Sparse Channel Estimation And Hybrid Precoding In Massive MIMO Systems

As one of the key technologies of 5G mobile communication,massive multipleinput multiple-output(MIMO)has the advantages of high spectral efficiency,low energy consumption and anti-interference.Due to the short wavelength of millimeter wave(mmWave),a large number of antennas can be deployed in a small physical space.Therefore,mmWave technology can be effectively combined with massive MIMO technology and widely used.The introduction of large-scale antennas and mmWave poses new challenges for channel estimation and precoding,such as excessive unknown channel parameters leading to huge pilot overhead,severe propagation loss resulting in low accuracy of channel estimation,and traditional precoding algorithms being not efficient enough.Focusing on the above problems,this dissertation studies the sparse channel estimation and hybrid precoding techniques of the massive MIMO system.The main research results are as follows.For the problem of large pilot overhead,the sparse channel estimation of the massive MIMO system is studied.This paper exploits the characteristics of the wireless channel in time-space-delay domain.In the time domain,the complex exponential basis expansion model(BEM)is used to capture the time-variation of the channel,which can reduce the number of parameters to be estimated.In the spatial domain,we propose the generalized spatial basis expansion model(G-SBEM)to model the spatial correlation of channels,which constructs redundant basis functions to reduce the angular energy leakage.In the delay domain,we utilize the sparsity of wideband wireless channels,and formulate the channel estimation as a compressive sensing(CS)problem,with a quasi-block-sparse matrix to be recovered.For this problem,we make the theoretical analysis of the characteristics of the measurement matrix,and decouple the design of the pilot values and pilot positions.Two greedy reconstruction algorithms are proposed to recover the channel parameters.Quasi-block simultaneous orthogonal matching pursuit(QBSO)algorithm divides seeking nonzero entry positions and calculating channel parameters into two phases.Based on QBSO algorithm,angle spread based QBSO(AS-QBSO)algorithm uses a priori angle spread information to increase the angle resolution.Simulation results show that the AS-QBSO algorithm can reduce the pilot overhead of 9.8% compared with the traditional least squares(LS)scheme.For the problem of low channel estimation accuracy,the precoding-assisted channel estimation for the mmWave MIMO system is studied.The channel estimation is decomposed into two phases: angle estimation and path gain estimation.For angle estimation,using the mmWave channel sparsity in the angle domain,combined with discrete prolate spheroidal BEM(DPS-BEM),we formulate the angle estimation as a block sparse signal recovery problem.Then,a new greedy reconstruction algorithm is proposed to adaptively reduce the angle regions and increase the angular resolution.For the path gain estimation,based on the estimated angle parameters,a heuristic algorithm is used to separately design the pilot analog precoding and digital precoding,aiming to maximize the received power of each distinguishable path.The simulation results show that compared with the traditional block-based orthogonal matching pursuit(BOMP)scheme,the proposed scheme can obtain a signal-to-noise ratio(SNR)gain of more than 4 dB.For the problem of low efficiency of the precoding algorithm,adaptive hybrid precoding and channel tracking schemes are studied.For the traditional adaptively-connected structure,we exploit the sparsity of the analog precoding matrix and propose the compressive sensing based adaptive hybrid precoding(CS-AHP)algorithm to improve the spectral efficiency.Based on the traditional adaptively-connected structure,we design a low-constrained adaptivelyconnected structure,which can obtain higher freedom than the conventional structure,thereby obtaining a higher achievable sum rate.For the low-constrained adaptively-connected structure,in order to provide the CSI required for precoding design,a hybrid structure channel tracking scheme is proposed.The pilot precoding matrix is obtained by angular rotation,and the channel parameters are tracked based on extended kalman filter(EKF).The simulation results show that compared with the traditional adaptive hybrid precoding algorithm,the proposed CS-AHP algorithm can obtain the achievable rate gain of 1.5 bps/Hz.Compared with the traditional channel tracking scheme,the proposed tracking scheme can reduce the pilot overhead of 11%.In summary,this paper studies the sparse channel estimation and hybrid precoding design for the massive MIMO system,and proposes CS based sparse signal reconstruction methods.The proposed algorithms effectively reduce the system pilot overhead,and improve channel estimation accuracy and system transmission rate.The methods in this paper have great significance for the theoretical research and technical realization of wireless communication.