Asynchronous DRL and Trajectory Prediction Based Vehicle Edge Computing Task Offloading

With the rapid expansion of the Internet of Vehicles (IoVs), the complex data generated by in-vehicle devices has surged, challenging vehicle computational resources to meet real-time processing demands. To tackle this challenge, Vehicular Edge Computing (VEC) provides improved computational capabilities via roadside units (RSUs) and mobile vehicles. Nevertheless, the rapid movement of these vehicles restricts the communication with base stations, making task offloading a more complex endeavor. This study introduces a task offloading strategy that utilizes trajectory prediction to boost task completion rates by forecasting the base stations that a vehicle is likely to encounter. Furthermore, we propose an optimization method grounded in deep reinforcement learning, specifically A3C, to streamline the offloading process. Our simulation findings reveal a notable increase in task completion rates, coupled with shorter completion times.