Deep reinforcement learning for task offloading in vehicular edge computing

Abstract

With the rapid growth of emerging vehicle applications, the great demands on low delay and energy of the vehicles are driven, vehicular edge computing (VEC) has been proposed to enhance the computation capacities at the edge of the vehicle. However, traditional VEC severs use fixed edge nodes to offload tasks making the system cost high greatly. To alleviate this issue, we propose a method which utilizes the idle resources on intelligent vehicles to assist edge computing offloading in the heterogeneous network. In order to minimize the system cost (the total latency and energy), we introduce deep reinforcement learning (DRL) and propose an improved Q-learning algorithm to jointly solve the task offloading and processing problem, and make the optimal offloading decision. The simulation results show that the improved Q-learning algorithm can effectively reduce the total system cost and improve the quality of service of VEC.