CU-DRL: A Novel Deep Reinforcement Learning-assisted Offloading Scheme for Supporting Vehicular Edge Computing
In recent years, data-driven and AI-based intelligent transportation systems have been greatly developed to alleviate the public's concern about the increasingly severe traffic congestion and traffic safety issues. For supporting various safety-related ITS applications, vehicular edge computing (VEC) has been proposed as a promising technology that can effectively provide computing power and storage capacity support for vehicles in close proximity. However, in the face of the instability of communication between vehicles and other devices caused by the high-speed motion of vehicles and the complex relative motion between vehicles, how to effectively realize the relatively stable arithmetic power sharing between vehicles and edge computing devices is a critical problem that must be solved to realize VEC. Therefore, in this paper, we propose a distributed online offloading method, called Candidate Utilization-based Deep Reinforcement Learning (CU-DRL) algorithm, by exploiting the deep reinforcement learning technique. We further evaluate and demonstrate the effectiveness and correctness of the proposed CU-DRL model through simulations.
