A Deep Reinforcement Learning Based EV-RSU Mapping Method for EVC-Assisted Multi-Tier VEC System

In the Internet of Vehicles (IoV), latency sensitive intelligent tasks require efficient processing, which can be improved through multi-tier vehicular edge computing (VEC). Previous study has proposed an advanced electric vehicle cluster (EVC) assisted multi-tier VEC system, which aims to jointly optimize offloading and resource allocation for road task vehicles (RTVs) and manage charging power for smart electric vehicles (EVs) using a game strategy to enhance system efficiency. Building on this previous work, our study employs a Deep Q-Network (DQN) algorithm to optimize the EV-RSU mapping, identifying the limitations of Nash equilibrium (NE) in handling large-scale vehicles and more complex tasks. The simulation results demonstrate that as the number of RTVs and the initial queue length in the RSU increasing the proposed DQN-based EV-RSU mapping method shows the minimum system average latency.