In this paper we present a novel "modular" approach for (weighted partial) MaxSAT Modulo Theories. The main idea is to combine a lazy SMT solver with a purely-propositional (weighted partial) MaxSAT solver, by making them exchange information iteratively: the former produces an increasing set of theory lemmas which are used by the latter to progressively refine an approximation of the final subset of the soft clauses, which is eventually returned as output. The approach has several practical features. First, it is independent from the theories addressed. Second, it is simple to implement and to update, since both SMT and MaxSAT solvers can be used as blackboxes. Third, it can be interfaced with external MaxSAT and SMT solvers in a plug-and-play manner, so that to benefit for free of tools which are or will be made available. We have implemented our approach on top of the MathSAT5 SMT solver and of a selection of external MaxSAT solvers, and we have evaluated it by means of an extensive empirical test on SMT-LIB benchmarks. The results confirm the validity and potential of this approach. © 2013 Springer-Verlag.
CITATION STYLE
Cimatti, A., Griggio, A., Schaafsma, B. J., & Sebastiani, R. (2013). A modular approach to MaxSAT modulo theories. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7962 LNCS, pp. 150–165). https://doi.org/10.1007/978-3-642-39071-5_12
Mendeley helps you to discover research relevant for your work.