作者：JiajiaGuo1,XiangyiLi1,MuhanChen1,PeiwenJiang1,TingtingYang2,WeimingDuan2,HaowenWang3,ShiJin1,QuanYu2单位：1.NationalMobileCommunicationsResearchLaboratory,SoutheastUniversity;2.PengChengLaboratory;3.LaboratoryofBroadbandWirelessTechnology,ShanghaiInstituteofMicrosystemandInformationTechnology,ChineseAcademyofSciences近年来，人工智能（ArtificialIntelligence，AI）在信号处理、信道估计、编码设计等通信领域取得了重大突破，打破了传统通信系统的设计瓶颈，作为一项突破性技术使得智能通信成为未来通信系统研究的热门方向之一。基于深度学习（DeepLearning，DL）的信道状态信息（ChannelStateInformation，CSI）反馈技术因其突出的性能优势得到了广泛关注，但目前相关研究仅使用模拟生成的数据集来训练和测试，无法保证AI算法在实际通信系统中仍然具有良好的性能。为了探索AI在通信系统中的实际性能，鹏城实验室等单位组织了全国人工智能大赛（NAIC）“AI+无线通信”赛道，初赛赛题为“基于AI的无线通信信道的压缩及恢复”。本文详细描述了该比赛的信道数据采集过程，同时为该比赛提供了一个基于DL的CSI反馈参考架构：QuanCsiNet，实现了真实信道场景采集的高维信道数据的压缩、量化、反馈和重建，为AI在未来通信系统中的实际部署和使用奠定了基础。
真实信道场景为图1所示的办公室场景，发射机在图中所示位置固定，接收机沿红点轨迹运动。对测量得到的真实信道数据进行图2所示的预处理，得到最终使用的数据集。图1实测信道数据场景示意图图2实测信道数据处理流程QuanCsiNet的网络结构如下图所示。
编码器用于对CSI进行特征提取和压缩，量化模块用于将压缩测量值用有限位表示，转化为比特流便于实际系统存储和传输；逆量化模块用于将比特流恢复成压缩测量值，译码器用于特征解压缩和信道恢复。图3QuanCsiNet网络结构具体来说，文章首次针对实测信道数据进行处理，提出了一种真实信道场景下的CSI反馈架构，为后续研究提供了可扩展的参考设计。引入了量化和逆量化模块，将反馈测量值转化为比特流，符合实际系统存储传输要求。
评估了基于DL的CSI反馈方案在实际信道环境中的性能，推动后续研究和实际部署。整体而言，本文对真实信道场景下采集的信道数据进行处理，设计了一种以比特流形式进行反馈的基于DL的CSI反馈架构，并衡量了其重建性能与复杂度，以期更多研究者为智能通信的实际应用作出贡献。
论文下载链接：（请戳此处）引用格式：JiajiaGuo,XiangyiLi,MuhanChen,PeiwenJiang,TingtingYang,WeimingDuan,HaowenWang,ShiJin,QuanYu,“AIEnabledWirelessCommunicationswithRealChannelMeasurements:ChannelFeedback”,JournalofCommunicationsandInformationNetworks,vol.5,no.3,pp.310-317,Sep.2020.本文由论文作者供稿。欢迎课题组投递成果宣传稿（可在文章下方留言联系我们）！作者简介JiajiaGuo（郭佳佳）receivedtheB.S.degreefromNanjingUniversityofScienceandTechnology,Nanjing,China,in2016,andtheM.S.degreefromUniversityofScienceandTechnologyofChina,Hefei,China,in2019.HeiscurrentlyworkingtowardsthePh.D.degreeininformationandcommunicationsengineering,SoutheastUniversity,China.Hisresearchinterestscurrentlyinclude,deeplearning,neuralnetworkcompression,massiveMIMO,andmachinelearningincommunications.XiangyiLi（李湘宜）receivedtheB.S.degreefromSchoolofMathematics,TianjinUniversity,Tianjin,China,in2017,andtheM.S.degreefromCentreforAppliedMathematics,TianjinUniversity,in2020.SheiscurrentlyworkingtowardthePh.D.degreeininformationandcommunicationsengineering,SoutheastUniversity,China.HermainresearchfocusesondeeplearningapplicationinwirelesscommunicationandmassiveMIMOsystems.MuhanChen（陈慕涵）receivedtheB.S.degreefromtheSchoolofInformationScienceandEngineering,SoutheastUniversity,Nanjing,China,in2019.SheiscurrentlyworkingtowardtheM.S.degreewiththeSchoolofInformationScienceandEngineering,SoutheastUniversity,Nanjing,China.Herresearchinterestscenterarounddeeplearningapplicationsinwirelesscommunicationsystems.PeiwenJiang（姜培文）receivedtheB.S.degreefromSoutheastUniversity,Nanjing,Chinain2019.HeiscurrentlyworkingtowardthePh.D.degreewiththeSchoolofInformationScienceandEngineering,SoutheastUniversity.Hisresearchinterestsincludedeeplearning-basedchannelestimationandsignaldetectionincommunications.TingtingYang（杨婷婷）receivedherB.Sc.andPh.D.degreesfromDalianMaritimeUniversity,China,in2004and2010,respectively.SheiscurrentlyaResearchProfessoratPengChengLaboratory,China.Herresearchinterestsareintheareasofmaritimewidebandcommunicationnetworks,AI-empoweredwirelesscommunications.SheservesastheAssociateEditor-in-ChiefoftheIETCommunications,aswellastheAdvisoryEditorforSpringerPlus.WeimingDuan（段为明）isnowaSeniorEngineerinPengChengLaboratory.HereceivedhisM.S.degreeincommunicationandinformationsystemfromtheUniversityofElectronicScienceandTechnologyofChinain1999.Inthesameyear,hejoinedHuaweiWirelessResearchDepartmentinShanghaiandhasworkedtherefor20years.Hehasworkedonbasebandalgorithmfor3G/4G,advancedreceiverfor4G,waveformconceptresearchfor5G,andhasalsobeendeeplyinvolvedinlow-levelalgorithmlibraryoptimizationtospeeduplargescalesystemsimulation.HaowenWang（王浩文）isaSeniorEngineerofLaboratoryofBroadbandWirelessTechnology,ShanghaiInstituteofMicrosystemandInformationTechnology,ChineseAcademyofSciences.HereceivedhisB.S.andM.S.degreesfromEEdepartmentandCollegeofSoftwareofFudanUniversity.InSIMIT,HaowenisaleaderofwirelesstechnologyRDgroup.Hehasmanyyearsofexperienceinthetestandverificationforthenewtechnologiesofwirelesscommunications.HisjobandresearchinterestsincludeRFdataacquisition,channelmeasurement,verificationandtestsolution.ShiJin（金石）[correspondingauthor]receivedhisB.S.degreeincommunicationsengineeringfromGuilinUniversityofElectronicTechnology,Guilin,China,in1996,hisM.S.degreefromNanjingUniversityofPostsandTelecommunications,Nanjing,China,in2003,andhisPh.D.degreeininformationandcommunicationsengineeringfromSoutheastUniversity,Nanjing,in2007.FromJune2007toOctober2009,hewasaResearchFellowwiththeAdastralParkResearchCampus,UniversityCollegeLondon,London,U.K.HeiscurrentlywiththeFacultyoftheNationalMobileCommunicationsResearchLaboratory,SoutheastUniversity.Hisresearchinterestsincludespacetimewirelesscommunications,randommatrixtheory,andinformationtheory.HeservesasanAssociateEditorfortheIEEETransactionsonWirelessCommunications,IEEECommunicationsLetters,andIETCommunications.Heandhiscoauthorshavebeenawardedthe2011IEEECommunicationsSocietyStephenO.RicePrizePaperAwardinthefieldofcommunicationtheoryandthe2010YoungAuthorBestPaperAwardbyIEEESignalProcessingSociety.QuanYu（于全）receivedhisB.S.degreeinradiophysicsfromNanjingUniversity,China,in1986,hisM.S.degreeinradiowavepropagationfromXidianUniversity,China,in1988,andhisPh.D.degreeinfiberopticsfromtheUniversityofLimoges,France,in1992.HeiscurrentlyaResearchProfessoratPengChengLaboratory.Hismainareasofresearchinterestarethearchitectureofwirelessnetworksandcognitiveradio.HeisanAcademicianoftheChineseAcademyofEngineeringandthefoundingEditor-in-ChiefoftheJournalofCommunicationsandInformationNetworks.版权文章，未经授权禁止转载。
详情见转载须知。
本文关键词：基于,实测,信道,的,赋能,无线通信,皇冠集团官网,反馈,作者
本文来源：皇冠集团官网-www.nursecoursefinder.com
