A Priority-Aware Intelligent Task Offloading Framework for Autonomous Vehicles in Vehicular Edge Computing Using Deep Reinforcement Learning

<title>Abstract</title> <p> Artificial Intelligence (AI) has become a key enabling technology for intelligent resource management in Vehicular Edge Computing (VEC), where autonomous vehicles continuously generate computational tasks with diverse latency requirements and dynamic resource demands. Efficiently determining where these tasks should be executed remains a challenging problem because onboard vehicles, neighboring vehicles, roadside units (RSUs), and cloud servers all possess heterogeneous and limited computational resources. Conventional task offloading approaches generally rely on static heuristics or optimization methods that cannot effectively adapt to rapidly changing traffic conditions and varying workloads. Moreover, existing studies seldom integrate intelligent decision-making, resource prediction, and caching into a unified AI-driven framework. This paper proposes an <italic>Artificial Intelligence-based framework</italic> for priority-aware task offloading in VEC. A Deep Reinforcement Learning (DRL) model is employed to learn optimal task offloading policies that minimize response time while adapting to dynamic network and resource conditions. To further improve the responsiveness of latency-critical applications, a Long Short-Term Memory (LSTM) model predicts future resource demands and proactively reserves RSU resources for high-priority tasks. Compared to other related studies, the proposed framework reduces the response time delay by 31.2%. </p>