This dissertation focuses on service scheduling and transshipment problems. The study of service scheduling is motivated by decisions facing service planners, who must inspect and maintain geographically dispersed infrastructure facilities. We study the problem of deciding which operations a service unit must perform at each customer location, given the sequence in which the unit periodically visits these locations. Each customer requires multiple service operations, and each operation has a time-varying completion or penalty cost that depends on the previous service time. The goal is to schedule the service start time for each customer and select the operations to perform so as to minimize the total completion cost. We first discuss how to solve a special case of this problem in which each site is visited only once per service cycle. We formulate this problem as a discrete time indexed network flow problem and prove that it is NP-hard in the ordinary sense. Then, we represent the problem as a multidimensional shortest path problem with path-dependent arc lengths. In this structure, arc costs depend on the total time spent for all customers. The resulting formulation is solvable via algorithms that have pseudo-polynomial run times. Computational results show that the shortest path approach outperformed the general network flow model. We then analyze the general case of this problem, in which each site can be visited more than once and prove that the problem is NP-Hard in the strong sense. We discuss the valid cuts and describe the preprocessor that reduces the problem size. Next, we examine an application to the general case of the problem and develop a fast and effective heuristic procedure that repeatedly applies the shortest path approach to subsequences that do not visit any customer more than once. Computational results for several problem instances show that the proposed heuristic identifies near optimal results very quickly, whereas a general purpose integer-programming solver (CPLEX) is not able to find an optimal solution even after many hours of computational time. Then we focus on techniques such as problem reduction, branching variables, and subdividing problem to smaller problems to get better solution times for the actual problem. Computational results show that these techniques can improve solution times substantially. Finally, we study a transshipment problem, in which the shipments need to be transported from their origin to destination and are subject to the logical and physical transportation network on which they rely. We consider a space-time network that allows one to formulate the problem as a multi-commodity network flow problem with additional side constraints and show the complexity results. We propose alternative models and propose algorithms for lower and upper bound calculations. / text

O problema de escalonamento de veículos com múltiplas garagens (MDVSP, do inglês Multi-Depot Vehicle Scheduling Problem) é um problema clássico de logística e transportes. O MDVSP também é a base para a solução de vários problemas correlatos, tais como o problema de escalonamento de veículos em tempo-real e soluções integradas com o escalonamento de veículos, tais como o escalonamento da tripulação e otimização da tabela de horários. Desta forma, aprimorar a solução deste problema pode ser considerado de grande relevância, a qual permitirá resolver grandes instâncias reais de forma eficiente, bem como permitir a solução de problemas correlatos. O objetivo desta dissertação é verificar a aplicabilidade da utilização da rede tempo-espaço e do método de geração de colunas modificado proposto, para a solução deste problema, e de sua variante com frota heterogênea, considerando grandes instâncias. Diversos testes foram realizados utilizando o gerador de instâncias aleatórias com base na distribuição de demandas proposto. Grandes instâncias, envolvendo milhares de viagens (entre 500-10.000) e dezenas de garagens (4-128) são resolvidas em tempos razoáveis. / The multiple-depot vehicle-scheduling problem (MDVSP) is a classic logistic and transportation problem. The MDVSP is also a subproblem for solving various related problems, such as the real time vehicle scheduling problem, disruption management; and integrated problems such as the vehicle and crew scheduling problems. Although several mathematical and solution method have been developed in the literature, large instances (involving thousands of trips and several depots) are still difficult to solve in a reasonable time. The objective of this research work is to verify the applicability of the use of the space-time network towards obtaining good solutions for large instances in short time. Time-space network was suggested by Kliewer et al (2006), and it is positioned with respect to two-dimensional axes, one representing time and the other one space or stations. The arcs represent deadheading movements; and waiting periods in the same station. Solution methods for the MDVS combining time space with integer linear programming solvers and column generation were developed. Extensive testing was carried out using random generated instances, based on demands distribution. Large instances, involving thousands of trips (between 1,000-10,000) and dozen (4-64) depots, are solved in reasonable times.

O problema de escalonamento de veículos com múltiplas garagens (MDVSP, do inglês Multi-Depot Vehicle Scheduling Problem) é um problema clássico de logística e transportes. O MDVSP também é a base para a solução de vários problemas correlatos, tais como o problema de escalonamento de veículos em tempo-real e soluções integradas com o escalonamento de veículos, tais como o escalonamento da tripulação e otimização da tabela de horários. Desta forma, aprimorar a solução deste problema pode ser considerado de grande relevância, a qual permitirá resolver grandes instâncias reais de forma eficiente, bem como permitir a solução de problemas correlatos. O objetivo desta dissertação é verificar a aplicabilidade da utilização da rede tempo-espaço e do método de geração de colunas modificado proposto, para a solução deste problema, e de sua variante com frota heterogênea, considerando grandes instâncias. Diversos testes foram realizados utilizando o gerador de instâncias aleatórias com base na distribuição de demandas proposto. Grandes instâncias, envolvendo milhares de viagens (entre 500-10.000) e dezenas de garagens (4-128) são resolvidas em tempos razoáveis. / The multiple-depot vehicle-scheduling problem (MDVSP) is a classic logistic and transportation problem. The MDVSP is also a subproblem for solving various related problems, such as the real time vehicle scheduling problem, disruption management; and integrated problems such as the vehicle and crew scheduling problems. Although several mathematical and solution method have been developed in the literature, large instances (involving thousands of trips and several depots) are still difficult to solve in a reasonable time. The objective of this research work is to verify the applicability of the use of the space-time network towards obtaining good solutions for large instances in short time. Time-space network was suggested by Kliewer et al (2006), and it is positioned with respect to two-dimensional axes, one representing time and the other one space or stations. The arcs represent deadheading movements; and waiting periods in the same station. Solution methods for the MDVS combining time space with integer linear programming solvers and column generation were developed. Extensive testing was carried out using random generated instances, based on demands distribution. Large instances, involving thousands of trips (between 1,000-10,000) and dozen (4-64) depots, are solved in reasonable times.

O problema de escalonamento de veículos com múltiplas garagens (MDVSP, do inglês Multi-Depot Vehicle Scheduling Problem) é um problema clássico de logística e transportes. O MDVSP também é a base para a solução de vários problemas correlatos, tais como o problema de escalonamento de veículos em tempo-real e soluções integradas com o escalonamento de veículos, tais como o escalonamento da tripulação e otimização da tabela de horários. Desta forma, aprimorar a solução deste problema pode ser considerado de grande relevância, a qual permitirá resolver grandes instâncias reais de forma eficiente, bem como permitir a solução de problemas correlatos. O objetivo desta dissertação é verificar a aplicabilidade da utilização da rede tempo-espaço e do método de geração de colunas modificado proposto, para a solução deste problema, e de sua variante com frota heterogênea, considerando grandes instâncias. Diversos testes foram realizados utilizando o gerador de instâncias aleatórias com base na distribuição de demandas proposto. Grandes instâncias, envolvendo milhares de viagens (entre 500-10.000) e dezenas de garagens (4-128) são resolvidas em tempos razoáveis. / The multiple-depot vehicle-scheduling problem (MDVSP) is a classic logistic and transportation problem. The MDVSP is also a subproblem for solving various related problems, such as the real time vehicle scheduling problem, disruption management; and integrated problems such as the vehicle and crew scheduling problems. Although several mathematical and solution method have been developed in the literature, large instances (involving thousands of trips and several depots) are still difficult to solve in a reasonable time. The objective of this research work is to verify the applicability of the use of the space-time network towards obtaining good solutions for large instances in short time. Time-space network was suggested by Kliewer et al (2006), and it is positioned with respect to two-dimensional axes, one representing time and the other one space or stations. The arcs represent deadheading movements; and waiting periods in the same station. Solution methods for the MDVS combining time space with integer linear programming solvers and column generation were developed. Extensive testing was carried out using random generated instances, based on demands distribution. Large instances, involving thousands of trips (between 1,000-10,000) and dozen (4-64) depots, are solved in reasonable times.