Modeling and Solving Semiring Constraint Satisfaction Problems by Transformation to Weighted Semiring Max-SAT

摘要

We present a variant of the Weighted Maximum Satisfiability Problem (Weighted Max-SAT), which is a modeling of the Semiring Constraint Satisfaction framework. We show how to encode a Semiring Constraint Satisfaction Problem (SCSP) into an instance of a propositional Weighted Max-SAT, and call the encoding Weighted Semiring Max-SAT (WS-Max-SAT). The clauses in our encoding are highly structured and we exploit this feature to develop two algorithms for solving WS-Max-SAT: an incomplete algorithm based on the well-known GSAT algorithm for Max-SAT, and a branch-and-bound algorithm which is complete. Our preliminary experiments show that the translation of SCSP into WS-Max-SAT is feasible, and that our branch-and-bound algorithm performs surprisingly well. We aim in future to combine the natural flexible representation of the SCSP framework with the inherent efficiencies of SAT solvers by adjusting existing SAT solvers to solve WS-Max-SAT.