Intelligent Reflecting Surface Optimization for MIMO Communication Using Deep Reinforcement Learning

This paper focuses on the optimization of the phase shifts of an intelligent reflecting surface (IRS) for an IRS-aided multiple input multiple output (MIMO) communication system. Motivated by the massive success of deep reinforcement learning (DRL) algorithms in handling high-dimensional continuous action spaces and tackling non-convex optimization problems, we propose a deep deterministic policy gradient (DDPG) framework for solving the formulated non-convex optimization problem. Numerical simulations demonstrate the robustness and efficiency of the proposed model in terms of spectral efficiency and algorithm run time when compared to a state-of-the-art scheme.