Population improvement with data oriented genetic operators - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2001

Population improvement with data oriented genetic operators

Résumé

In this paper, we consider various particular cases of the one-machine scheduling problem. The general problem considered is denoted by 1/r_i,S_{sd}/sum_{w_iT_i} in the Graham's notation, i.e. the one-machine problem with non identical release dates, sequence dependent set-up times and the weighted sum of tardiness as measure of performance. We will also consider the makespan criterion. We propose new genetic operators called "data oriented" because they use not only the permutation (order of the operations) contained in the chromosome but also an analysis of the corresponding solution. One family of data oriented operators focuses on the set-up times minimization and thus indirectly on the tardiness costs, the other one focuses directly on the tardiness cost. When the release dates are identical, each chromosome is the direct representation of the solution (direct encoding). When the release dates are not identical, the only way to remain efficient is to use active generators such as Baker's or Carlier's generators to convert the permutation contained in the chromosome (considered now as priority order of the operations) into an active schedule (indirect encoding). Only the tardiness oriented operators may be used in the latter case. Experimental results show the efficiency of the new proposed operators.
Fichier non déposé

Dates et versions

inria-00100512 , version 1 (26-09-2006)

Identifiants

  • HAL Id : inria-00100512 , version 1

Citer

Marie-Claude Portmann, Mohamed Ali Aloulou. Population improvement with data oriented genetic operators. Proceedings of the Genetic and Evolutionary Computation Conference - GECCO'2001, 2001, San Francisco, USA, pp.339-346. ⟨inria-00100512⟩
45 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More