Heurísticas Busca Tabu para o problema de programação de tripulações de ônibus urbano

• Autor(a) principal: Marinho, Euler Horta
• Data de Publicação: 2008
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da Universidade Federal Fluminense (RIUFF)
• Texto Completo: https://app.uff.br/riuff/handle/1/17835
• topic: Metaheurística; Busca Tabu; Transporte urbano; Programação de tripulações de ônibus urbano; Relaxação adaptativa; Reconexão por caminhos; Metaheurísiticas; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::TEORIA DA COMPUTACAO::COMPUTABILIDADE E MODELOS DE COMPUTACAO

The Bus Driver Scheduling Problem (BDSP) attempts to setup work schedules for drivers and conductors, with the best possible utilization of resources. The solving of BDSP is a subject for great relevance. It is due to the fact that most of operational costs involved in the process have come from labor costs. The BDSP belongs to a class of combinatorial problems called staff scheduling or rostering. Such problems are difficult to solve, since a large number of labor and operational rules should be observed. Moreover, they belong to the class NP-Hard, since the Set Covering and Set Partitioning problems can be reduced to this type of problem. Thus, the exclusive use of exact methods is limited to solving models with a small number of different restrictions. For realistic models, with a large number of different restrictions. For realistic models, with a large number of different restrictions, one common approach is to use heuristic techniques. In this work a solution method, based on the Tabu Search metaheuristic, was proposed for solving the BDSP for a Brazilian company of public transport from Belo Horizonte city. Firstly, a strategy named BT-FI-IC was implemented and compared to an algorithm fro literature based on Variable Neighborhood Search method (VNS-RTL). The results had demonstrated the superiority of the BT-FI-IC strategy in relation to quality of the final solution and speed in producing good quality solutions. With the aim of compare the Tabu Search to a mathematical programming method, a complete column generation approach was implemented to solve the BDSP. For this, eleven artificially created instances, with smaller dimensions were created from the real problem. The BT-FI-IC method was improved with an Adaptive Relaxatiom mechanism (BTAR-FI-IC) and with a Path Relinking mechanism (BTPR-FI-IC). The solutions obtained through the three strategies were compared to the optimal solutions and also to the best known solution for the real problem. The BTAR-FI-IC method was the most efficient one, showing robustness and ability to produce good solutions quickly.