Observation scheduling for a state-of-the-art SAREOS: Two adaptive multi-objective evolutionary algorithms

Abstract

An observation scheduling for Gaofen-12 (OSFGF-12), a state-of-the-art earth observation satellite installed a novel synthetic aperture radar, is investigated in this paper. Considering a novel observation capability, uninterrupted imaging mode, we build a three-objective optimization model OSFGF-12 to minimize the loss rate of ground targets, the invalid observation and the times of opening sensor. We design two adaptive multi-objective evolutionary algorithms, ALNS + NSGA-II and ALNS + NSGA-III, where an adaptive large neighborhood search algorithm is adopted as the breeding algorithm, and NSGA-II and NSGA-III are respectively adopted as selection mechanisms. A supplemental selection mechanism is proposed to remedy NSGA-III, by which the iteration values of hypervolume obtained by ALNS + NSGA-III increase stably, and ALNS + NSGA-III achieves the Pareto frontier much faster. Extensive simulation instances are designed to test these two algorithms.