Sparsity Adaptive SWOMP Channel Estimation Algorithm For Massive MIMO Systems

Massive multi-input multi-output(MIMO)technology,which is adopted as the key technology of 5G has attracted the attention of researchers.In order to fully utilize the spatial multiplexing capability and array gain provided by massive MIMO,the base station(BS)needs to obtain accurate channel state information(CSI).When the BS is equipped with massive antennas,the pilot overhead required by conventional channel estimation methods becomes overwhelming,leading to unacceptable loss of spectrum efficiency.Extensive experimental studies have shown that the angular spread on the BS side is limited due to the fewer scatterers around BS.Hence,the channel matrix shows sparsity structure in the angular domain.In this dissertation,by firstly analyzing and proving the sparsity of the channel in the angular domain,two distributed compressed sensing based sparse channel estimation algorithms are proposed which exploit the common sparsity of the channel to extend the conventional stagewise weak orthogonal matching pursuit(SWOMP)algorithm from a single measurement vector(SMV)problem to a multiple measurement vector(MMV)problem.With the prior channel sparsity level information,the so called distributed SWOMP algorithm reselects the larger coefficients of the channel matrix,which can decrease the overestimation problem of the conventional SWOMP algorithm.Considering the sparsity level information is hard to be acquired accurately in practice,a novel termination condition is designed and incorporated into the distributed SWOMP algorithm to develop a distributed sparsity adaptive SWOMP(SA-SWOMP)algorithm,which can adaptively adjust the iteration numbers for different sparsity levels.In addition,a pilot matrix design via particle swarm optimization(PSO)algorithm is provided which is well adapted for the distributed compressed sensing based channel estimation algorithms and improves their performance further.Simulation results show good performance of the proposed algorithms in terms of convergence speed,normalized steady state mean square error(NMSE),pilot overhead demand and indicate reduced complexity and robustness against different channel sparsity levels and signal-to-noise ratio(SNR)values.