Robust Transmission Design for RIS-aided Full-Duplex-RSMA V2X Communications via Multi-agent DRL

The proliferation of mobile devices and acceleration of spectral efficiency have become a pivotal requirement for the unprecedented connectivity, and performance of vehicle-to-everything (V2X) networks. This paper investigates an unconventional framework of reconfigurable intelligent surface (RIS)-integrated full-duplex (FD) rate-splitting multiple access (RSMA) communication systems, which aims to maximize the spectral efficiency of the V2X network. In particular, a robust spectral-efficient design for the considered RIS-integrated FD-RSMA system via joint beamforming design, and power allocation under imperfect channel state information is investigated. To tackle the non-convexity of the original sum-rate maximization problem, we adopt a deep reinforcement learning (DRL)-based multi-agent (MA) proximal policy optimization (PPO) algorithm which leverages Markov decision process (MDP) formulation. Simulation results demonstrate the effectiveness of the integration of RIS, RSMA, and FD schemes for V2X networks over half-duplex (HD) and multi-user linear precoding schemes. Furthermore, the superiority of the proposed MA-PPO algorithm is validated over the counterpart PPO, and deep deterministic policy gradient algorithm (DDPG). Additionally, adopting RSMA yields a substantial 34.8
$\%$
performance boost compared to SDMA at 50 m/s velocity, highlighting RSMA's adaptive robustness amidst dynamic CSI fluctuations in vehicular networks.