An Effective Method for Satellite Mission Scheduling Based on Reinforcement Learning

With the increasing number of satellites in orbit and the growing observation missions, how to make an allocation scheme with the maximization of total profit effectively has become increasingly important. In this paper, an effective method based on Reinforcement Learning is proposed to solve satellite mission scheduling problem, in which the arrival missions are arranged immediately without waiting all missions collected. Firstly, a mathematical model based on Markov Decision Process is established, whose goal is to find an optimal policy to maximize the accumulated reward. Then, Asynchronous Advantage Actor-Critic algorithm with neural network is used to assign missions to different satellites. The simulation experiments with comparison to first come first service algorithm and genetic algorithm are conducted, which demonstrates that the proposed method performs well with respect to real-time speed and solution quality.