Research And Application Of Compressed Sensing Technology In FDD Massive MIMO Systems

As one of the key technologies of 5G mobile communication systems,massive MIMO can greatly improve the performance of the systems.Equipped with more antennas in both sides,channel estimation in massive MIMO systems requires a large number of pilots and feedback resources.Therefore,it is significant to research how to reduce the overhead of channel estimation in massive MIMO systems.This thesis mainly addresses FDD massive MIMO systems,focusing on compressed sensing-based CSIT estimation technology.Firstly,the basic theory and the application of compressed sensing technology in massive MIMO channel estimation are summarized.The sparsity decomposition,measurement matrix and reconstruction algorithms of compressed sensing are studied.After that,the sparse model and corresponding reconstruction algorithms are analyzed and compared detailedly.The sparse characteristics of massive MIMO channel are also analyzed.The main work of this thesis is focused on the compressed sensing based channel estimation of the multi-user Massive MIMO systems.Secondly,CSIT estimation and feedback with temporal correlation are investigated.The sparse characteristics of temporal correlation channel are analyzed.The prior channel information can be exploited to estimate CSIT in current slot.In addition,in order to deal with potential model mismatch,AMSP algorithm is proposed in the thesis.Simulation results show that the proposed AMSP algorithm has good robustness against model mismatch while maintaining good NMSE performance.Thirdly,the distributed compressed CSIT estimation technology in massive MIMO channel is also studied.The channel sparsity characteristics under multi-user FDD massive MIMO system are analyzed,including the sparse structure of the channel of single user and the joint sparse structure among multiple users.Closed-loop pilot and CSIT feedback resource adaptation framework are studied,and an improved ESA control algorithm and a RCMF algorithm are proposed based on SA control algorithm.Simulation results show that the JOMP algorithm has good NMSE performance even with short pilot length,and the JOMP-ESA algorithm and the JOMP-Rand algorithm both have good convergence performance.Finally,this thesis study the compressed channel estimation scheme based on user grouping technology.User grouping is realized by using JSDM method,and the similarity measurement methods and clustering methods for user grouping are studied.The joint sparse characteristics of user channels in the same group after grouping are studied in detail,and the overhead required for CSIT estimation is reduced by using compressed sensing.The simulation results show that users can be divided into several groups effectively according to users' second order channel statistics information.And the overhead required for CSIT estimation can be obviously reduced by using the joint sparse characteristics of user channels in the same group.