Font Size: a A A

Research On Massive MIMO Channel Modeling And Channel Estimation

Posted on:2022-07-09Degree:DoctorType:Dissertation
Country:ChinaCandidate:J Q ChenFull Text:PDF
GTID:1488306350988549Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
This paper focuses on one of the key technologies in the physical layer of the fifth generation(5G)mobile communication,namely massive multiple-input multiple-output(MIMO)channel modeling and channel estimation technology.By deploying large-scale antenna array on the base station side,the advantages of high-frequency spectral efficiency and high-energy efficiency of massive MIMO system have been widely recognized by both the academia and industry.The transmission rate and quality of wireless communication system are directly affected by wireless propagation environments.Therefore,the research on wireless channels is the basis of massive MIMO system implementation and performance evaluation.Channel modeling is used to describe the propagation mechanism of wireless signal with a series of parameters on the basis of full understanding of wireless electromagnetic wave,which is considered as one of the most effective methods to evaluate the performance of massive MIMO technology.By obtaining accurate channel state information(CSI),channel estimation can guarantee the implementation of channel equalization,precoding,beamforming and resource allocation in massive MIMO systems.However,the introduction of large-scale antenna array brings new challenges to channel modeling and estimation of massive MIMO systems.In the aspect of channel modeling,the new characteristics of massive MIMO channel need to be deeply investigated when channel modeling.In addition,the introduction of massive MIMO technology to some other 5G application scenarios(such as vehicle to vehicle(V2V)street environments)will aggravate the complexity of channel scene description.In the aspect of channel estimation,the pilot overhead of massive MIMO channel estimation increases sharply,which has become the bottleneck of effective information transmission and performance improvement.At the same time,the research of high-precision channel estimation algorithms has always been one of the important goals of massive MIMO channel estimation.This paper focuses on the key issues of massive MIMO channel modeling and channel estimation,and the main research contents and contributions are as follows:(1)A three-dimensional(3D)confocal ellipsoid channel model for massive MIMO system with employing a uniform planar antenna-array(UPA)is proposed.The main propagation characteristics of massive MIMO channels are discussed,including the elevation characteristics,near-field effect and spatial non-stationarity.Firstly,under spherical wavefront assumption,a 3D confocal-ellipsoid channel model is established to describe the near-field effect,and the channel impulse responses(CIRs)are derived,including line-of-sight(LoS)components and non-LoS(NLoS)components.Secondly,the birth-death visible-region mixed method(BVMM)is proposed to characterize the non-stationary properties of clusters in both array and time axes.Finally,the statistics of massive MIMO channels are studied through spatial-temporal correlation functions(STCFs)and Doppler shift standard deviation.(2)A massive MIMO channel model based on irregular geometry is proposed to characterize the channel characteristics in street scattering environments.Firstly,the irregular semi-ellipsoid is adopted to model the street scattering environment,in which the irregular structure is caused by the vertical buildings on both sides of streets.The boundaries of different scatterer distribution areas are given in detail,and then the geometric relationships of the transmitter,scatterer and receiver constrained by different boundaries of the scattering areas are derived.Secondly,in order to characterize statistical characteristics of the channel,the marginal probability density function(PDF)of the angle-of-departure(AOD),the joint PDF of time-of-arrival(TOA)/AOD and PDF of Doppler shifts are derived in detail.Finally,the impacts of street width and antenna array on the statistics of the channel are studied.(3)To address the issue of high pilot-overhead in downlink channel estimation of massive MIMO systems,sparse channel representation and estimation methods based on compressed sensing(CS)theory are proposed by taking advantage of the channel sparsity in the angle domain.Firstly,aiming at the problems of sparse channel representation based on discrete Fourier transform(DFT)matrix,a sparse channel representation method based on classification dictionary learning is proposed,and the process of sparse channel representation based on classified K-SVD(C-K-SVD)dictionary learning algorithm is presented in detail.Secondly,the pattern coupled Bernoulli Gaussian(PC-BG)model is proposed to describe the prior structure of the sparse channel.Meanwhile,since it is difficult to conduct accurate Bayesian estimation for the PC-BG prior model,a new channel estimation method combining the generalized approximate message passing(GAMP)technology with the expectation maximization(EM)method is proposed.Finally,the performance of the proposed sparse channel representation and estimation method is studied by simulation.(4)To describe the hierarchical-block structure resulting from the joint sparsity of the channel and antenna array,a hierarchical-block sparse channel estimation model and the corresponding channel parameter estimation algorithm are proposed.Firstly,a new hierarchical-block prior model is designed to describe the structural sparsity of the channel,which adopts a set of hyperparameters to describe the two levels of block-sparsity,i.e.,external block-sparsity and internal block-sparsity.Secondly,a channel parameter inference algorithm based on VBI technology is proposed,named as hierarchical-block VBI(HB-VBI).Meanwhile,to address the issue of high-dimensional matrix inverse in each iteration of HB-VBI,a method of calculating approximated posterior distribution of channel vectors based on GAMP technology is presented,which can avoid high-dimensional matrix inversion and therefore reducing the computational complexity.Thirdly,based on the Gaussian message propagation in a factor graph,a concise derivation from belief propagation(BP)to GAMP is given,which can reveal the internal relationship between them.Finally,the performance of the proposed hierarchical-block prior model and its parameter estimation method is analyzed by simulation.
Keywords/Search Tags:Massive MIMO, Wireless channel, Geometry-based stochastic model, Dictionary learning, Bayesian inference, Message passing
PDF Full Text Request
Related items