Deep Reinforcement Learning-Based Joint Task Offloading and Radio Resource Allocation for Platoon-Assisted Vehicular Edge Computing

Platoons, formed by smart vehicles driving in the same patterns, bring potential benefits to road traffic efficiency while providing a promising paradigm to execute computation tasks with onboard computing resources. However, constrained resources of individual vehicles (IV), limited wireless coverage of vehicular communication nodes as well as high mobility of running platoons pose critical challenges on task scheduling and resource management. To address these challenges, we propose a platoon-based vehicular edge computing mechanism, which exploits computation capabilities of both platoons and edge computing enabled Roadside Units (RSUs), and jointly optimizes task offloading target selection and resource allocation. Taking aim at minimize delay cost and energy consumption of the platoon-based task execution, we leverage deep deterministic policy gradient (DDPG) to design a learning algorithm, which efficiently determines target computation servers and obtains optimized resource scheduling strategies. Numerical results demonstrate that our algorithm significantly reduces delay and energy costs in comparing its performance to that of benchmark schemes.