Deep Reinforcement Learning for Practical Phase Shift Optimization in RIS-aided MISO URLLC Systems

Reconfigurable intelligent surfaces (RISs) can assist the wireless systems in providing reliable and low-latency links to realize the requirements in Industry 4.0. In this paper, the practical phase shift optimization in a RIS-aided ultra-reliable and low-latency communication (URLLC) system at a factory setting is performed by applying a novel deep reinforcement learning (DRL) algorithm named as twin-delayed deep deterministic policy gradient (TD3). First, the system achievable rate in finite blocklength (FBL) regime is identified for each actuator then, the problem is formulated where the objective is to maximize the total achievable FBL rate, subject to non-linear amplitude response and the phase shift values constraint. Since the amplitude response equality constraint is highly non-convex and non-linear, we employ the TD3 to tackle the problem. The considered method relies on interacting RIS with industrial scenario by taking actions which are the phase shifts at the RIS elements, to maximize the total FBL rate. We assess the performance loss of the system when the RIS is non-ideal, i.e., non-linear amplitude response with/without phase quantization and compare it with ideal RIS. The numerical results show that optimizing phase shifts in non-ideal RIS via the considered TD3 method is highly beneficial to improve the performance.