Améliorer la propagation : l'Importance d'être Inconsistant

Abstract : Abstract Classical CSP solving methods often alternate two steps : propagation and exploration. Propagation reduces the domains of the variables according to the constraints. It can be seen as a discrimination of the search space in two sub-spaces : the inconsistent one that can be deleted from the solving process, and the undetermined one which may contain the solutions. The exploration step divides the undetermined sub-space into several sub-spaces in which the search continues. This step is usually implemented in solvers by split heuristics relying onto the domains of the variables and/or the constraints of the problem. This article introduces a new step into the solving process called elimination. It divides the search space into two sub-spaces : the undetermined one and the consistent one. It allows the solver to benefit more from the constraints, thus obtaining more significative frontiers for the exploration. It is based on a key observation : the consistent part of a problem is equivalent to the inconsistent part of the complementary problem. This new step is implemented in the AbSolute continuous solver. This solver combines methods from Abstract Interpretation and Constraint Programming. Our elimination technique can be easily added in it. Preliminary results show significative improvements of the solving process.
Document type :
Conference papers
Complete list of metadatas

Cited literature [17 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01735167
Contributor : Marie Pelleau <>
Submitted on : Thursday, March 15, 2018 - 3:15:14 PM
Last modification on : Tuesday, March 26, 2019 - 9:25:22 AM
Long-term archiving on : Monday, September 10, 2018 - 10:20:56 PM

File

limportance-detre-inconsistant...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01735167, version 1

Citation

Ghiles Ziat, Marie Pelleau, Charlotte Truchet, Antoine Miné. Améliorer la propagation : l'Importance d'être Inconsistant. Treizièmes journées Francophones de Programmation par Contraintes, Jun 2017, Montreuil sur Mer, France. ⟨hal-01735167⟩

Share

Metrics

Record views

125

Files downloads

43