Tractable conservative Constraint Satisfaction Problems

  • Authors:
  • Andrei A. Bulatov

  • Affiliations:
  • -

  • Venue:
  • LICS '03 Proceedings of the 18th Annual IEEE Symposium on Logic in Computer Science
  • Year:
  • 2003

Quantified Score

Hi-index 0.02

Visualization

Abstract

In a constraint satisfaction problem (CSP) the aim is tofind an assignment of values to a given set of variables, subjectto specified constraints. The CSP is known to be NP-completein general. However, certain restrictions on theform of the allowed constraints can lead to problems solvablein polynomial time. Such restrictions are usually imposedby specifying a constraint language. The principalresearch direction aims to distinguish those constraint languageswhich give rise to tractable CSPs from those whichdo not.We achieve this goal for the widely used variant of theCSP, in which the set of values for each individual variablecan be restricted arbitrarily. Restrictions of this typecan be expressed by including in a constraint language allpossible unary constraints. Constraint languages containingall unary constraints will be called conservative. Wecompletely characterize conservative constraint languagesthat give rise to CSP classes solvable in polynomial time.In particular, this result allows us to obtain a complete descriptionof those (directed) graphs H for which the LISTH-COLORING problem is polynomial time solvable.