Multi-Satellite Cooperative Computing Task Offloading Strategy Based on Deep Reinforcement Learning
Hufan Cao, Yizhuang Peng, Houpeng Wang, Haolin Jia, Linghai Kong, Suzhi Cao · 2024
With the proposition of an integrated space-air-ground-sea communication network concept, satellites have become a key bridge connecting the globe. Low Earth Orbit (LEO) satellites, in particular, have become an important force in achieving global seamless communication due to their low latency and wide coverage advantages. LEO satellite networks can provide computing services to ground users under extreme conditions, which is of significant practical importance. The coordination among satellite edge nodes not only improves the utilization of space-based resources but is also key to enhancing the QoS of satellite edge computing services. Therefore, we propose a computation offloading strategy based on deep reinforcement learning (DRL) for multi-satellite cooperative computing scenar-ios (DQN-SCCO). First, the optimization problem is formulated as a Markov decision process (MDP), and the optimal solution to the problem is approached through a deep Q-network (DQN). Simulation results show that DQN-SCCO can effectively reduce the task set response delay and improve the task offloading success rate compared to baseline algorithms.