Finding solutions by finding inconsistencies - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2018

Finding solutions by finding inconsistencies

Ghiles Ziat
Marie Pelleau
Antoine Miné

Résumé

In continuous constraint programming, the solving process alternates propagation steps, which reduce the search space according to the constraints, and branching steps. In practice, the solvers spend a lot of computation time in propagation to separate feasible and infeasi-ble parts of the search space. The constraint propagators cut the search space into two subspaces: the inconsistent one, which can be discarded, and the consistent one, which may contain solutions and where the search continues. The status of all this consistent subspace is thus indeterminate. In this article, we introduce a new step called elimination. It refines the analysis of the consistent subspace by dividing it into an indeterminate one, where the search must continue, and a satisfied one, where the constraints are always satisfied. The latter can be stored and removed from the search process. Elimination relies on the propagation of the negation of the constraints, and a new difference operator to efficiently compute the obtained set as an union of boxes, thus it uses the same representations and algorithms as those already existing in the solvers. Combined with propagation, elimination allows the solver to focus on the frontiers of the constraints, which is the core difficult part of the problem. We have implemented our method in the AbSolute solver, and present experimental results on classic benchmarks with good performances.

Domaines

Autre [cs.OH]
Fichier principal
Vignette du fichier
finding-solutions-finding.pdf (360.25 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01885769 , version 1 (02-10-2018)

Identifiants

  • HAL Id : hal-01885769 , version 1

Citer

Ghiles Ziat, Marie Pelleau, Charlotte Truchet, Antoine Miné. Finding solutions by finding inconsistencies. CP 2018 - 24th International Conference on Principles and Practice of Constraint Programming, Aug 2018, Lille, France. pp.1-16. ⟨hal-01885769⟩
181 Consultations
266 Téléchargements

Partager

Gmail Facebook X LinkedIn More