Joint Beamforming for RIS-Assisted MU-MISO Systems Based on Model-Driven Deep Learning

This paper aims to maximize the weighted sum rate of a multi-user multiple-input single-output system that is assisted by reconfigurable intelligent surfaces (RIS). However, optimizing the phase shifts and beamforming vector at the access point remains a challenge due to the non-convex objective function and constraints. To address this problem, we propose a solution that transforms the non-convex optimization problem into a weighted minimum mean square error optimization problem. We derive a closed-form solution for active beamforming and use the power iteration algorithm to update the passive beamforming. Additionally, we introduce model-driven deep learning (DL) to obtain key variables, which accelerates convergence and reduces algorithm complexity. Simulation results demonstrate the excellent performance of our proposed algorithm with the aid of model-driven DL, achieving the same performance as the state-of-the-art algorithm in just 6% of the runtime.