DRL-Based Partial Task Offloading for Multiple Vehicles in VEC Networks

As 6G and Internet of Vehicles (IoV) technologies advance, vehicles are becoming increasingly interconnected and intelligent. This evolution emphasizes the need for advanced autonomous decision-making in intelligent vehicles, which must manage limited computing and communication resources against strict task deadlines. Balancing energy consumption and computing task latency in such environments is becoming critically important. To address these challenges, this paper presents a vehicle edge computing (VEC) task offloading model rooted in deep reinforcement learning (DRL). This model is designed to cater to the multiple vehicles within the VEC system, specifically tailored for partial computation offloading schemes. Given the constraints of vehicle computing power, completing tasks promptly may be challenging. To address this, tasks can be offloaded to a base station (BS) VEC server, which possesses more robust computing capabilities. To boost convergence efficiency and improve overall system performance, we employed a DRL algorithm founded on the Actor-Critic approach to enhance the training efficacy of our model. According to our simulation results, this DRL algorithm based on Actor-Critic effectively expedites convergence, enhances performance, and reduces the overall system cost.