Abstract
The study of electromagnetic detection satellite scheduling problem (EDSSP)has attracted attention due to the detection requirements for a large number oftargets. This paper proposes a mixed-integer programming model for the EDSSPproblem and an evolutionary algorithm framework based on reinforcement learning(RL-EA). Numerous factors that affect electromagnetic detection are consideredin the model, such as detection mode, bandwidth, and other factors. Theevolutionary algorithm framework based on reinforcement learning uses theQ-learning framework, and each individual in the population is regarded as anagent. Based on the proposed framework, a Q-learning-based geneticalgorithm(QGA) is designed. Q-learning is used to guide the population searchprocess by choosing variation operators. In the algorithm, we design a rewardfunction to update the Q value. According to the problem characteristics, a newcombination of <state, action> is proposed. The QGA also uses an eliteindividual retention strategy to improve search performance. After that, a tasktime window selection algorithm is proposed To evaluate the performance ofpopulation evolution. Various scales experiments are used to examine theplanning effect of the proposed algorithm. Through the experimentalverification of multiple instances, it can be seen that the QGA can solve theEDSSP problem effectively. Compared with the state-of-the-art algorithms, theQGA algorithm performs better in several aspects.