Research For Localization Under Extremely Large Antenna Array

Massive-Multiple Input Multiple Output(Massive MIMO)technology has achieved great success,by increasing the number of antennas on the array to obtain beamforming and multiplexing gain for achieving more system capacity and spectrum efficiency effectively.However,based on the latest channel measurement results,with the increase of the antenna number,spatial non-stationary characteristics are unavoidable,where the signals from the mobile station can’t be received by all the antennas of the array,causing the problem of different spatial spectrums for different antennas and posing new challenges to conventional methods.Extremely Large Aperture Array(ELAA),as one version of very large antenna array,can effectively expand the number of antennas and take advantage of the non-stationarity to achieve higher gain,while its path loss obeys near-field propagation,thus having significant potential for localization.Considering highly accurate positioning as the important scenario of 6G intelligent sensing multi-service,this paper has studied the location in ELAA system under spartially non-stationary channel conditions as follows:This paper first investigates the ELAA system model under spatially non-stationary channel conditions based on the visible region(VR)of the diffractor and the near-field spherical wave model,and then analyses the approximation error of the parabolic wave to the spherical wave model.The existing classic array processing algorithms,e.g.Multiple Signal Classification(MUSIC)and Minimum Variance Distortionless Response(MVDR)are simulated and analyzed in ELAA system.Considering the conventional Estimation of Signal Parameters using Rotational Invariance Techniques(ESPRIT)algorithm can’t be implemented in this system,thus a fourth-order-statistics-based ESPRIT-like algorithm is investigated in this thesis as its alternative.Secondly,this thesis focuses on the improvement of compress-sensing-based localization algorithm under ELAA scenario.Since these algorithms are based on over-complete discrete dictionary matching,the continuous search area needs to be discretized,but the realistic target position commonly does not fall exactly on the discrete point.To solve the above off-grid errors,the existing off-grid refinement algorithm is improved to multidimensional cases,which effectively reduces the errors under near-field positioning.Next,considering the positioning problem under spatially non-stationary channel conditions,the Logistic Stick Breaking Process(LSBP)in image processing is extended and combined with the developed spartial spectrum sparse mixture model in this thesis,where the off-grid location of the ELAA system is realized by adaptive partition of stable channel subarrays.Utilizing additional antenna position information,the simulation results show that the clustering process LSBP is more accurate,encouraged by clustering adjacent antennas with the similarity of recieved paths,and the proposed model achieves more joint gain among the partitioned stationary subarray than conventional counterparts,utilizing geometry while it also outperforms the other two-step methods.Finally,under the multi-band distributed array scene,a wideband Bayesian model based on Dirichlet process clustering is presented in this thesis,which effectively clusters received signals with the same components in the spatial spectrum,and achieves joint gain in the wideband case.At the same time,the positioning error reduction is confirmed through comparisons with other wideband methods.