Research On Hybrid Precoding Technology Of Millimeter Wave Massive MIMO System

The development of wireless communication technology entered the fifth generation(5G)of mobile communication moving on the direction of Beyond 5G to meet the increasing requirement for high data rates and high-quality services.The 5G of mobile communication technology can greatly increase the communication rate to gigabit per second,which significantly reduces latency to the millisecond level and supports for a large number of the increasing devices.The wide bandwidth of millimeter wave and massive MIMO(Multiple Input Multiple Output,MIMO)are the key technologies for the 5G communication systems to increase communication capacity.The large number of antennas in base stations makes the traditional full-digital precoding processing technology hardware cost and energy consumption too high to be feasible.Hybrid precoding processing technology becomes the research trend of the 5G wireless communication system,this thesis mainly studies the hybrid precoding technology in the wireless communication of millimeter wave system combined with massive MIMO that compromised the complexity and spectrum efficiency.Firstly,for a single-cell downlink system where a base station serves multiple users,a hybrid precoding based on machine learning and adaptive connection 1-bit quantization is proposed.The analog precoder at the base station has an adaptive connection structure.We firstly determine the connection form of the RF(Radio Frequency,RF)chain and antenna in the analog precoding according to the channel state information of each user,and then the machine learning adaptive crossentropy is optimized to obtain the analog precoding matrix for the best reachable rate,and finally the ZF(Zero Forcing,ZF)method is applied to design digital precoding to eliminate interference between data streams.Simulation shows that the proposed ACN based MLACE(Adaptive Connection Network based Machine Learning Adaptive Cross-Entropy,ACN based MLACE)scheme can achieve good available rate with the same low hardware complexity and achieve a trade-off between hardware complexity and available rate.Secondly,the joint optimization method is used for the situation where the base station serves a single user with full connection.The problem of maximum system capacity is firstly transformed into the problem of maximum determinant of equivalent channel matrix.Then three hybrid precoding schemes based on joint optimization are proposed respectively.The first scheme is based on the principle of maximum beamforming gain,which is selected from the DFT(Discrete Fourier Transform,DFT)codebook to form the analog precoding matrix at the transmitter and the analog combination matrix at the receiver.According to the maximization of the equivalent channel matrix determinant,it can be further transformed into two parts: the contribution of each RF chain and the contribution of other RF chains.The joint single-user and full-connection optimization algorithms of JLIO-SVD(Joint List Iteration Optimization based SVD,JLIO-SVD)and JLIO-GA(Joint List Iteration Optimization based Gradiend Ascent,JLIO-GA)are respectively proposed,which can show excellent spectrum efficiency.The JLIO-GA achieves a better trade-off between computational time complexity and spectrum efficiency.And it can be used as an effective hybrid precoding algorithm for single-user millimeter-wave massive MIMO systems.Furthermore,The ACN-SVD(Adaptive Connection based SVD,ACN-SVD)alternation optimization algorithm is proposed for the situation that the base station serves for a single-user adaptive connection.We firstly design an adaptive connection structure for the analog precoding based on the principle of maximizing the signal power at the transmitting antenna,and determine it according to the unitary matrix of the SVD(Singular Value Decomposition,SVD)of the channel matrix.Then the analog precoding is designed based on the minimum Frobenius distance between the optimal full-digital precoding and the hybrid precoding,and the digital precoding is designed with the SVD of the optimal precoding conjugate transpose and the product of the analog precoding.The simulation shows that the ACN-SVD algorithm can achieve better spectrum efficiency with lower computational complexity.