Skip to Main content Skip to Navigation

Tabu-NG : hybridation de programmation par contraintes et recherche locale pour la résolution de CSP

Abstract : A very large number of combinatorial problems belong to the family of Constraint Satisfaction Problems (CSP): configuration, planning, scheduling, resource allocation... These problems share a common description that generally allows a clear and intuitive modeling. In this thesis, we proposed and studied a new hybrid method for solving CSPs. We named this method Tabu-NG for Tabu Search based on NoGood. The name is a bit simplistic because it is a hybrid of filtering algorithm, constraint propagation, Tabu Search and nogood management. The method was applied to two types of problems. The first is the Frequency Assignment Problem (FAP) in military radio networks, particularly the problems proposed from 1993 (benchmarks of the European project CALMA) until 2010 (benchmarks of a DGA project). The second is the academic problem of k-coloring of graphs on the DIMACS instances. The method has improved some high scores currently known. In both problems we dealt with unary and binary constraints, and also with n-ary constraints and function optimization under constraints for FAP. The principles of Tabu-NG are general and can be applied to other CSPs. It can also accommodate specific heuristics to problems, we practiced it on the problems cited, and in this sense we believe we can qualify the method as metaheuristic without abusing this definition.
Complete list of metadatas
Contributor : Mohammad Dib <>
Submitted on : Wednesday, November 23, 2011 - 5:21:43 PM
Last modification on : Tuesday, January 8, 2013 - 11:31:59 AM
Long-term archiving on: : Friday, February 24, 2012 - 2:31:12 AM


  • HAL Id : tel-00644178, version 1



Mohammad Dib. Tabu-NG : hybridation de programmation par contraintes et recherche locale pour la résolution de CSP. Algorithme et structure de données [cs.DS]. Université de Technologie de Belfort-Montbeliard, 2010. Français. ⟨tel-00644178⟩



Record views


Files downloads