Search

In this paper, we studied mission planning problem for Earth observation satellites using a competitive learning strategy. In particular, a scheme based on Q-network is proposed to solve scheduling problem of a single satellite, while a profit-based competitive strategy is introduced to address the scheduling conflict problem inherent with multi-satellite missions. More specifically, the Q-network is integrated with the competitive strategy to improve the algorithmic scalability and to reduce the communication burden, and the search-window is designed to improve scheduling efficiency by eliminating the obvious unbefitting choices in the learning process. Simulation results verified the performance of the proposed schemes. [ABSTRACT FROM AUTHOR]