Downlink Channel Estimation And Pilot Optimization Based On Compressive Sensing For Massive MIMO Systems

Massive multi-input multi-output(MIMO)is considered to be one of the key technologies for future wireless communication system.Multiple-antenna systems allow for a high spatial multiplexing gain and a high array gain.In order to obtain the gains of massive MIMO,the accurate channel state information(CSI)should be known.However,in a frequency-division duplexing(FDD)MIMO system,the CSI is difficult to be acquired.Since the pilot overhead increases linearly with the number of transmit antennas.Conventional approaches of downlink channel estimation,such as minimal mean square error(MMSE)and least squares(LS),cannot deal with massive MIMO system effectively.Because the number of pilots is proportional to the number of transmit antennas.In a massive MIMO system,with the increasing of the number of transmit antennas,limited local scatterers at the base station bring about the user channel matrices tending to be sparse.The massive MIMO channel estimation can be transformed to a problem of reconstructing sparse signals by compressive sensing(CS)theory,and then the problem of channel estimation can be solved by CS reconstruction algorithm.Particularly,minimizing the value of mutual coherence of sensing matrix in CS framework can be used as a criterion of optimizing pilot sequences.This paper focuses on the channel estimation and pilot optimization algorithms based on CS framework for FDD massive MIMO systems.According to the criterion of minimizing the value of mutual coherence of the measurement matrix,two kinds of pilot optimization methods are proposed.(1)Pilot optimization method based on random Gaussian matrix: an iterative algorithm is employed to obtain the optimized pilot symbols from a random Gaussian matrix.(2)Pilot optimization method based on partial Fourier matrix: several rows of Fourier matrices are selected as pilot symbols by random search algorithm,which leads to that the measurement matrix in CS framework has a smaller value of mutual coherence.Simulation results show that,the reconstruction error of the sparse signal can be effectively reduced by decreasing the value of mutual coherence of the measurement matrix.In the noisy case,the optimal pilot sequences obtained by the two pilot optimization methods proposed in this paper can effectively decrease the mean square error of the channel estimation of Massive MIMO system by 1 ~ 3dB,as compared with the unoptimized pilot sequences.In addition,the performance gain of the optimized pilot sequences obtained by the partial Fourier matrix is better than that of the optimized pilot sequences based on the random Gaussian matrix.