From Set Constraint Models to SAT Instances
Résumé
On the one hand, Constraint Satisfaction Problems(CSP) are a declarative and expressive approach for modelingproblems. On the other hand, propositional satisfiability problem(SAT) solvers can handle huge SAT instances up to millions ofvariables and clauses. In this article, we present an approach fortaking advantage of both CSP modeling and SAT solving. Ourtechnique consists in expressively modeling set constraint prob-lems as CSPs that are automatically treated by some reductionrules to remove values that do not participate in any solution.These reduced CSPs are then encoded into ”good” SAT instancesthat can be solved by standard SAT solvers. We illustrate ourtechnique on the Sports Tournament Scheduling problem, andwe show that we obtain competitive results compared to an ad-hoc solver. Our technique is simpler, more expressive, and lesserror-prone than direct SAT modeling. The SAT instances thatwe automatically generate are rather small and can efficientlybe solved up to huge instances. Moreover, the reduction phaseenables to push back the limits and treat even larger problems.
Domaines
Informatique [cs]Origine | Fichiers produits par l'(les) auteur(s) |
---|