Research On Beamforming In Millimeter Wave Massive MIMO Systems

Millimeter wave communication and Massive MIMO are the key technologies of 5G,which can effectively improve the spectrum efficiency and energy efficiency of the systems.As is known to all,beamforming plays an important role in millimeter wave Massive MIMO systems,while the high complexity brought by the surge in antennas number in the millimeter wave Massive MIMO systems prevents the application of traditional beamforming algorithms.To this end,the dissertation focuses on the research of beamforming technology in millimeter wave Massive MIMO systems.The main work and contributions of this dissertation are listed as follows.(1)To improve the performance of millimeter wave Massive MIMO systems,we take the phase of elements in the analog matrix as variables to design the analog matrix,and propose hybrid precoding algorithms based on decoupling the transceiver for hybrid analog and digital structure.For the fully-connected structure,the particle swarm optimization algorithm is adopted to solve the phase of the elements in the analog precoding matrix column by column,and the least-squares method is used to design the digital precoding matrix.Then,the analog precoding matrix and digital precoding matrix are optimized iteratively to further improve the performance of the algorithm;For the partially-connected structure,based on the characteristic that each column of the analog precoding matrix has many zero elements,the objective function of hybrid precoding is transformed into a series of independent sub-problems,where each sub-problem minimizes the Euclidean distance between a sub-matrix of the optimal precoding matrix and the product of a sub-matrix of the analog precoding matrix and one row of the digital precoding.The particle swarm optimization algorithm and the least-squares method are used to solve the sub-problem iteratively.Theoretical analysis shows that the two proposed algorithms share similar complexity of classic algorithms.Simulation results highlight the superior spectral efficiency of the proposed algorithm in the fully-connected structure case,and show that the spectral efficiency of the proposed algorithm approaches the theoretical upper limit in the partially-connected structure case.(2)To bypass the SVD operation of the channel matrix of antenna dimensions and to reduce the complexity of the hybrid precoding algorithm,three hybrid precoding algorithms based on the joint of transceiver are proposed for the hybrid analog and digital structure.The concept of equivalent channel is introduced,and the SVD operation of the low dimensional equivalent channel matrix is used to jointly design the digital part of the transceiver.It is derived that the optimization problem of the analog part of the transceiver is equivalent to maximizing the equivalent channel gain.Then three schemes are proposed to solve the analog part.The first scheme utilizes the idea of successive interference cancellation to transform the design problem of the analog part into a column-by-column optimization problem,and uses the coordinate descent method to solve the problem,and the design of analog part through iteration.In order to further reduce the complexity,the second scheme considers the analog precoding matrix as the optimization unit,and completes the design of the analog part by iteratively performing SVD operation of the equivalent channel matrix and phase extraction.Finally,the codebook based third scheme is proposed for reducing the feedback of the systems.This scheme completes the design of the analog part at one time by maximizing the equivalent channel gain.Simulation results show that the proposed algorithms can achieve spectrum efficiency similar to the classical algorithm with lower complexity.(3)To avoid obtaining channel information in advance and to further reduce the complexity of the millimeter wave Massive MIMO systems,for the analog structure,we study the fast beam search scheme for line of sight(LoS)and non line of sight(NLoS)scenarios when the channel state information is unknown.Firstly,to understand the essential difference of existing search schemes,we introduce the concepts of state utilization efficiency,information representation efficiency,and feedback efficiency.These concepts are further used to analyze the existing schemes,and it is concluded that the higher the state utilization efficiency of the search program is,the lower the complexity is.According to this conclusion,it is deduced that using Gray mapping to encode the subinterval number of each stage can make the state utilization efficiency reach 100%and the search success ratio is the highest.Then,the multi-layer beam search scheme based on beam overlapping is proposed according to this coding scheme.Through analysis,the state utilization efficiency,information representation efficiency,and feedback efficiency of the proposed scheme are both 100%,and the search complexity is only 50%of the binary search.Secondly,to make the scheme universal,the beam search scheme in the NLoS scenario are also studied.Through analysis,it is found that the NLoS scenario has the problem of misjudgment caused by multipath compared with the LoS scenario,and then a beam search scheme suitable for NLoS scenario is proposed.