Uso de rotas elementares no CVRP

• Autor(a) principal: Pecin, Diego Galindo
• Data de Publicação: 2010
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFG
• dARK ID: ark:/38995/00130000006cc
• Texto Completo: http://repositorio.bc.ufg.br/tede/handle/tede/12516
• topic: CVRP; Enumeração; ESPPCC; ESPPRC; Geração de colunas; Rotas elementares; Column generation; CVRP; Elementary routes; ESPPCC; ESPPRC; CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO

This dissertation addresses the optimization of the Elementary Shortest Path Problem with a Capacity Constraint (ESPPCC) and describes algorithms for its resolution that make use of concepts such as Label-Setting, Bidirectional Dynamic Programming and Decremental State Space Relaxation. These algorithms were used in a robust CVRP’s Branch-and Cut-and-Price framework as the column generation mechanism. The resulting BCP was used to obtain results (lower bounds, processing time and the number of branching nodes generated) to several CVRP’s test instances. These results are compared with previous ones obtained with the original BCP, which is based on k-cycle elimination. Elementary routes are also explored in a route enumeration context, which allows the enumeration of all possible relevant elementary routes, i.e., all routes that have a chance of being part of an optimal CVRP’s solution. If the number of relevant routes is not too large (say, in the range of tenths of thousands), the overall problem may be solved by feeding a general MIP solver with a set-partition formulation containing only those routes. If this set-partition can be solved, the optimal solution will be found and no branch will be necessary. Sometimes this leads to very significant speedups when compared to traditional branch strategies.