Research On Channel Estimation For Large Scale MIMO Systems

With the commercialization of fourth generation(4G)communication network,more attentions are paid to the research of fifth generation(5 G)communication network targeting the various use scenarios in 2020 and beyond.In order to meet the needs of user experience of 100Mbps in 5G continuous wide area coverage network,large-scale Multiple Input Multiple Output(MIMO)technology as one of the key technologies of 5G has outstanding performance in system capacity,spectrum efficiency and link reliability under perfect channel state information(CSI).However,in practice,the precise CSI is unknown,in order to obtain relatively precise CSI for diversity combination,coherent detection and other processes,so this thesis focuses on the channel estimation technique for large-scale MIMO systems,our research and innovation are as follows:First,the current research status of large-scale MIMO technology,State-of-The-Art 3D(Three Dimensional)/Massive MIMO technology from standardization organizations and the study on massive MIMO channel estimation are summarized.Second,as foundation of large-scale MIMO channel estimation,the traditional pilot aided channel estimation(PACE)method for 3D MIMO systems is discussed,which extends the 2D pilot pattern to 3D pilot pattern by utilizing space correlation and multiple dimension sampling in order to reduce the pilot overhead.Moreover,for the channel recovery,simulation results demonstrate that winner filter based interpolation algorithm can obtain a smaller mean square error(MSE)at some expense of algorithm complexity compared with three-dimensional linear interpolation algorithm.However,the pilot constraints of traditional uniform pattern are limited to multi-dimensional sampling theorem,and the pilot overhead has a theoretical lower bound.So we introduce the compressed sensing(CS)technique into 3D pilot aided channel estimation to reduce pilot overhead in 3D MIMO systems.Moreover,a random search method based non-uniform pilot allocation algorithm with low computational complexity is proposed to further improve the CS performance.Simulation results demonstrate that compared with the traditional uniform pilot for 3D PACE,the average gain of the proposed non-uniform pilot for CS-based channel estimation can get the improvement of 3.58dB relative to the same pilot overhead.In massive MIMO FDD(Frequency Division Duplex)systems where reciprocal channel does not exist,the pilot overhead for the downlink traning becomes intolerable as the pilot overhead increases with the number of base station antennas.In order to reduce pilot overhead in FDD systems,a spatial correlated channel is firstly modeled in this thesis,the channel can be represented in a sparse form in spatial-frequency domain.Then,the CS theory is applied to develop an effective mechanism for channel estimation.Moreover,based on the inherent common sparsity in the user channel matrices,this thesis proposes an improved orthogonal matching pursuit(OMP)algorithm to reduce the pilot overhead and improve the channel estimation accuracy.Simulation results demonstrate that the proposed algorithm can significantly reduce the pilot overhead and increase the channel estimation accuracy.Finally,a summary of the full text is made and future research directions are discussed.