Joint Task Offloading and Service Migration in RIS Assisted Vehicular Edge Computing Network Based on Deep Reinforcement Learning

To address the increasingly complex computing tasks of intelligent vehicles, we consider a framework for Reconfigurable intelligent surface (RIS) assisted vehicular edge computing (VEC) networks. We aim to maximize the weighted sum throughput of all vehicular user equipments (VUEs) while limiting the latency of all VUEs in each time slot to a certain range by jointly optimizing computational edge servers for all VUEs, the deployment location of the RIS and its reflecting beamforming matrix. We propose a deep reinforcement learning (DRL) based algorithm to solve the problem. Evaluation results show the effectiveness of the proposed algorithm and verify that RIS deployment is a valid solution to enhance the communication and computation in VEC network.