GPU-accelerated optimization of CubeSat constellation design considering cloud cover uncertainty

Abstract

Over the past decades, the miniaturization of commercial off-the-shelf electronics has led to the development of nanosatellites, especially on the basis of CubeSat standard. And recently, space community begins to realize that constellations of hundreds of CubeSats have potential to provide comparable or even greater capabilities than a single traditional large satellite in real space missions, but with lower cost and higher robustness. However, to design such a satellite constellation for reaping its benefits entails solving large-scale multi-objective optimization problems. One obvious problem is to decide the orbit elements of each CubeSat in the constellation with the purpose of optimizing some performance measures. In order to reduce the burden on the orbit design engineers and speed up the design process of future constellation missions, this study developed a tool named ACCELERATOR, which is essentially a multi-objective genetic algorithm with GPU-based modules, for obtaining optimal orbital configuration of the CubeSat constellations in reasonable time. Moreover, the primary limiting factor in the optical satellite constellation scheduling problem, namely the cloud cover uncertainty, was considered during the iterations of the genetic algorithm with Monte Carlo method. Therefore the resulting optimal constellations were believed to be robust enough to cope with the uncertainty while in the operations phase of the satellite constellation.