AC-3d an efficient arc-consistency algorithm with a low space-complexity

Abstract

Arc-consistency algorithms prune the search-space of Constraint Satisfaction Problems (CSPs). They use support-checks to find out about the properties of CSPs. Their arc-heuristics select the constraint and their domain-heuristics select the values for the next support-check. We shall combine AC-3 and DEE and equip the resulting hybrid with a double-support domain-heuristic. The resulting hybrid AC-3d is easy to implement and requires the same data structures as AC-3 thereby improving on AC-7’s space-complexity. We shall present experimental results which indicate that AC-3 can compete with AC-7.