Evolutionary Multiobjective Satellite Range Scheduling With Learning-Guided Population Generation

Most existing evolutionary approaches to satellite range scheduling seek optimal solution in terms of the request satisfaction. The scheduling demand of managing ground station resource is seldom considered, which restricts their real-world applications. To effectively generate a set of more rational satellite range schedules, this paper establishes the multi-objective satellite range scheduling mathematical model. Unlike existing approaches, we propose a general population generation approach to solve the problem without relying on any specific kind of evolutionary algorithm, so different types of evolutionary algorithms can be extended to satellite range scheduling without modifying the original framework and search strategy. The idea is to utilize the request satisfaction and resource utilization knowledge learnt from parent schedules to guide the generation and updating of new solutions. Furthermore, an iterative rewriting operator is designed to guide a biased faster convergence towards the low request failure region in objective space. The proposed approach has been applied to five different types of classical and state-of-the-art evolutionary algorithms and examined on benchmark problems. Experimental results illustrate the search efficiency enhancement and good adaptability to different evolutionary algorithms, which show the broad application prospect for satellite range scheduling.