Solving Over-Constrained Satellite Remote-Sensing Scheduling Problems

The high demand for satellite imagery and the relative scarcity of commercial imaging satellites combine to produce scheduling problems in which one or more of the resources is over constrained. Over-constrained scheduling problems can be problematic for constraint programming methods because of the overhead associated with propagating constraints and because of the time spent backtracking over a large number of tasks that are very unlikely to be scheduled. For this reason, we have developed and evaluated a hybrid scheduling method that we call priority segmentation. Priority segmentation breaks the problem up into sub-problems based on the priority of the tasks, finds an optimal solution to each sub-problem, and then combines the sub-problem solutions. We present preliminary results evaluating this approach for different sub-problem sizes on randomly generated satellite imaging problems. Our preliminary results indicate that priority segmentation produces solutions that are an improvement over a greedy solution when the problems are highly over constrained. When the problems are only slightly over constrained, the amount of improvement over greedy is very small. Although this approach has provided satisfactory results for our customersʹ highly over-constrained problems, we are pursuing alternative methods that provide better performance; both in terms of expected solution quality and computational requirements.