Research On Hybrid Precoding For Millimeter Wave Massive MIMO Systems

With the extensive development of 5G research,massive MIMO technology and millimeter wave communication technology have received extensive attention.The most prominent feature of millimeter wave communication technology is its large bandwidth,which can effectively support high-throughput wireless communication.However,compared with the microwave signal,it has a greater order of path loss than microwave signals.To solve this problem,it can be combined with massive MIMO technology,that is,a large number of antenna arrays are loaded in a small range of transceivers.To achieve the goal of high link quality and high data rate in communication systems,one potential solution is to use precoding techniques.For millimeter wave massive MIMO systems,hybrid precoding techniques are generally employed.This paper compares and analyzes the architectural differences between traditional massive MIMO systems and millimeter wave massive MIMO systems from the perspective of single-user and multi-user.Aiming at the characteristics of millimeter wave channel,the ray tracing channel model is introduced,and the traditional digital precoding algorithm is simulated and compared in the millimeter wave channel environment.For the single-user millimeter wave massive MIMO systems,this paper first introduces the OMP-based hybrid precoding,and finds that it has the problem of matrix inversion and candidate matrix.In order to solve these problems,the generalized inverse of Banachiewicz-Schur partitioned matrix is used to deduce the matrix inversion in detail,and the array response vector multiplied by the residual inner product is searched by using the characteristics of fast searching global optimal value of bird swarm algorithm.Then a hybrid MIB-OMP precoding based on BSA is proposed.Simulation results show that the proposed algorithm has improved spectral efficiency and bit error rate performance compared with OMP-based hybrid precoding without matrix inversion and candidate matrix,and its beam pattern effect is consistent with the full digital precoding.For multi-user millimeter-wave massive MIMO systems,this paper first introduces two-stage hybrid precoding based on ZF and two-stage hybrid precoding based on MMSE.Since ZF precoding only considers inter-user interference,and does not consider the influence of noise,while MMSE precoding considers both,so this paper improves on the basis of MMSE-based two-stage hybrid precoding.In order to solve the codebook problems of these two algorithms,the KSVD dictionary learning method is introduced,and the original codebook is updated by KSVD.From the two stages of sparse coding and dictionary updating,a hybrid precoding of MMSE based on KSVD is proposed.The simulation results show that compared with the two-stage hybrid precoding based on ZF and the two-stage hybrid precoding based on MMSE,the proposed algorithm has significantly improved the system spectral efficiency and bit error rate performance,and its beam pattern effect and fully digital MMSE precoding is relatively close.