Heuristic approaches to the double vehicle routing problem with multiple stacks

• Ano de defesa: 2017
• Autor(a) principal: Silveira, Ulisses Eduardo Ferreira da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Heurística; Programação heurística; Veículos a motor; Levantamento e carregamento; Ciência da Computação
• Link de acesso: http://www.locus.ufv.br/handle/123456789/11596

In a world, that in a fast pace, has become increasingly needed in consumable and nonconsumable goods, the logistics in the transportation of these products has been put to the test, being one of the most important stages in the relationship between the pro-duction process and the end user. It is said that at least 30% of the costs between the industry and the end user are solely determined by the cost of transportation. A novel problem arose followed by a question that was encountered in a real-life scenario. The Double Routing Vehicle Problem with Multiple Stacks (DVRPMS) consists in a Dou-ble Traveling Salesman Problem with Multiple Stacks (DTSPMS) with multiple vehicles. Both problems appeared for the urgent need of optimizing intermodal transportation in the european context. It consists in gathering costumer inquires from a pickup region and loading them in a set of stacks inside a container that must not be rearranged for security reasons. The container moves to a delivery region and the items gathered must be delivered according the last-in-first-out policy of the stacks. In this work, four heuristics were proposed based on the Iterated Local Search (ILS), Variable Neighborhood Descent and Simulated Annealing (SA) metaheuristics. The DVRPMS was extended to a modi-fied version where the items offered are bigger than the vehicle fleet capacities. An exact model approach is proposed and three other heuristics, based on the ILS, SA and Tabu Search are proposed and tested. The approaches presented in this work were tested by computational experiments and a statistical analysis was made to chose the best com-bination of parameters. Good results were found, providing a better average than the current literature.