Abstract
—Designing a robust channel estimation method for Reconfigurable Intelligent Surfaces (RIS)-assisted multiple-input multiple-output (MIMO) systems with a limited number of pilot signals remains quite challenging, given the large number of channel coefficients to estimate. This work designs a novel, practical, and robust channel estimation method for RIS-assisted MIMO systems. Its main feature is to leverage score-based diffusion model (SDM) that uses gradients of probabilistic density and annealed Langevin dynamics to perform channel estimation in various channel environments with great accuracy. Our novel SDM-based channel estimation for MIMO-RIS is underpinned by a lightweight network, which is distilled from RefineNet, to enhance the accuracy of the SDM score. Simulation results show that our proposed method has a close-to-lower-bound in-distribution performance, and a robust out-of-distribution performance. Our proposed method significantly outperforms both conventional and machine learning based methods commonly used for MIMO systems, enabling more opportunities to improve channel estimation accuracy in complex scenarios.