Federated Learning-Based Optimization for RIS-Enhanced 6G-V2X Communications

To address the performance challenges of reconfigurable intelligent surfaces (RIS) in full-duplex (FD) 6G vehicle-to-everything (V2X) communications, we propose a novel analytical framework. This framework evaluates an FD base station (BS) simultaneously communicating with one uplink (UL) and one downlink (DL) vehicle, each assisted by a separate RIS. Departing from traditional methods, we introduce a Federated Learning (FL)-based approach for enhanced performance. We jointly optimize the phase-shift matrices at both RISs to maximize the sum-rate under discrete phase-shift constraints, managing the non-convexity using Karush-Kuhn-Tucker (KKT) conditions. Our approach significantly improves sum-rate performance, outperforming DRL-RIS-V2X, DDQN-RIS-V2X, and NO-RIS-V2X configurations. We also examine the UL-DL rate trade-off in FD systems, highlighting the importance of power management. Empirical results demonstrate notable improvements in signal strength and system robustness, confirming the effectiveness of the proposed FL-RIS-V2X approach in optimizing V2X communications.