The Research Of Massive MIMO Nonlinear Precoding Technology

As one of the key technologies of 5G system,massive multiple input multiple output(massive MIMO)systems,utilizes massive transmitting antennas to simultaneously transmit multiple data streams to achieve greater spatial diversity gain compared with traditional MIMO systems.With channel status information acquired at the base station,precoding technology eliminates interference between data streams through preprocessing algorithms to ensure reliable data transmission and reduce detection complexity at the receiver.Therefore,precoding technology in massive MIMO system has important research significance.Generally,precoding technology is divided into linear precoding and nonlinear precoding.In order to obtain higher performance gain,this paper focuses on nonlinear precoding technology in massive MIMO system.The main contents of this thesis are as follows:Firstly,in this thesis,single-user and multi-user massive-MIMO systems are studied,then the system modeling for single-user and multi-user massive-MIMO systems are established,where the effects of inter-stream interference and inter-user interference in massive-MIMO system are analyzed.Then two typical nonlinear precoding as tomlinsonHarashima precoding(THP)and vector perturbation(VP)precoding are simulated and analyzed in massive MIMO system.Subsequently,this thesis improves the conventional THP algorithm in massiveMIMO system.Considering the performance of QR decomposition assisted traditional THP algorithm is limited by the worst sub-channel,this thesis proposes the geometric mean decomposition(GMD)assisted THP pre-coding algorithm,termed as GMD-THP.The proposed scheme makes the gain of all sub-channels equally through GMD decomposition,thus enhancing the performance of the worst sub-channel and greatly improving the total system performance.Besides,considering that the traditional THP algorithm cannot achieve multi-user massive-MIMO transmission,this paper extends the proposed GMD-THP single-user massive-MIMO system to the multi-user massiveMIMO system with the proposed BD-GMD-THP algorithm,which is based on block diagnalization(BD)criterion to achieve multi-user transmission with THP nonlinear precoding assisted massive-MIMO system.Afterwards,the constructive interference(CI)assisted zero forcing(ZF)nonlinear precoding method is investigated in this paper.The CI precoding method transforms the harmful interference signals into beneficial signals by placing them in the right place of the constellation via jointly precoding with the transmit data,which greatly improves the performance of the traditional linear ZF precoding system.Moreover,this method does not need to do the modulus operation at the transceiver,which simplifies the corresponding massive-MIMO system.In this thesis,based on the investigation of strict phase rotation based strict-ZF-CI and non-strict phase rotation based ZF-CI precoding algorithms for MPSK modulation,a non-strict-ZF-CI precoding method for QAM modulation is proposed in this thesis to achieve high order modulation for ZF-CI.Finally,in order to reduce the complexity of the ZF-CI precoding algorithm,this thesis tries to solve the disturbance problem of ZF-CI with low complexity,where a fast disturbance solution based on dual neural network(DNN)and conjugate gradient(CG)descent algorithms are considered.The minimum mean square error(MMSE)assisted non-strict-MMSE-CI and block diagonalization assisted non-strict-BD-CI precoding algorithms are further proposed to improve the bit error rate(BER)performance of constructive interference precoding algorithm.