A hybrid genetic algorithm is used to schedule tasks for a satellite that can be modeled as a robot whose goal is to retrieve objects from a two-dimensional field. The objective is to find a schedule that maximizes the value of objects retrieved. Typical of the real-world tasks to which this corresponds is the scheduling of ground contacts for a communications satellite. An important feature of our application is that the amount of time available for running the scheduler is not necessarily known in advance. This requires that the scheduler produce reasonably good results after a short period, but that it also continue to improve its results if allowed to run for a longer period. We satisfy this requirement by developing what we call a sustainable genetic algorithm.