An Improved Many-Objective Evolutionary Algorithm for Multi-Satellite Joint Large Regional Coverage

Multi-satellite joint regional coverage aims to select the optimal combination of satellite resources to acquire the image information of the specified area. Meanwhile, more than three objectives are usually considered simultaneously during this process. Therefore, it is a typical many-objective optimization problem that is NP-hard. Most existing many-objective optimization algorithms cannot preserve extreme solutions due to the failure of Pareto dominance. In this paper, through introducing the idea of S-CDAS into the traditional NSGA-III, an improved many-objective evolutionary algorithm named NSGA-III for extreme solutions preservation (ESP-NSGA-III) is proposed with problem-specific genetic operations to generate regional coverage schemes. A comparative study is conducted with other six state-of-the-art many-objective evolutionary algorithms. Hypervolume (HV) and pure diversity (PD) metrics are used to evaluate the performance of algorithms. The simulation results show that ESP-NSGA-III has good comprehensive performance and improves the diversity of original algorithms. The maximum difference of the coverage rate between ESP-NSGA-III and other six algorithms is 0.2576 so that satisfactory regional coverage scheme can be obtained by ESP-NSGA-III. Our proposed methods are not only applicable to regional coverage tasks, but also have important reference significance for solving other real-world problems.