No abstract

This article deals with the vehicle routing problem with time windows (VRPTW). This problem consists in determining a leastcost set of trips to serve customers during specific time windows. The proposed solution method is a memetic algorithm (MA), a genetic algorithm hybridised with a local search. Contrary to most papers on the VRPTW, which minimize first the number of vehicles, our method is also able to minimize the total distance travelled. The results on 56 classical instances are compared to those of the best metaheuristics. The efficiency of the MA is similar for the classical criterion, but it becomes the best algorithm available for the total distance, being much faster and improving 20 best-known solutions. Résumé. Cet article concerne le problème de tournées de véhicules avec fenêtres horaires (Vehicle Routing Problem with Time Windows ou VRPTW). Ce problème consisteà déterminer un ensemble de tournées de coût total minimal pour servir des clients dans des fenêtres horaires spécifiques. Nous proposons un algorithme mémétique (MA, algorithme génétique hybridé avec une recherche locale) pour le résoudre. Contrairementà la plupart des articles sur le VRPTW, qui minimisent en priorité le nombre de véhicules, notre méthode peut aussi minimiser la distance totale parcourue. Les résultats sur 56 problèmes-tests classiques sont comparésà ceux des meilleures métaheuristiques. Pour le critère classique, le MA offre des performances similaires, mais il devient le meilleur algorithme disponible pour la distance totale, enétant bien plus rapide et en améliorant 20 des meilleures solutions connues.

No abstract

This paper proposes lower and upper bounds for the split-delivery capacitated arc-routing problem (SDCARP), a variant of the capacitated arc-routing problem in which an edge can be serviced by several vehicles. Recent papers on related problems in node routing have shown that this policy can bring significant savings. It is also more realistic in applications such as urban refuse collection, where a vehicle can become full in the middle of a street segment. This work presents the first lower bound for the SDCARP, computed with a cutting plane algorithm and an evolutionary local search reinforced by a multistart procedure and a variable neighborhood descent. Tests on 126 instances show that the new metaheuristic outperforms on average a published memetic algorithm; achieves small deviations to the lower bound; and finds 44 optima, including 10 new ones.