Enhancing V2X Communication With Active RIS: A Mixed Action DRL Approach With Perfect and Imperfect CSI

In this work, we investigate the use of active reconfigurable intelligent surfaces (A-RIS) to enhance the performance of vehicle-to-everything (V2X) communication systems, addressing limitations seen in traditional vehicular communication systems. We tackle this challenge by formulating an optimization problem aimed at maximizing the uplink sum rate for vehicle-to-infrastructure (V2I) links. In particular, we optimize transmit precoders, phase-shift matrices, transmit power, and spectrum sharing for vehicle-to-vehicle (V2V) links while ensuring that V2V links maintain a minimum signal-to-interference-plus-noise ratio (SINR). To address the complex hybrid control scenarios in this context, we propose a mixed action deep reinforcement learning based algorithm. This algorithm employs two agents, delivering nearly optimal performance and effective control capabilities across various scenarios. We also compare our proposed algorithm with two conventional benchmark schemes: deep deterministic policy gradient (DDPG) with discrete actions (DA), a mixed action greedy policy and alternating optimization (AO) approach. Additionally, we assess the effectiveness of our algorithm under the presence of imperfect channel state information (ICSI).Simulation results demonstrate the efficacy of our proposed algorithm, showing that A-RIS significantly enhances vehicular communication quality and hence achieves high performance with low overhead. We analyze the impact of factors such as the number of A-RIS elements, vehicle speed, loss, execution time, amplification power, and CSI error.