基于BP神经网络的星上任务可调度性预测方法

An Earth Observation Satellite Task Schedulability Prediction Method based on BP Artificial Network

传统成像卫星任务规划方法已无法应对现代对地观测活动越来越高的复杂性和动态性,需要通过任务可调度性预测方法提高成像卫星任务规划效能。在分析实际影响成像卫星任务执行因素,如任务重叠度、任务收益、气象信息等基础上,设计了一种基于BP神经网络的星上任务可调度性预测方法,并设计多个不同算例验证其效果。仿真实验结果表明,该方法在使用不完全正确的预测信息的情况下具有较高的正确率,并且在动态环境下具有较好的鲁棒性。该方法能够为未来的任务规划调度方法设计提供借鉴依据。

Formal Earth Observation Satellite(EOS)task planning and scheduling algorithms can hardly duel with the rising complication and dynamics of Earth Observation actives,thus task schedulability prediction methods are needed to guarantee the efficiency of EOS task planning and scheduling methods.As a result,based on the analysis of practical factors that affect the task completion,such as mission overlap,profit and environment information,a EOS task schedulability prediction method is designed.Then some instances are produced to test this method.The computational result shows that this method has a low error rate using imperfect forecast information,and is robust under dynamic conditions.So it can be used to improve the efficiency of future task planning and scheduling methods.