Most of the currently active Earth-observing satellites are entirely controlled from the ground: observation plans are regularly computed on the ground (typically each day for the next day), uploaded to the satellite using visibility windows, and then executed onboard as they stand. Because the possible presence of clouds is the main obstacle to optical observation, meteorological forecasts are taken into account when building these observation plans. However, this does not prevent most of the performed observations to be fruitless because of the unforeseen presence of clouds. To fix this problem, the possibility of equipping Earth-observing satellites with an extra instrument dedicated to the detection of the clouds in front of it, just before observation, is currently considered. But, in such conditions, decision upon the observations to be performed can be no longer made offline on the ground. It must be performed online onboard, because it must be performed at the last minute when detection information is available and because visibility windows between Earth-observing satellites and their control centers are short and rare. With agile Earth-observing satellites which are the next generation ones, decision-making upon observation requires the computing of an as short as possible attitude trajectory allowing the satellite to point to the right ground area within its visibility window. In this paper, we show the results of an experiment consisting in using a continuous constraint satisfaction problem solver (RealPaver) to compute such optimal trajectories online onboard.
展开▼
