A memetic algorithm for a bi-objective and stochastic CARP - Laboratoire de Mathematiques Accéder directement au contenu
Communication Dans Un Congrès Année : 2005

A memetic algorithm for a bi-objective and stochastic CARP

Résumé

The Capacitated Arc Routing Problem or CARP is raised for instance by urban refuse collection. The paper shows how to modify an efficient memetic algorithm for the CARP to tackle two other problems: a CARP with stochastic demands and a bi-objective CARP in which the second objective is the cost of the longest trip or makespan. The two new algorithms are finally combined for the bi-objective case with stochastic demands. The resulting MOGA uses average demands to compute a priori a set of non-dominated solutions that will resist random demand fluctuations on the field. Under mild simplifying assumptions, mathematical expressions are derived for the two objectives (expected total cost and expected makespan), thus avoiding simulation or noising techniques to evaluate chromosomes. The algorithm is evaluated on classical CARP instances, modified to have random demands. The non-dominated solutions obtained are very robust, with null standard deviations in most cases. A simulation phase is performed a posteriori to evaluate the behavior of solutions on the field. It confirms the expected values computed by the MOGA, the accuracy of mathematical expressions and the validity of assumptions. The general method presented could be applied to other combinatorial optimization problems.
Fichier non déposé

Dates et versions

hal-00069455 , version 1 (17-05-2006)

Identifiants

  • HAL Id : hal-00069455 , version 1

Citer

Gérard Fleury, Philippe Ph. Lacomme, Christian Prins, Marc Sevaux. A memetic algorithm for a bi-objective and stochastic CARP. 6th Metaheuristics International Conference, MIC 2005, 2005, Vienna, Austria. pp.368-37. ⟨hal-00069455⟩
192 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More