A DRL-Based Optimization Framework for RIS-Assisted Vehicular Edge Computing: AoI and Energy Perspective

This letter proposes a hierarchical deep reinforcement learning (DRL) framework for resource allocation in reconfigurable intelligent surface (RIS)-assisted vehicular edge computing (VEC) systems. The objective is to minimize the Age of Information (AoI) and improve energy efficiency under dynamic vehicular conditions. The framework decouples the complex joint optimization into two layers: the upper layer employs a Twin Delayed Deep Deterministic Policy Gradient (TD3)-based controller to optimize RIS phase shifts, while the lower layer uses a Federated Multi-Agent DDPG (FMADDPG) algorithm for distributed power allocation. Simulation results, based on realistic urban mobility, show that the proposed approach outperforms benchmark methods such as Proximal Policy Optimization (PPO), Block Coordinate Descent (BCD), and conventional DDPG in terms of AoI reduction, energy savings.