquantumeffectsandhasimportantapplicationsinareassuchaspracticalquantumcommunicationsystems.2.1.2QuantumKeyDistributionTodate,therearemanydifferentprotocolsforQKD,allofthemcanbedividedintotwomaincategories:prepare-and-measure(PM)protocol,entanglement-based(EB)protocol.Fortheformer,thetransmittergeneratesarandombitsequenceandthenencodesthemonquantumstates,whicharesubsequentlysenttoreceivertomeasure.Forthelatter,onepartypreparesenoughentangledstatesfordistributingoverthechanneltotheotherparty,andthenpurifiesandmeasurestheentangledstatestoobtainthesecurekeys.BecausethePMschemeiseasiertoimplement,itisoftenusedtostructurepracticalsystems,inwhichthe"PreparedbybothpartiesandMeasuredbycenter"schemeismostlyadopted.Inaddition,itcanbefurtherdividedintotwotypes:DV-QKDandCV-QKD.ThesuccessfuldemonstrationoftheBB84QKDin1989provedthetheoreticalunconditionalsecurityofQKD.AlthoughQKDistheoreticallyunconditionallysecure,imperfectionsinthepracticaldevicescanexposethesystemtothreats.Therefore,Decoy-stateprotocolandMeasurement-device-independent(MDI)QKDareshowninTable2-1,whichaddressthevulnerabilitiesofweaklycoherentsourcesanddetectordevicesrespectively,enabletheunconditionalsecurityofQKDtobeguaranteedinthenon-perfectdevicecase.ThisisamajorPreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023advanceintheprotocolizationofQKD.Twin-Field(TF)QKDprotocol,forthefirsttime,breaksthePLOBboundwithoutaquantumrepeater,becomingawidelyrecognizedtechnicalsolutionforultra-long-rangeQKD.Withanewtransmissiondistanceof833kmin2022[2-1],TF-QKDisastepclosertobringingthe1,000kmquantumcommunication.Table21ThestagesofdevelopmentoftheQDKprotocolComparedwithDV-QKD,CV-QKDhasthecapabilityofMbit/shigh-speedkeyformationatshortandmediumtransmissiondistances,whichissuitableforhigh-speedmetropolitanareanetworkapplications.ThedevelopmentoftheCV-QKDsystemarchitectureisdividedintothreestages,withtherandomlocaloscillation(RLO),locallocaloscillation(LLO)anddiscretemodulateddigitalsystem,amongwhichthediscretemodulateddigitalsystemisexpectedtobecomethemainstreamcommercialsolutionforCV-QKDinthefuture.In2022,theLLO-CV-QKDsystemdemonstratedinthemetropolitanareawasreportedwithasecurekeyrateof21.53Mbit/sat25kmdistance[2-2],realizingLLO-CV-QKDwithultra-highsecurekeyrateandlayingasolidfoundationforCV-QKDwithevenhighersecurekeyrate.InthevariousQKDprotocolsmentionedabove,noneofthedevicesecurityriskshavebeencompletelyavoided,althoughMDI-QKDhasaddressedtheflawofattacker-controlledprobes.Theidealsolutionwouldbetoapplyanentanglement-basedDI-QKDsystemthatdealswiththesecurityvulnerabilitiesthatallowanattackertocontrolalldevicesandcanreachanupperlimitofinformation-theoreticsecurityatthephysicallevel.In2022,theBritish,GermanandChineseresearchteams,simultaneouslyreportedthreeexperimentaladvancesinDI-QKDproof-of-principle,enabling3.32bit/sinaDI-QKDsystembasedontheE91protocol[2-3],apredictiveentanglement-basedDD-QKDwithaBERof0.078[2-4],anda200mfiberDI-QKDbasedonpolarizationentangledphotons[2-5].Itisimportanttonotethatthesetechniquesaretheoreticallyvalidatedandarecurrentlydifficulttoindustrializeduetotheverystrongcapabilitiesofthehypotheticalattacker.PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023ThesatellitetransmissionQKDsystemisalsoamajordevelopmenttechnologyforquantumcommunication,anditsmaingoalistoconductsatellite-groundhigh-speedQKDexperimentswiththehelpofasatelliteplatform,andtoproceedwithwide-areaquantumkeynetworkexperiments.SatelliteQKDnetworkshaveuniqueadvantages.Ontheonehand,comparedwithopticalfibertransmissionQKD,satellitetransmissionQKDhaslowerlossandcansignificantlyincreasethetransmissiondistance.Ontheotherhand,satellitescanbeusedasrepeaters,whichcaneffectivelyimprovetheapplicationscopeandsecurityofQKD.Inrecentyears,countrieshaveattachedgreatimportancetothedevelopmentofsatelliteQKDandhavecarriedoutaseriesofexperimentsonsatellitequantumnetworks.In2022,China'sMozisatellitehasreachedthecurrentfarthestQKDof1200km[2-6]andlaunchedtheworld'sfirstQKDmicro-nano-satellite"Jinan-1"[2-7].AsQKDisonthecommercializationtrack,integratedphotonicsprovidesapowerful,miniaturizedandcost-effectiveplatformtoimplementQKDtransmitterandreceiverdevices.ThedesignofintegratedQKDsystemsrequirestheselectionofdifferentopticaldesignsaswellasmaterialplatformsdependingontherequirementsoftheapplication.Silicon-basedplatformsofferprovenprocessingplatformsbutrequiretheuseofhybridintegratedlasersources;InPplatformsallowmonolithicintegrationoflasersandhigh-speedphasemodulators,butdevicesizeaswellascostaspectsstillneedtobeimproved.Futuredevelopmentsinfull-chipQKDtendtousenotjustoneofthesematerials,butacombinationofseveralmaterialstodesigndevicessuitableforthesystem,therebyreplacingalargenumberofhigh-performancediscretedevices,reducingdevicecostandsize,improvingsystemintegration,andfurtherpromotingthelarge-scalecommercializationofquantumcommunicationsystems.2.1.3QuantumRandomNumberGeneratorQuantumRandomNumberGenerator(QRNG)isasystemthatgeneratestruerandomnumbersbasedontheprinciplesofquantumphysicswiththecharacteristicsofunpredictability,irreducibility,andunbiasedness,whichisavitaldeviceinquantumcommunicationsystemsandcanbeappliedinQKDsystems.IntheQRNGsystem,thecorrespondingquantumstateneedstobepreparedfirst.Afterward,thequantumstateismeasuredandtherawdataisobtained.ThequantumrandomnessPreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023containedintherawdatacanbequantifiedbymodelingaswellasbycalculation.Basedontheresultsofthequantizationanalysis,therawdataarepost-processedtoobtainthefinaltruerandomnumber.QRNGsaredividedintotwomaincategories:discreteandcontinuous,dependingontherandomsourceused.ThediscreteQRNGmainlyusessignalssuchassinglephotonsourcesandentangledphotonpairsascarriersofrandomvariables.Theschemeissimpleinprincipleandhasobviousquantumuncertainty,buttherandomnumbergenerationrateofthisschemeislow,whichismainlylimitedbythelinewidthoftherandomsourceandthedetectionefficiencyofthesingle-photondetector.ThecontinuousQRNGusesthetruerandomnessofthespontaneousradiationphotonphasetoconverttherandomfluctuationphaseintolightintensity,whichisthencapturedandquantizedbyahigh-speedanalog-to-digitalconvertertoobtaintherawquantumrandomnumber.Thisschemeisnotrestrictedbythesaturationcountrateofsingle-photondetectorsandsubstantiallyincreasesthegenerationrateofrawrandomnumbers.Currently,thedevelopmentdirectionofQRNGtechnologyisfocusedonincreasingthegenerationrateofquantumrandomnumbers,miniaturizationofquantumrandomnumbergeneratingdevices,andreducingthecostofquantumrandomnumbergenerators.TherandomnumbergenerationrateisthemostimportantmetricforQRNG.In2022,GhentUniversity,togetherwiththeTechnicalUniversityofDenmarkandthePolitecnicodiBariinItaly,experimentallydemonstratedanultra-fastgenerationrateof100Gbit/s[2-8],raisingthenewrecordforvacuumquantumrandomnumbergenerationbyanorderofmagnitude.Besides,QRNGchipswithstableperformance,lowcost,andhighvolumeproductionhavebecomeanurgentrequirementforcryptographicsystems.Manycompaniesandresearchinstitutesareconductingminiaturizationandchip-basedresearch,andavarietyoftechnologysolutionsanddeviceformsarebecomingcommerciallyavailableforQRNGproducts,withthehighestrandomnumbergenerationratesincreasingto10Gbit/s.Korea'sSKTandSamsunglaunchedGalaxyQuantum3smartphonetopromotechip-basedQRNGinmobileterminalauthenticationandinformationencryptionapplications.Inthefuture,QRNGisexpectedtoentertheconsumermarketrapidlyastheQRNGchip-basedtechnologymaturesandcost-effectivenessisrealized.PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs20232.22.2StandardizationActivitiesforQKDNQKDanditsnetworkingtechnologieshaveattractedalotofinterestinmultipleSDOs,e.g.,ISO,IEC,ITU,IEEE,IETF,ETSI,asshowninFigure2-1.ThestatusofQuantumKeyDistributionNetworks(QKDN)standardizationindifferentSDOswillbebrieflyreviewedinthefollowingsub-clauses.Figure21QKDNstandardizationtimeline2.2.1ITU-TITU-TwasthefirstSDOtostandardizeQKDasanetwork.InJuly2018,ITU-TSG13initiatedthefirstworkitem(i.e.,Y.3800)onQKDandbroughtintheconceptofQuantumKeyDistributionNetwork(QKDN)firstly.Afterwards,therearemorethan40workitemsconductedby4differentgroupsinITU-TundertheumbrellaofQKDN,whichcanbedividedinto4branchesasfollows:■StudyGroup13(Q16/13andQ6/13):focusonnetworkaspectsofQKDN■StudyGroup17(Q15/17,formerlyQ4/17):focusonsecurityaspectofQKDN■StudyGroup11(Q2/11):focusonQKDNhighlayerprotocolsandsignaling■FocusGrouponQuantuminformationtechnologyforNetworks(FG-QIT4N):tostudytheimplicationsofQITsforbothquantumandICTnetworkPreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023ITU-TStudyGroup13AlandscapediagramfortheQKDNstandardizationworkinSG13isasillustratedinFigure2-2.SG13hasthefollowingworkitemsonQKDNaslistedinTable2-2.Figure22:QKDNstandardizationlandscapeinITU-TSG13Table21ThestagesofdevelopmentoftheQDKprotocolPreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023ITU-TStudyGroup17AlandscapediagramfortheQKDNstandardizationworkinSG17isillustratedinFigure2-3.SG17hasthefollowingworkitemsonQKDNaslistedinTable2-3.PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023Figure23:QKDNstandardizationworkitemsinSG17Table23QKDNrelatedworkitemsinITU-TSG17PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023ITU-TStudyGroup11AlandscapediagramfortheQKDNstandardizationworkinSG11isillustratedinFigure2-4.SG11hasthefollowingworkitemsonQKDNprotocols,aslistedinTable2-4.Figure24:QKDNstandardizationworkitemsinSG11Table24QKDNrelatedworkitemsinITU-TSG11ITU-TFG-QIT4NFG-QIT4NhasthefollowingworkitemsonQKDNaslistedinTable2-5.Table24QKDNrelatedworkitemsinITU-TSG11PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023ETSIinitiatedtheindustryspecificationgroup(ISG)onQKDin2008.ETSIISG-QKDhaspublishedninespecificationsonQKDuntil2019andhaveseveralworkitemsongoingaslistedinTable2-6.ThepreviousworkmainlyfocusedonQKDlink-levelissues,includingQKDopticalcomponents,modules,internalandapplicationinterfaces,practicalsecurity,etc.NotethatETSIhasalsoinitiatedthestudyofQKDnetworkarchitecturesrecentlyandthespecificationofQKDsecuritycertificationbasedoncommoncriteria.Table26:QKDrelatedworkitemsinETSIPreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023ISO/IECJTC1/SC27initiatedthestudyperiod"Securityrequirements,testandevaluationmethodsforquantumkeydistribution"in2017.In2019,thestudyperiodwascompleted,andanewworkitemISO/IEC23837(Part1&2)wasestablishedaslistedinTable2-7.Table27:QKDrelatedworksitemsinISO/IECJTC12.32.3Implicationsfor6GTheQuantumkeycloudplatformobtainsquantumkeysfromQKDorQRNG,andstoresandmanagesthekeysafely.Throughthesecuritymechanism,thequantumkeyscanbedistributedtotheusersecurityterminalandprovidehigh-levelsecurityprotectiontotheusersecurityterminaleveninthefaceofchallengesofquantumcomputing.TheQuantumkeycloudplatformcanprovidequantumencryptionservicesforthegovernment,enterprises,andindividualstoprotectthestorageandtransmissionofdatasafely.Atpresent,ChinaUnicomhasbuiltaQuantumkeycloudplatforminXiong’anNewAreaandcarriedoutquantumencryptiontechnologyresearchandapplicationdemonstration,suchasquantumencryptedcall,quantumpublicnetworkclusterintercom,quantumvideoconference,andquantumUAVpatrol.Thefollowingdescribestwooftherepresentativeapplicationscenarios,namely,quantumencryptedcallandquantumpublicnetworkclusterintercom.PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs20232.3.1ApplicationScenario1:QuantumencryptedcallFigure25SystemdiagramofQuantumencryptedcallIntheapplicationscenarioofquantumencryptedcall,thesecurityterminalusesthepre-chargedquantumkeysas“theidentityauthenticationkeys”and“thebasicencryptionkeys”.Whenmakingacallorsendingamessage,theQuantumkeycloudplatformselectsasetofquantumkeysas“thesessionkeys”,andencryptsthemwith“thebasicencryptionkeys”andsendsthemtothesecurityterminal.Thesecurityterminaldecrypts“thesessionkeys”with“thebasicencryptionkeys”,andthen“thesessionkeys”canbeusedtoprotectthevoiceanddatastreamofthesecurityterminal.Inaddition,wehavedevelopedaspecialApp,throughwhichthequantumsecurityterminalcanrealizeencryptedtransmissionoftext,voice,pictures,files,andothercontents.AndtheAppsupportsthefunctionof"burnafterreading".PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs20232.3.2ApplicationScenario2:QuantumpublicnetworkclusterintercomFigure26SystemdiagramofQuantumpublicnetworkclusterintercomIntheapplicationscenarioofquantumpublicnetworkclusterintercom,theterminalandtheCommandanddispatchingplatformintegratethequantumSDK,whichcanobtainthequantumkeysfromtheQuantumkeycloudplatformandperformquantumencryptiontoensurethesecurityoftheclustervoice,video,imageandotherservicedataandoperationalsignalingwhentransmittedoverthepublicnetwork.IftheeavesdropperusesterminalsCandDwithoutintegratingquantumencryptionfunctiontoillegallyentertheclusterintercomsystemofterminalsAandB,whichintegratingquantumencryptionfunction,whenterminalsAandBsendvoiceandvideomessages,terminalsCandDcannotcrackthereceivedquantumencryptedinformation,thatis,theycannotreceivevoiceorvideoinformationnormally.However,terminalsAandBcanreceivethemessagesnormally,whichweresentbyterminalsCandD.PreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs20233.QuantumMachineLearning(QML)Itishighlyexpectedthatthe6thgeneration(6G)communicationsystemswilllayafoundationofpervasivedigitization,ubiquitousconnectionandfullintelligence.Tosatisfythedramaticallyincreasedcommunicationsystemperformanceandrichdiversityofinnovativeservices,QuantumMachineLearning(QML)isemergedduetoitsinformationprocessingcapabilitybeyonditsclassicalcounterpart,whichisachievedbycombiningtheestablishedbenefitsofquantummechanismandmachinelearning.Inthewhitepaperofversion2021,weintroducedtheconceptsandbasicparadigmsofQMLonahighlevel.Whereinquantum-enhancedmachinedlearningcanbefurthercategorizedaccordingtothethreebranchesofML(i.e.,supervisedlearning,unsupervisedlearning,andreinforcementlearning).Inparticular,quantum-enhancedreinforcementlearninghasapotentialtorevolutionizethefieldofartificialintelligence(AI).Inthisfollowing,wewillgetinsightintotheresearchofquantum-enhancedmachinelearningbyanalyzingtworepresentativeworksindetail.Thefirstworkin[3-1]gainsspeed-upofreinforcementlearningbyprobingtheenvironmentinsuperpositionsandprovidesageneralmethodofquantumimprovementsinthethreeparadigmsofmachinelearning.Thesecondworkin[3-2]showsanexperimentperformedtoreconstructanunknownphotonicquantumstatewithalimitedamountofcopies,forwhichtheperformanceintermsoffidelitiescanbeimprovedwhenassistedbysemi-quantumreinforcementlearningapproach.3.13.1Quantum-EnhancedReinforcementLearningReinforcementLearning(RL)[3-3]isanareaofmachinelearningconcernedwithhowintelligentagentsreactinanenvironmentwithatargetofmaximizingthereward.ThefocusofRLisonfindingabalancebetweenexploration(ofunchartedterritory)andexploitation(ofcurrentknowledge)[3-4].Ascomparedtosupervisedlearning,labeledtrainingdataisnotrequiredforreinforcementlearning.However,partiallysupervisedRLalgorithmscancombinetheadvantagesofsupervisedandRLalgorithms.OnepowerfulfeatureofRLissuitablefordealingwithlargeenvironments.Reinforcementlearningistypicallyusedforsolvingcontrolandclassificationproblems.ConventionalandnotableRLalgorithmssuchasQ-learningandmulti-armedbandittakeasaninputthecurrentstateofthenetworkandenablethepredictionofthenextstate.APreliminaryStudyofAdvancedTechnologiestowards6GEra:QITs2023promisingapplicationofRLincommunicationcontributestoschedulingparametersoptimizationacrossvariouslayers.Additionally,deeplearningcanbecombinedwithRLtofacilitatelearninglong-termtemporaldependencesequencesinsuchawaythattheaccumulationoferrorswon’tgrowveryfast[3-5].Inquantum-enhancedreinforcementlearning,aquantumagentinteractswithaclassicalorquantumenvironmentandoccasionallyreceivesrewardsforitsactions,whichallowstheagenttolearnwhattodoinordertogainmorerewards.Therearevariouswaysofachievingquantumspeedup.Forexample,in[3-6]aquantumagentwhichhasquantumprocessingcapabilityisprovidedinachievingaquadraticspeed-upforactivelearning.Alternatively,theworkin[3-1]gainsspeed-upbyprobingtheenvironmentinsuperpositions.Furthermore,ageneralmethodofquantumimprovementsinthethreeparadigmsofmachinelearningisprovidedin[3-1].Thissectionwillintroducethemajorworkin[3-1].TheQMLcanberepresentedbyanagent-environmentparadigm,wherealearningagentAinteractswithinteractswithanunknownenvironmentEviatheexchangeofmessages,interchangeablyissuedbyA(calledactions)andE(calledpercepts).Forreinforcementlearning,theperceptspacealsocontainsthereward.Inthequantumextension,bothAandEarequantumsystems,wherethesetsofactionsandperceptsbecomeHilbertspacesandformorthonormalbases.TheagentandtheenvironmentactonacommoncommunicationregisterRC(capableofrepresentingbothperceptsandactions).T