Low-Earth-Orbit Satellite Assisted Edge Computing for Vehicular Networks: A Task Priority-Based Delay Minimization Approach

Abstract

With the rapid advancement of the Internet of Vehicles (IoV), new types of vehicle applications are emerging continuously. These applications impose increasingly stringent requirements on delay and quality of service standards, which are difficult to meet for vehicle terminals with limited resources. Meanwhile, in the complex computing task system of vehicles, there exists a close correlation between task priorities and computing tasks. As a core technology of space-ground integrated networks, low Earth orbit (LEO) satellite communication integrated with vehicle networking can ensure high-efficiency and reliable real-time data transmission. However, additional delay and energy consumption are incurred during the communication between vehicle terminals and satellites. Introducing edge computing into satellite-assisted vehicular networking can satisfy the computing demands of vehicle terminals and reduce the processing delay of vehicle applications, with computing offloading being the key technology. We focus on the LEO satellite assisted vehicular edge computing network. Considering the varying sensitivities of different vehicle tasks to delay and energy consumption, it precisely sets task priorities and proposes a task-priority scheduling scheme. With the objective of minimizing the average delay under constraints of energy consumption, the problem is modeled as a Markov decision process (MDP) and addressed by employing the proximal policy optimization (PPO) algorithm within the framework of deep reinforcement learning (DRL). Simulation results demonstrate that the proposed computational offloading algorithm can effectively decrease the system’s average delay, outperforming other benchmark testing methods significantly.