Gallai-Edmonds decomposition as a pruning technique

Abstract

We introduce a generic propagation mechanism for constraint programming. The method is based on the results of matching theory which is a mature and well-studied subject of graph theory. A first benefit of our new pruning technique comes from the fact that it can be applied on several global constraints whose solution is representable by a matching in a particular graph. In this work we describe a filtering scheme for such a family based on the Gallai-Edmonds Structure Theorem. In a number of important cases our method achieves hyper-arc consistency in polynomial time.