Reusing CSP Propagators for QCSPs

摘要

Quantified Constraint Satisfaction Problems are considerably more difficult to solve than classical CSP and the pruning obtained by local consistency is of crucial importance. In this paper, instead of designing specific consistency operators for constraints w.r.t each possible quantification pattern, we propose to build them by relying on classical existential propagators and a few analysis of some mathematical properties of the constraints. It allows to reuse a large set of constraints already carefully implemented in existing solvers. Moreover, multiple levels of consistency for quantified constraint can be defined by choosing which analysis to use. This can be used to control the complexity of the pruning effort. We also introduce QeCode, a full-featured publicly available quantified constraint solver, built on top of Gecode.