Research On Massive MIMO Channel Estimation Technologies Based On Compressive Sensing

Massive multiple input multiple output(MIMO)is the key technology of the 5th mobile communication networks.Compared with traditional MIMO,the massive MIMO has significant superiorities in energy efficiency,frequency efficiency and stability.However,the complex channel estimation seriously influences the efficiency of massive MIMO systems,since the complexity of channel estimation increases with the number of antennas.This dissertation studies new techniques for channel estimation and pilot allocation of massive MIMO system,based on the compressive sensing technology.The main content of this dissertation is summarized as follows:(1)A new antenna sharing path delay estimation method is proposed to improve the accuracy of sparse channel estimation.Firstly,based on the observation that the path delays of BS antenna are closely similar,the base station(BS)antennas share the path delay information with each other,to obtain the consistent estimation of path delays.After that,for the massive MIMO system(where the number of BS antenna is more than 50)that the path delays in the antenna array may be different,an antenna division strategy is proposed to ensure that the path delays in one small antenna zone are similar,and then the antenna sharing strategy is utilized in each antenna zone to improve the accuracy of path delay estimation by calculating the consistent estimation of path delays.Finally,in order to further improve the accuracy of path gain estimation,a decision aided path gain estimation method is proposed.In this method,some of the detected data are utilized as the pilots to estimate the channel.Simulation results show that the proposed algorithm can provide more than 10dB SNR gain over the traditional subspace pursuit(SP)method(2)To improve the accuracy of greedy algorithm,an improved Homotopy algorithm is proposed.Firstly,we weight the traditional Homotopy method,substitute the regularization parameter by the weights,and adjust the weight adaptively according to the channel coefficients,to improve the convergence rate of the algorithm.After that,in order to improve the accuracy of the proposed algorithm,we share the weight information within the antenna array and calculate the consistent estimation of the weight during each iteration.In order to solve the problem that ordinary Homotopy method can not converge in the complex domain,we extend the conventional weighted Homotopy to the complex field,and prove its convergence by theory analysis.Finally,we simplify the calculation of step size,and hence reduce the algorithm complexity.Simulation results show that the proposed algorithm can provide more than 6 dB SNR gain over traditional greedy algorithm.(3)In order to reduce the number of required pilots,a new auxiliary information based channel estimation algorithm is proposed for the time division duplexing(TDD)massive MIMO system.Inspired by the observation that the path delays of adjacent OFDM symbol are closely similar.the estimated uplink path delay is utilized as auxiliary information to improve the accuracy of downlink path delay estimation.For the scenario that the path delays in one uplink-downlink process may be different in fast time-varying channel,a probability weighted channel estimation algorithm is proposed based on the uplink path delay and variation of path delays.In this algorithm,we firstly estimate the probabilities that the elements in the current channel impulse response(CIR)are nonzero.Then,we efficiently prepress the downlink channel estimation according to these probability information.Simulation results show that the proposed method can reduce the pilot overhead by nearly 20%over the traditional methods.(4)In order to improve the channel estimation performance in massive MIMO OFDM system,.one efficient pilot design scheme is proposed.Firstly,by utilizing the block structure in massive MIMO channel,the pilot optimization problem is transformed into a optimization problem by minimizing the block consistent value of pilot matrix,based on the block coherence(BC)in compressive sensing.After that,the simultaneous perturbation stochastic approximation(SPSA)method and cross-entropy optimization(CEO)method are proposed to solve the optimization problem.The SPSA method estimates the gradient value of the objective function,and then optimizes the pilot pattern by correcting the search direction in each iteration.The CEO method obtains the optimal pilot pattern by estimating the probability density function of the minimum objective function.Simulation results show that both of the schemes outperform the equispaced pilot pattern,exhaustive pilot pattern and random pilot pattern in convergence speed and channel estimation performance.Specifically,comparing with the traditional schemes,SPSA and CEO can provide over 5dB SNR gain under the same MSE.