Research On CSIT Acquisition In Massive MIMO System Under FDD Mode

The fifth-generation mobile communication system(5G)can enhance 4G in all aspects,so that it has become one of the hottest researches in communications field nowadays.Massive Multiple-Input Multiple-Output(Massive MIMO)technology means to configure a large number of antennas at the transmitter and the receiver,with the purpose of providing services to multiple users on the same time-frequency resource simultaneously.As one of the key technologies of 5G,massive MIMO has a lot of advantages,such as large system capacity,high spectrum efficiency,and so on.However,the key of massive MIMO being able to exploit these advantages is that the transmitter must obtain downlink channel state information(CSI).In the frequency division duplex(FDD)massive MIMO systems,the transmitter acquires the CSI at transmitter(CSIT)through the feedback of the receiver because of the lack of channel reciprocity.Nevertheless,as the number of antennas increases dramatically,the amount of feedback from CSIT increases,which causes the poor performance of massive MOMO systems.Therefore,this thesis focuses on the issue of CSIT acquisition in massive MIMO systems under FDD mode,and aims to reduce the feedback overhead as well as ensure the accuracy of the CSIT.The main contributions are as follows:(1)The massive MIMO channel model is analyzed,and the compressed sensing(CS)technology is described in detail.On these bases,a CSIT acquisition method based on spatial sparsity is proposed for MIMO systems.Also,to improve the traditional sparse adaptive matching pursuit(SAMP)which is widely used,a novel SAMP method is put forward.This method could reconstruct signal quickly and accurately by the advantages of segmentation idea,initial sparsity estimation and variable step size.The numerical results prove the superiority of the proposed method both in reconstruction accuracy and computation time.Moreover,compared with CSI acquisition methods based on orthogonal matching pursuit(OMP),Subspace Pursuit(SP)or SAMP,simulation results demonstrate that the proposed method improves the normalized mean squared error(NMSE)performance effectively in massive MIMO systems.(2)In massive MIMO systems,the channel shows a strong spatial correlation due to the deployment of large-scale antenna arrays and the small distance between antennas.Therefore,two compressed feedback methods by taking advantage of spatial correlation property are proposed.Firstly,a CSI compression feedback method based on low complexity principal component analysis(PCA)is proposed to solve the problem of high computational complexity of the Karhunen-Loeve transform(KLT)for the unknown transient correlation matrix of the base station.Secondly,on the issues of channel data with complex nonlinear structure can not be handled perfectly by PCA,a CSI compression feedback method with Laplacian Eigenmaps(LE)nonlinear processing is proposed.Theoretical analysis and simulation results show that,compared with discrete cosine transform(DCT)sparse compression method,the proposed two methods can reduce the amount of feedback while maintaining a certain accuracy,and have lower bit error rate(BER).In addition,compared with the KLT method,both proposed methods have lower computational complexity and shorter average running time.