PPO-Based Joint Task Offloading and Resource Allocation for Vehicular Edge Computing Via V2I and V2V Communications

Advances in Intelligent Transportation Systems and Vehicular Edge Computing (VEC) demand task-offloading and resource-allocation schemes that can guarantee millisecond-level latency and high reliability. Yet most existing work overlooks the idle spectrum and computing resources available through Vehicle-to-Vehicle (V2V) links, leaving substantial efficiency untapped. This paper introduces a joint task offloading and resource allocation framework driven by Proximal Policy Optimization (PPO). We frame the problem as a Markov Decision Process (MDP) and train a PPO agent to select, for every task, among local execution, Vehicle-to-Infrastructure (V2I) offload, and V2V offload while allocating radio and CPU resources in real time. Extensive simulations show that the proposed method surpasses state-of-the-art baselines—random, greedy, and Deep Q-Network (DQN) policies—on both average latency and task completion ratio. Harnessing V2V links trims latency by 21.8 % and raises the completion ratio by 10.5 percentage points. The PPO agent also converges faster and maintains robust performance under highly dynamic traffic conditions, highlighting its practicality for real-world vehicular networks.