Age of Information and Energy Optimization in Cooperative Multi-RSU C-V2X Networks

In this paper, we propose a deep reinforcement learning (DRL)-based framework to optimize the Age of Information (AoI) in a cooperative multi-RSU (roadside unit) cellular vehicle-to-everything (C-V2X) network. We consider a real-time coordination scenario where an emergency vehicle, such as an ambulance, triggers the RSUs to alert nearby vehicles to yield, while also notifying downstream RSUs. To address the dynamic and decentralized nature of this problem, we implement and compare three multi-agent DRL algorithms: Multi-Agent Proximal Policy Optimization (MAPPO), Multi-Agent Soft Actor-Critic (MASAC), and Distributed Exploration Multi-Agent Deep Deterministic Policy Gradient (DE-MADDPG). Simulation results show that MAPPO achieves the lowest AoI and highest reward convergence, indicating its effectiveness in coordinating RSU behavior.