Abstract: The agile earth observing satellites mission scheduling problem is a complicated combinatorial optimization problem with long time windows and multiple time windows constraints.In order to deal with this problem, the original problem was proposed to be divided into task & resource matching problem and single satellite task processing problem. A novel learnable genetic algorithm was proposed to solve the task & resource matching problem. The knowledge model learned and extracted knowledge from the iterative process of the genetic algorithm, brought back feedback and guided the search process of the algorithm.Backward time slack and best position search method were designed to solve the single satellite task processing problem. Experiment results demonstrated the effectiveness of the proposed approach.

Key words: agile satellite, earth observing, multi-satellites cooperation, mission scheduling, decomposition optimization, learnable genetic algorithm

摘要： 多敏捷对地观测卫星联合对地观测调度问题是一个具有长时间窗、多时间窗等复杂约束的组合优化问题。为了解决该问题，提出将原问题分解为任务资源匹配及单星任务处理两个子问题的分解优化思路。设计了学习型遗传算法解决任务资源匹配子问题，算法中的知识模型在算法迭代过程中学习和提取知识，反馈并引导算法对任务资源匹配的搜索寻优过程。采用后移滑动策略及最优插入位置搜索策略解决单星任务处理子问题，并采用基于规则的方式处理其他约束。实验结果证明了所提方法的有效性。

关键词: 敏捷卫星, 对地观测, 多星联合, 任务调度, 分解优化, 学习型遗传算法