Recollection: an Alternative Restoration Technique for Constraint Programming Systems

Abstract: Search is a key service within constraint programming systems, and it demands the restoration of previously accessed states during the exploration of a search tree. Restoration proceeds either bottom-up within the tree to roll back previously performed operations using a trail, or top-down to redo them, starting from a previously stored state and using suitable information stored along the way. In this paper, we elucidate existing restoration techniques using a pair of abstract methods and employ them to present a new technique that we call recollection. The proposed technique stores the variables that were affected by constraint propagation during fix points reasoning steps, and it conducts neither operation roll-back nor recomputation, while consuming much less memory than storing previous visited states. We implemented this idea as a prototype within the Gecode solver. An empirical evaluation reveals that constraint problems with expensive propagation and frequent failures can benefit from recollection with respect to runtime at the expense of a marginal increase in memory consumption, comparing with the most competitive variant of recomputation.