Research On Channel Estimation For Millimeter Wave MIMO Systems Based On Compressed Sensing

With the rapid development of wireless communication technology,people have put forward the demand of higher speed and lower delay for communication performance,which makes the combination of millimeter wave and large-scale MIMO technology become the trend of current mobile communication development.The huge number of antennas,complex hybrid beamforming structure and high path loss in millimeter wave large-scale MIMO systems make the traditional channel estimation scheme no longer applicable.The sparse reconstruction algorithm of compressed sensing provides theoretical support for the restoration of large-scale millimeter wave channel matrix.Therefore,based on the channel characteristics of millimeter wave and the technical idea of compressed sensing,this thesis proposes millimeter wave channel estimation schemes in different scenarios.Aiming at the problem of misjudgment of K-beam search scheme in non line of sight(NLoS)scenarios,an improved proposed search scheme based on layered search(PSSLS)is designed,and a channel estimation algorithm based on PSSLS is proposed.On the basis of K-beam search,PSSLS adds additional judgment of relative power threshold(RPT)when determining the angle search range of the next phase in each phase,which improves the power of searching the global optimal communication path in NLoS scenario as much as possible.The simulation results show that compared with the original K-beam search scheme,the probability of PSSLS to obtain the global optimal communication path in NLoS scenario is increased by about 20%.Aiming at the bottleneck of slow convergence speed and low estimation accuracy in channel estimation based on sparsity and low rank of millimeter wave channel,the traditional alternating direction method of multipliers(ADMM)is improved,and a millimeter wave channel estimation algorithm based on ADMM with adaptive penalty(ADMMAP)is designed in this thesis.Firstly,the received signal matrix Y and the virtual channel gain matrix HV are regarded as two optimization objectives,and the channel estimation problem is transformed into a two objective global optimization problem by combining the sparsity and low rank of the channel.Then,the solution of the problem is divided into multiple iterations by ADMMAP,and the penalty coefficients ? used in each iteration are adaptively changed according to the set rules.Simulation results show that compared with ADMM channel estimation algorithm,the convergence speed of ADMMAP channel estimation algorithm is improved by about 54%,and the normalized mean square error of channel estimation result and actual channel matrix is reduced by about 0.04dB.