Efficient Collaborative Computing for Multilayer LEO Satellites With Spatiotemporal Dynamics: A Long-Term Continuous Timescale Optimization

With the proliferation of smart devices and the expansion of human production and living areas, low Earth orbit (LEO) satellite computing is needed to meet the computing demands over a wide area. Due to the limited resources that a single LEO satellite can carry, it is essential to realize intersatellite collaborative computing to achieve efficient onboard processing. However, the high spatiotemporal dynamics of satellite networks pose significant challenges to the establishment of connection and offloading decisions in collaborative computing. Facing these challenges, we propose a multilayer LEO satellite collaborative computing framework by integrating LEO satellites from different orbits. Considering the continuity of task generation, we establish a temporal model for on-board processing. On this basis, we optimize intersatellite offloading decisions to minimize the average system cost over a long-term timescale. To address the constantly changing environment information and the strict constraints on task completion time, we propose using the proximal policy optimization (PF-PPO) algorithm to solve the problem. Extensive simulation results illustrate the effectiveness of the proposed algorithm, which can achieve lower system costs compared with benchmark methods under different conditions. It is also proved that our algorithm has stable performance for the system over a long-term continuous timescale.