High Mobility Channel Estimation Research Over Massive MIMO-OTFS System

Massive multi-input multi-output(MIMO)is one of the key technologies of the 5-th Generation Mobile Networks(5G).Deploying a large number of array antennas can provide extremely high beam domain resolution to improve the spectral and energy efficiency of the system.Then,combined with orthogonal frequency division multiplexing(OFDM),it can reduce inter-symbol interference in multipath environments.For massive MIMO-OFDM systems,many techniques are highly dependent on the acquisition of accurate channel state information,such as beamforming and precoding techniques.However,with the development of 5G wireless communication systems,high mobility scenarios are attracting more and more attention.Due to the short coherence time of data processing and feedback delay,the channel information obtained at the base station tends to be outdated,and the link performance will be significantly degraded.In recent years,the emergence of new orthogonal time frequency space(OTFS)modulation waveforms has brought opportunities to solve problems in high mobility wireless communications.In this paper,for high mobility massive MIMO systems,due to the existence of time-selective fading channels with a large number of antennas,the acquisition of accurate channel state information will lead to problems of low accuracy and large complexity.Therefore,this paper aims to use uplink channel estimation and downlink channel estimation is the main content,and a Bayesian statistical inference strategy and a massive MIMO-OTFS scheme are proposed respectively.The specific work is as follows:Aiming at the problem of high mobility massive MIMO uplink channel estimation,in order to solve the huge computational complexity of traditional methods applied to high mobility massive MIMO channel estimation,this paper uses sparse Bayesian learning(SBL)based on the framework and the idea of fast Bayesian inference,a high mobility massive MIMO uplink channel estimation scheme based on the fast expectation maximization based variational Bayesian(EM-VB)algorithm is proposed.This scheme uses the idea of compressed sensing to transform the time-varying channel estimation problem into a sparse signal recovery problem.Then,by using the sparse Bayesian framework to build a channel parameter graph model,the sparsity of the channel matrix is effectively mined,and then the uplink channel estimation is achieved with extremely low training overhead and computational complexity with the help of powerful compressed sensing tools.The simulation result analysis proves the effectiveness and reliability of the proposed fast EMVB algorithm based on the SBL framework.In addition,for the high mobility massive MIMO downlink channel estimation problem,in order to solve the unacceptable training overhead and computational complexity in the frequency division duplexing(FDD)system,this paper uses a new OTFS modulation waveform,which a highly dynamic time-varying channel can be effectively converted into a two-dimensional quasi-static channel,and a high mobility massive MIMO-OTFS downlink channel estimation scheme based on OTFS technology is proposed.This scheme utilizes the spatial sparse characteristics of massive MIMO channels and OTFS modulation to construct a three-dimensional(3D)sparse representation system for massive MIMOOTFS high mobility channels.The downlink channel intrinsic parameters are then accurately reconstructed by means of the uplink channel estimation process and the channel parameter reciprocity that exists between uplink and downlink.Based on this,by analyzing the energy dispersion effect of the channel in the delay-Doppler domain,the complex mapping mechanism of massive MIMO-OTFS transmitting and receiving 3D signal space is revealed.Finally,a reasonable pilot design and user scheduling strategy are completed to form a delayDoppler-angle domain channel tracking strategy with extremely low overhead and complexity.The simulation results demonstrate the high efficiency and reliability of the proposed massive MIMO-OTFS scheme,which outperforms traditional methods even in FDD mode.