Towards automated reformulation of specications

Several state-of-the-art systems and languages for constraint solving adopt a clear separation between the specification of a problem and its instances. Some of them additionally perform a limited form of reasoning on the spec, with the goal of choosing the most appropriate solver. In this paper we propose a more sophisticated form of reasoning on problem specs, with the goal of reformulating them so that they are more efficiently solvable. To this end, we present a reformulation technique that highlights constraints that can be safely “delayed”, and solved afterwards. Our main contribution is the characterization (with soundness proof) of safe-delay constraints wrt a syntactic criterion on the spec, thus obtaining a mechanism for the automated reformulation of specs applicable to a great variety of problems, e.g., graph coloring and job shop scheduling. Another contribution is a preliminary experimentation on the effectiveness of the proposed technique, which reveals promising time savings.