Failed Value Consistencies for Constraint Satisfaction

Christophe Lecoutre (1) , Olivier Roussel (1)

1 CRIL - Centre de Recherche en Informatique de Lens

In constraint satisfaction, basic inferences rely on some properties of constraint networks, called consistencies, that allow the identification of incon- sistent instantiations (also called nogoods). Two main families of consistencies have been introduced so far: those that permit us to reason from variables such as (i, j)-consistency and those that permit us to reason from constraints such as relational (i, j)-consistency. This paper introduces a new family of consistencies based on the concept of failed value (a value pruned during search). This family is orthogonal to previous ones.

Domaines: Intelligence artificielle [cs.AI]

Christophe Lecoutre, Olivier Roussel. Failed Value Consistencies for Constraint Satisfaction. 15th International Conference on Principles and Practice of Constraint Programming (CP'09), 2009, Lisbon, Portugal. pp.554-568. ⟨hal-00865327⟩