Research On Channel Estimation Methods Of MIMO-OTFS System

Due to the non-stationary fading and fast time-varying characteristics of highmobility wireless channel,radio waves will be subjected to severe doubly dispersive and selective fading.The Doppler shift leads to severe inter carrier interference of traditional OFDM,which results in system performance plummeting.Orthogonal Time Frequency Space(OTFS)technologies as an alternative modulation scheme of OFDM is applied in high-mobility communication scenarios.In OTFS system,modulation symbols are placed in the Delay-Doppler domain and mapped to the time domain through a series of twodimensional transformations.Compared with traditional OFDM,it realizes Doppler diversity on the basis of frequency diversity and makes more full use of resource grid.Moreover,the delay-Doppler channel presents better separateness,compactness and sparsity,which can more conveniently track the real-time state of the fast time-varying channel.This thesis mainly discusses the channel estimation methods of OTFS system in high-mobility communication scenarios.This thesis firstly introduces the modulation and demodulation principle of OTFS system in detail,deduces its input-output relation under integer Doppler and fractional Doppler,and gives the principle and implementation process of embedded pilot-aided channel estimation method based on this.This method embedds pilot pulse in the all-zero protection band to suppress the interference between pilot and data,which can accurately estimate the channel at the cost of a large pilot overhead.Therefore,this thesis also discusses the channel reconstruction method based on compressed sensing recovery algorithm,and mainly discusses the performance of OMP,2D-SOMP and SBL algorithms.The 2D-SOMP algorithm structure the traditional OMP algorithm to obtain better NMSE performance under integer Doppler based on the two-dimensional sparse structure of the Delay-Doppler channel under integer Doppler.SBL algorithm based on Sparse Bayes Learning is more suitable for solving block-sparse Delay-Doppler channel under fractional Doppler.Simulation results show that these two methods have better NMSE performance than traditional OMP algorithm and embedded pilot-aided method.In addition,the channel estimation method of MIMO-OTFS system is further discussed in this thesis.In the MIMO-OTFS system,multiple antennas increase the complexity of the Delay-Doppler channel.In the traditional embedded pilot-aided channel estimation method,each transmit(Tx)antenna arranges its own pilot,guard,and information symbols on the delay–Doppler grid for transmission.The pilot symbol is used to estimate the channels from that Tx antenna to each receive(Rx)antenna.This method has high pilot overhead and is not suitable for the MIMO system with a large number of antennas.Therefore,this thesis discusses the channel structure of MIMO-OTFS in detail,and uses CVR to map it to the angle domain to make its four-dimensional channel sparsity more obvious.By taking advantage of this sparsity,4D-SOMP algorithm recover the four dimensions of the channel one by one.This thesis also proves through simulation that the4D-SOMP algorithm has better NMSE performance than embedded pilot and traditional OMP algorithm in MIMO-OTFS system,and is more suitable for MIMO scenarios.