DRL-based Task Scheduling and Edge Collaboration for LEO Satellite Networks

Low Earth Orbit (LEO) satellites equipped with mobile edge computing (MEC) servers can break through the coverage limitation of ground base stations and provide ubiquitous services to users, which is an indispensable technique in future networks. However, the satellite’s energy fluctuates with its movement relative to the sun, restricting its battery load capacity, which further hinders it from supplying continuous services to users. In this paper, to maximize the sum of all users’ quality of experience ($\mathrm{Q o E}$) while meeting satellite energy constraints, we propose a dynamic task scheduling and resource allocation scheme, enabled by the cooperation between satellites and the cloud center. As this issue is a long-term decision and optimization problem, we model it as a Markov decision process (MDP) and exploit the improved Proximal Policy Optimization(PPO) algorithm to address it. Simulation results verify the effectiveness of the algorithm, which can significantly improve users’ QoE compared with other baseline algorithms.