Joint Multi-User Channel Estimation for RIS-Assisted Massive MIMO Systems

Accurate channel estimation plays a pivotal role in realizing the passive beamforming gain of reconfigurable intelligent surfaces (RIS). Multi-user RIS-assisted channels exhibit a common sparse structure in the angular domain, which offers the potential to enhance channel estimation performance. Nonetheless, most existing methods simply partition the multi-user channel estimation problem into multiple sub-problems, which regrettably neglect the exploitation of the shared property across different users. In this paper, we formulate a unified multi-user channel estimation problem and propose a computationally efficient message passing approach to jointly extract the common sparsity. We first introduce a new sparse Bayesian learning (SBL) framework for joint multi-user RIS-assisted channel estimation, where some auxiliary variables and Dirac delta distributions are introduced to handle the intricate interplay of numerous unknown variables within the joint sparse channel representation. Subsequently, we devise a reassembled message passing algorithm for the associated joint Bayesian inference, incorporating a three-stage expected propagation approximation (EPA) procedure along with a novel reassembling technology to facilitate variable decoupling and ensure reliable sparse signal recovery. Simulation results demonstrate the superiority of the proposed algorithm over the state-of-the-art counterparts.