Intelligent Task Offloading Method for Vehicular Edge Computing Based on Improved-SAC
Tiyin Xiao, Yuanyuan Qi, Tao Shen, Yan Feng, Lin Huang · 2022 IEEE 5th Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC) · 2022
In the era of rapid development of the Internet of Vehicles (10V), how to use vehicular edge computing to process high-complexity intelligent tasks in real time is an urgent problem to be solved. To cope with the offloaded intelligent task interruption in IoV caused by mobility, this paper proposes an offloading model based on improved Soft Actor-Critic (Improved-SAC) algorithm to minimize the latency. Using the divisibility to task partition, we divide intelligent tasks into multiple layers to reduce task interruption. The problem formulation of minimizing the latency is obtained by jointly considering vehicular mobility, location, computing power and coverage of edge nodes. This paper utilizes Value-Difference Based Exploration (VDBE) and dual experience playback pool to improve the efficiency of sample data utilization and the exploration ability of SAC, and introduces Improved-SAC algorithm into knowledge plane. The numerical results show that the Improved-SAC algorithm can reduce the latency by up to 64% compared with Q-learning and DQN algorithm, and generate near-optimal offloading decisions within 70ms.