Massive-MIMO-for-Next-Generation-Wireless-Systems

IEEECommunicationsMagazine�February20141860163-6804/14/$25.00©2014IEEEGOINGLARGE:MASSIVEMIMOMassivemultiple-inputmultiple-output(MIMO)isanemergingtechnologythatscalesupMIMObypossiblyordersofmagnitudecomparedtothecurrentstateoftheart.Inthisarticle,wefollowuponourearlierexposition[1],withafocusonthedevelopmentsinthelastthreeyears;mostparticularly,energyefficiency,exploitationofexcessdegreesoffreedom,time-divisionduplex(TDD)calibration,techniquestocombatpilotcontamination,andentirelynewchannelmea-surements.WithmassiveMIMO,wethinkofsystemsthatuseantennaarrayswithafewhundredantennassimultaneouslyservingmanytensofterminalsinthesametime-frequencyresource.ThebasicpremisebehindmassiveMIMOistoreapallthebenefitsofconventionalMIMO,butonamuchgreaterscale.Overall,massiveMIMOisanenablerforthedevelopmentoffuturebroadband(fixedandmobile)networks,whichwillbeenergy-efficient,secure,androbust,andwillusethespectrumefficiently.Assuch,itisanenablerforthefuturedigitalsocietyinfra-structurethatwillconnecttheInternetofpeopleandInternetofThingswithcloudsandothernetworkinfrastructure.Manydifferentconfigu-rationsanddeploymentscenariosfortheactualantennaarraysusedbyamassiveMIMOsystemcanbeenvisioned(Fig.1).Eachantennaunitwouldbesmallandactive,preferablyfedviaanopticalorelectricdigitalbus.MassiveMIMOreliesonspatialmultiplexing,whichinturnreliesonthebasestationhavinggoodenoughchannelknowledge,onboththeuplinkandthedownlink.Ontheuplink,thisiseasytoaccomplishbyhavingtheterminalssendpilots,basedonwhichthebasestationestimatesthechannelresponsestoeachoftheterminals.Thedownlinkismoredifficult.InconventionalMIMOsystemssuchastheLongTermEvolu-tion(LTE)standard,thebasestationsendsoutpilotwaveforms,basedonwhichtheterminalsestimatethechannelresponses,quantizethethusobtainedestimates,andfeedthembacktothebasestation.Thiswillnotbefeasibleinmas-siveMIMOsystems,atleastnotwhenoperatinginahigh-mobilityenvironment,fortworeasons.First,optimaldownlinkpilotsshouldbemutuallyorthogonalbetweentheantennas.Thismeansthattheamountoftime-frequencyresourcesneededfordownlinkpilotsscaleswiththenum-berofantennas,soamassiveMIMOsystemwouldrequireupto100timesmoresuchresourcesthanaconventionalsystem.Second,ABSTRACTMulti-userMIMOoffersbigadvantagesoverconventionalpoint-to-pointMIMO:itworkswithcheapsingle-antennaterminals,arichscat-teringenvironmentisnotrequired,andresourceallocationissimplifiedbecauseeveryactiveter-minalutilizesallofthetime-frequencybins.However,multi-userMIMO,asoriginallyenvi-sioned,withroughlyequalnumbersofserviceantennasandterminalsandfrequency-divisionduplexoperation,isnotascalabletechnology.MassiveMIMO(alsoknownaslarge-scaleantennasystems,verylargeMIMO,hyperMIMO,full-dimensionMIMO,andARGOS)makesacleanbreakwithcurrentpracticethroughtheuseofalargeexcessofserviceantennasoveractiveterminalsandtime-divisionduplexoperation.Extraantennashelpbyfocus-ingenergyintoeversmallerregionsofspacetobringhugeimprovementsinthroughputandradiatedenergyefficiency.OtherbenefitsofmassiveMIMOincludeextensiveuseofinexpen-sivelow-powercomponents,reducedlatency,simplificationoftheMAClayer,androbustnessagainstintentionaljamming.Theanticipatedthroughputdependsonthepropagationenviron-mentprovidingasymptoticallyorthogonalchan-nelstotheterminals,butsofarexperimentshavenotdisclosedanylimitationsinthisregard.WhilemassiveMIMOrendersmanytraditionalresearchproblemsirrelevant,ituncoversentirelynewproblemsthaturgentlyneedattention:thechallengeofmakingmanylow-costlow-precisioncomponentsthatworkeffectivelytogether,acquisitionandsynchronizationfornewlyjoinedterminals,theexploitationofextradegreesoffreedomprovidedbytheexcessofserviceanten-nas,reducinginternalpowerconsumptiontoachievetotalenergyefficiencyreductions,andfindingnewdeploymentscenarios.ThisarticlepresentsanoverviewofthemassiveMIMOcon-ceptandcontemporaryresearchonthetopic.ACCEPTEDFROMOPENCALLErikG.Larsson,ISY,LinköpingUniversity,SwedenOveEdforsandFredrikTufvesson,LundUniversity,SwedenThomasL.Marzetta,BellLabs,Alcatel-Lucent,UnitedStatesMassiveMIMOforNextGenerationWirelessSystemsEDFORS_LAYOUT.qxp_Layout1/30/141:08PMPage186IEEECommunicationsMagazine�February2014187thenumberofchannelresponseseachterminalmustestimateisalsoproportionaltothenumberofbasestationantennas.Hence,theuplinkresourcesneededtoinformthebasestationofthechannelresponseswouldbeupto100timeslargerthaninconventionalsystems.Generally,thesolutionistooperateinTDDmode,andrelyonreciprocitybetweentheuplinkanddown-linkchannels,althoughfrequency-divisionduplext(FDD)operationmaybepossibleincer-taincases[2].WhiletheconceptsofmassiveMIMOhavebeenmostlytheoreticalsofar,stimulatingmuchresearchparticularlyinrandommatrixtheoryandrelatedmathematics,basictestbedsarebecomingavailable[3],andinitialchannelmea-surementshavebeenperformed[4,5].THEPOTENTIALOFMASSIVEMIMOMassiveMIMOtechnologyreliesonphase-coherentbutcomputationallyverysimplepro-cessingofsignalsfromalltheantennasatthebasestation.SomespecificbenefitsofamassiveMU-MIMOsystemare:�MassiveMIMOcanincreasethecapacity10timesormoreandsimultaneouslyimpr