Surrogate Assisted Efficient Multi-objective Optimization for an Observation Satellite Constellation

Observation satellite constellation uses onboard remote sensors to cooperatively obtain the ground information from space, which has been widely used in Earth resource detection and military reconnaissance. To efficiently improve both the resolution and coverage performances of the observation satellite constellation, a surrogate-assisted efficient multi-objective optimization scheme is developed in this paper. The multi-objective optimization problem is defined at first to simultaneously improve the payload resolution and coverage performance. To reduce the computational cost for constellation multi-objective optimization, a radial basis function assisted non-dominated sorting genetic algorithm II method (RBF-NSGA-II) is proposed. In this approach, RBF surrogate is used to approximate the expensive constellation simulation model for optimization. During the optimization process, the surrogate is adaptively refined via k-means clustering method, which leads the search to the Pareto frontier rapidly. Finally, the proposed RBF-NSGA-II is applied to the constellation multi-objective optimization problem compared with the standard NSGA-II algorithm. The optimization results indicate that RBF-NAGA-II outperforms the competitive NSGA-II in terms of the hypervolume index. Moreover, the optimization cost of RBF-NSGA-II is reduced by 75%, which demonstrates the efficiency of the proposed method. After RBF-NSGA-II based optimization, 169 feasible Pareto solutions are obtained. Compared with the initial solution, the coverage rate of the optimized constellation configuration is increased by 37.08% at most. At the same time, the resolution is increased by 47.22% at most. The optimization results illustrate the effectiveness and practicability of the surrogate-assisted multi-objective optimization scheme for the studied observation satellite constellation.