Application of SARSA-Based Reinforcement Learning Approach for Resource Allocation in Vehicular Edge Computing

Vehicular Edge Computing (VEC) holds significant promise for enhancing the performance and resource utilization of Internet of Vehicles (loV) applications. However, the dynamic nature of vehicular networks, characterized by high mobility and limited capabilities of VEC servers, poses significant challenges for resource allocation. This paper proposes a Reinforcement Learning (RL) based approach, employing the SARSA algorithm, to learn optimal resource allocation strategies in Multi-access Edge Computing (MEC) environments designed for vehicular networks. Our strategy aims to minimize service delays and balance resource utilization effectively by dynamically deciding whether to offload tasks to edge servers or cloud servers. We evaluate the proposed approach using the EdgeCloudSim simulator, comparing its performance against conventional static and heuristic resource allocation methods. Simulation results, considering various parameters and application characteristics, demonstrate the superior efficiency of our SARSA-based resource allocation solution for VEC compared to traditional methods, in terms of overall service time and resource utilization.