Channel Estimation for Double-RIS-Assisted Multi-User MIMO System in the Presence of Obstructed Links Using Deep Learning

Maximizing the performance of reconfigurable intelligent surfaces (RIS) in wireless communication systems critically depends on accurate channel estimation. This challenge becomes more pronounced in the presence of double RISs, particularly in severely obstructed environments, where the number of channel coefficients increases substantially and more pilot overhead is required. To address these limitations, we propose a deep learning-based two-stage dual-reflection cascaded channel estimation for multi users multiple-input multiple-output (MIMO) system. In the first stage, the dual-reflection channel of a reference user is estimated and in the second stage, the channels of the remaining users are estimated using the reference estimate. The proposed approach operates the first RIS in a known fixed mode, while the second RIS in a dynamically adjustable mode, reducing the number of pilots and channel coefficients to be estimated. Simulation results demonstrate that the proposed deep learning method significantly outperforms existing methods as far as estimation accuracy is concerned especially in low SNR environments.