Research On Channel State Information Acquisition Methods For FDD Massive MIMO System Based On Compressed Sensing

Massive MIMO has become one of the core technologies of the current and future wireless communication because it can effectively enhance the channel capacity and spectrum efficiency of mobile communication systems and realize green communication.In order to effectively utilize the advantages of massive MIMO technology,accurate acquisition of channel state information is essential.Based on the sparsity and compressed sensing technology,the acquisition methods of massive MIMO downlink channel state information for the most widely used FDD system are studied in this thesis.Firstly,the development history and prospect of mobile communication technology are briefly described.The evolution and advantages of MIMO technology are introduced.Previous studies and experiments of the acquisition of downlink channel state information for massive MIMO systems are summarized.Then,the sparsity characteristics of FDD massive MIMO channels in the virtual angular domain and delay domain are analyzed.For the angular domain,the performances of various angular domain transformation methods are compared.Based on the OMP-GSO algorithm,an algorithm called OMP-GSOAdp is proposed,which adaptively terminates iterations and obtains almost the same performance as OMPGSO with fewer computation iterations.Meanwhile,an over-complete transformation matrix based on dictionary learning is also proposed.Simulations show that compared to the widely used DFT matrix,the proposed OMP-GSO-Adp method and the optimized transformation matrix based on dictionary learning achieve better sparse representation performance for ULA antenna arrays.For the delay domain,the corresponding massive MIMO channel model is presented.The sparsity of the channel impulse response in the delay domain is expounded.The spatial-temporal common sparsity of delay-domain MIMO channels are discussed.Next,the compressed sensing based downlink channel estimation scheme based on the sparsity of delaydomain is studied.A non-orthogonal pilot design which selects sub-carriers with equal intervals is analyzed.Meanwhile,under the assumption of spatial-temporal common sparsity,the adaptive structured subspace pursuit(ASSP)downlink channel estimation algorithm is studied.In addition,under the assumption of partial spatial-temporal common sparsity,an algorithm called partial adaptive structured subspace pursuit(PASSP)is proposed,which utilizes the estimation results of previous antenna groups as the prior knowledge of sparsity for current antenna group.Simulations demonstrate that compared to conventional compressed sensing method,ASSP can obtain better estimation performance with less pilot overhead.Meanwhile,compared with ASSP,the proposed PASSP method can adapt to the imperfect structured sparsity pattern,and obtain better estimation performances.Finally,the downlink channel estimation scheme based on the sparsity of the angular domain is studied.For the uplink,the channel estimation is formulated into a sparsity recovery problem by exploiting the sparsity of the angular domain.Besides,a method is proposed to optimize the correlation of columns in the pilot matrix.The uplink channel can be recovered even with non-orthogonal pilot sequences.For the downlink,a channel state information acquisition scheme based on angular reciprocity is analyzed,and maximum-reuse pilot design and minimum-interference pilot design are given.In addition,in view of the mismatch between the uplink and downlink angular domain support sets,the modification scheme for the support set is proposed with partial orthogonal pilots or non-orthogonal pilots.Meanwhile,a compressed channel state information feedback scheme is presented,which only quantizes a small amount of components via RVQ due to the sparsity in angular domain.The simulation results show that compared with the LS estimation based on orthogonal pilots,the analyzed downlink channel information acquisition scheme can obtain good channel estimation performance while reducing the pilot overhead.The proposed support set modification scheme improves estimation performance under medium and high SNR without the increase of feedback overhead.Compared with conventional channel quantization feedback scheme,the proposed feedback scheme significantly reduces the number of required feedback bits.