Resumen de Reformulation of constraint models into SMT
In this thesis we focus on reformulate constraint satisfaction problems (CSP) into SAT Modulo Theories (SMT). SMT is an extension of SAT where the literals appearing in the formulas are not restricted to contain only propositional variables, instead they can have predicates from other theories, e.g., linear integer arithmetic. We present two systems developed to reformulate CSPs into SMT (fzn2smt and WSimply). The first one, reads instances written in FlatZinc and solved using an external SMT solver, and it has been extended to also solve optimization problems (COP) which are not supported by SMT solvers. The second one reads CSP, COP and weighted CSPs (WCSP) written in its own high level declarative language, which in addition to reformulate into SMT also reformulates into pseudo-Boolean and linear programming formats. We also present an incremental optimization algorithm based on using Binary Decision Diagrams (BDD) to solve WCSPs.
