Joint Channel Extrapolation and Scattering Environment Sensing for Multi-User TDD Massive MIMO-OFDM Systems

In this paper, we consider joint channel extrapolation and scattering environment sensing for multi-user timedivision duplexing (TDD) massive multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems. Unlike conventional angle-delay domain channel modeling techniques, we adopt a location domain channel modeling approach to leverage the spatial overlap of scatterers across different users. By exploiting this, we propose a novel two-stage joint multi-user channel extrapolation and scattering environment sensing algorithm. In the coarse estimation stage, a low complexity Spatial and Temporal Multiple Signal Classification (ST-MUSIC) algorithm is utilized to perform independent channel extrapolation and scatterer localization for each user. In the refined estimation stage, an aggregated location grid constructed from the coarse estimation result is used to enable a sparse representation of the location domain channels of all users. And by combining the inverse-free variational Bayesian inference (IF-VBI) and the expectation maximization (EM) algorithm, an EM-IF-VBI algorithm is designed to jointly refine the location domain channel coefficients and aggregated location grid of all users, which can exploit the spatial overlap of scatterers across different users to simultaneously achieve more accurate channel extrapolation and scattering environment sensing. Simulation results show that our proposed method significantly outperforms existing baseline methods.