O foco central desta tese é o desenvolvimento de algoritmos para o problema de recoloração convexa de caminhos. Neste problema, é dado um caminho cujos vértices estão coloridos arbitrariamente, e o objetivo é recolorir o menor número possível de vértices de modo a obter uma coloração convexa. Dizemos que uma coloração de um grafo é convexa se, para cada cor, o subgrafo induzido pelos vértices dessa cor é conexo. Sabe-se que este problema é NP-difícil. Associamos a este problema um poliedro, e estudamos sua estrutura facial, com vistas ao desenvolvimento de um algoritmo. Mostramos várias inequações válidas para este poliedro, e provamos que várias delas definem facetas. Apresentamos um algoritmo de programação dinâmica que resolve em tempo polinomial o problema da separação para uma classe grande de inequações que definem facetas. Implementamos um algoritmo branch-and-cut baseado nesses resultados, e realizamos testes computacionais com instâncias geradas aleatoriamente. Apresentamos adicionalmente uma heurística baseada numa formulação linear que obtivemos. Estudamos também um caso especial deste problema, no qual as instâncias consistem em caminhos coloridos, onde cada cor ocorre no máximo duas vezes. Apresentamos um algoritmo de 3/2-aproximação para este caso, que é também NP-difícil. Para o caso geral, é conhecido na literatura um algoritmo de 2-aproximação. / The focus of this thesis is the design of algorithms for the convex recoloring problem on paths. In this problem, the instance consists of a path whose vertices are arbitrarily colored, and the objective is to recolor the least number of vertices so as to obtain a convex coloring.Acoloring of a graph is convex if, for each color, the subgraph induced by the vertices of this color is connected. This problem is known to be NP-hard. We associate a polyhedron to this problem and investigate its facial structure. We show various classes of valid inequalities for this polyhedron and prove that many of them define facets.We present a polynomial-time dynamic programming algorithm that solves, in polynomial time, the separation problem for a large class of facet-defining inequalities.We report on the computational experiments with a branch-and-cut algorithm that we propose for the problem. Additionally, we present a heuristic that is based on a linear formulation for the problem. We also study a special case of this problem, restricted to instances consisting of colored paths in which each color occurs at most twice. For this case, which is also NP-hard, we present a 3/2-approximation algorithm. For the general case, it is known a 2-approximation algorithm.

We consider a one-warehouse multi-retailer system where the warehouse orders or receives from its supplier and replenishes multiple retailers with direct shipping or multi-stop routing over a finite time horizon. The warehouse has the knowledge of external (deterministic) demands at the retailers and manages their inventories while ensuring no stock-out. We consider two problems with direct shipping policy and two problems with routing policy. For the direct shipping policy, the problem is to determine the optimal replenishments for the warehouse and retailers such that the system-wide costs are minimized. In one problem, the warehouse decides about how much and when to ship to the retailers while in the other problem, inventory level of the retailer has to be raised up to a predetermined level whenever replenished. We propose strong mixed integer programming formulations for these problems. Computational experiments show that our formulations are better than their competitors and are very successful in solving the problems to optimality. For the routing policy, the problem is to decide on when and in what sequence to visit the retailers and how much to ship to a retailer so as to minimize system-wide costs. In one problem, the warehouse receives given amounts from its supplier while in the other the warehouse decides on its own replenishments. We propose branch-and-cut algorithms and heuristics based on strong formulations for both problems. Computational results reveal that our procedures perform better than their competitors in the literature for both problems.

La qualité de service (QoS), liée au partage des ressources, prend tout son sens dans le cadre des réseaux multimédias. L'intégration de celle-ci dans les protocoles de routage, nécessite la prise en compte des phénomènes de congestion. Cela a favorisé l'apparition du protocole MPLS (Multi Protocol Label Switching). Cette nouvelle technologie, grâce à son routage par LSP (Label Swithed Path), permet une gestion plus fine des ressources disponibles dans le réseau. Nous traitons en première partie de ce travail le problème du routage des LSPs dans les réseaux IP/MPLS. Nous en formulons une modélisation originale qui tient compte de la QoS. Nous proposons aussi une heuristique de résolution (ILSP-OLS-ACO) qui gère un grand nombre de contraintes opérationnelles, tels que la bande passante, les contraintes d'affinités ou de sécurité sur les LSPs. Celle-ci fournit des solutions quasi-optimales tout en permettant le passage à l'échelle (grands réseaux, milliers de LSPs). La deuxième partie de notre travail concerne la conception optimale de topologie d'accès. L'originalité de l'approche réside dans le fait de prendre en compte le trafic générés par les clients ainsi que les coûts des équipements. Nous élaborons une modélisation basée sur la programmation linéaire en nombres entiers. Nous proposons pour la résoudre une méthode exacte basée sur des techniques de « Branch and Cut ». Nous proposons aussi une heuristique combinant une technique de « clustering » et une technique de recherche locale, qui permet d'obtenir très rapidement des solutions quasi-optimales.

In this thesis, we use integer programming techniques to solve the resource constrained shortest path problem (RCSPP) which seeks a minimum cost path between two nodes in a directed graph subject to a finite set of resource constraints. Although NP-hard, the RCSPP is extremely useful in practice and often appears as a subproblem in many decomposition schemes for difficult optimization problems. We begin with a study of the RCSPP polytope for the single resource case and obtain several new valid inequality classes. Separation routines are provided, along with a polynomial time algorithm for constructing an auxiliary conflict graph which can be used to separate well known valid inequalities for the node packing polytope. We establish some facet defining conditions when the underlying graph is acyclic and develop a polynomial time sequential lifting algorithm which can be used to strengthen one of the inequality classes. Next, we outline a branch-and-cut algorithm for the RCSPP. We present preprocessing techniques and branching schemes which lead to strengthened linear programming relaxations and balanced search trees, and the majority of the new inequality classes are generalized to consider multiple resources. We describe a primal heuristic scheme that uses fractional solutions, along with the current incumbent, to search for new feasible solutions throughout the branch-and-bound tree. A computational study is conducted to evaluate several implementation choices, and the results demonstrate that our algorithm outperforms the default branch-and-cut algorithm of a leading integer programming software package. Finally, we consider the dial-a-flight problem (DAFP), a new vehicle routing problem that arises in the context of on-demand air transportation and is concerned with the scheduling of a set of travel requests for a single day of operations. The DAFP can be formulated as an integer multicommodity network flow model consisting of several RCSPPs linked together by set partitioning constraints which guarantee that all travel requests are satisfied. Therefore, we extend our branch-and-cut algorithm for the RCSPP to solve the DAFP. Computational experiments with practical instances provided by the DayJet Corporation verify that the extended algorithm also outperforms the default branch-and-cut algorithm of a leading integer programming software package.

O foco central desta tese é o desenvolvimento de algoritmos para o problema de recoloração convexa de caminhos. Neste problema, é dado um caminho cujos vértices estão coloridos arbitrariamente, e o objetivo é recolorir o menor número possível de vértices de modo a obter uma coloração convexa. Dizemos que uma coloração de um grafo é convexa se, para cada cor, o subgrafo induzido pelos vértices dessa cor é conexo. Sabe-se que este problema é NP-difícil. Associamos a este problema um poliedro, e estudamos sua estrutura facial, com vistas ao desenvolvimento de um algoritmo. Mostramos várias inequações válidas para este poliedro, e provamos que várias delas definem facetas. Apresentamos um algoritmo de programação dinâmica que resolve em tempo polinomial o problema da separação para uma classe grande de inequações que definem facetas. Implementamos um algoritmo branch-and-cut baseado nesses resultados, e realizamos testes computacionais com instâncias geradas aleatoriamente. Apresentamos adicionalmente uma heurística baseada numa formulação linear que obtivemos. Estudamos também um caso especial deste problema, no qual as instâncias consistem em caminhos coloridos, onde cada cor ocorre no máximo duas vezes. Apresentamos um algoritmo de 3/2-aproximação para este caso, que é também NP-difícil. Para o caso geral, é conhecido na literatura um algoritmo de 2-aproximação. / The focus of this thesis is the design of algorithms for the convex recoloring problem on paths. In this problem, the instance consists of a path whose vertices are arbitrarily colored, and the objective is to recolor the least number of vertices so as to obtain a convex coloring.Acoloring of a graph is convex if, for each color, the subgraph induced by the vertices of this color is connected. This problem is known to be NP-hard. We associate a polyhedron to this problem and investigate its facial structure. We show various classes of valid inequalities for this polyhedron and prove that many of them define facets.We present a polynomial-time dynamic programming algorithm that solves, in polynomial time, the separation problem for a large class of facet-defining inequalities.We report on the computational experiments with a branch-and-cut algorithm that we propose for the problem. Additionally, we present a heuristic that is based on a linear formulation for the problem. We also study a special case of this problem, restricted to instances consisting of colored paths in which each color occurs at most twice. For this case, which is also NP-hard, we present a 3/2-approximation algorithm. For the general case, it is known a 2-approximation algorithm.

Étant donnés un graphe simple non orienté G = (V, E) et un sommet particulier r dans V appelé racine, un arbre enraciné, ou r-arbre, de G est soit le graphe nul soit un arbre contenant r. Si un vecteur de capacités sur les sommets est donné, un sous-graphe de G est dit borné si le degré de chaque sommet dans le sous-graphe est inférieur ou égal à sa capacité. Soit w un vecteur de poids sur les arêtes et p un vecteur de profits sur les sommets. Le problème du r-arbre borné maximum (MBrT, de l’anglais Maximum Bounded r-Tree) consiste à trouver un r-arbre borné T = (U, F) de G tel que son poids soit maximisé. Si la contrainte de capacité du problème MBrT est relâchée, nous obtenons le problème du r-arbre maximum (MrT, de l’anglais Maximum r-Tree). Cette thèse contribue à l’étude des problèmes MBrT et MrT.Tout d’abord, ces deux problèmes sont formellement définis et leur complexité est étudiée. Nous présentons ensuite des polytopes associés ainsi qu’une formulation pour chacun d’entre eux. Par la suite, nous proposons plusieurs algorithmes combinatoires pour résoudre le problème MBrT (et donc le problème MrT) en temps polynomial sur les arbres, les cycles et les cactus. En particulier, un algorithme de programmation dynamique est utilisé pour résoudre le problème MBrT sur les arbres. Pour les cycles, nous sommes amenés a considérer trois cas différents pour lesquels le problem MBrT se réduit à certains problèmes polynomiaux. Pour les cactus, nous montrons tout d’abord que le problème MBrT peut être résolu en temps polynomial sur un type de graphes appelé cactus basis. En utilisant une série de décompositions en sous-problèmes sur les arbres et les cactus basis, nous obtenons un algorithme pour les graphes de type cactus.La deuxième partie de ce travail étudie la structure polyédrale de trois polytopes associés aux problèmes MBrT et MrT. Les deux premiers polytopes, Bxy(G,r,c) et Bx(G,r,c) sont associés au problème MBrT. Tous deux considèrent des variables sur les arêtes de G, mais seuls Bxy(G,r,c) possède également des variables sur les sommets de G. Le troisième polytope, Rx(G,r), est associé au problème MrT et repose uniquement sur les variables sur les arêtes. Pour chacun de ces trois polytopes, nous étudions sa dimension, caractérisons certaines inégalités définissant des facettes, et présentons les moyens possibles de décomposition. Nous introduisons également de nouvelles familles de contraintes. L’ajout de ces contraintes nous permettent de caractériser ces trois polytopes dans plusieurs classes de graphes.Pour finir, nous étudions les problèmes de séparation pour toutes les inégalités que nous avons trouvées jusqu’ici. Des algorithmes polynomiaux de séparation sont présentés, et lorsqu’un problème de séparation est NP-difficile, nous donnons des heuristiques de séparation. Tous les résultats théoriques développés dans ce travail sont implémentés dans plusieurs algorithmes de coupes et branchements auxquels une matheuristique est également jointe pour générer rapidement des solutions réalisables. Des expérimentations intensives ont été menées via le logiciel CPLEX afin de comparer les formulations renforcées et originales. Les résultats obtenus montrent de manière convaincante la force des formulations renforcées. / Given a simple undirected graph G = (V, E) with a so-called root node r in V, a rooted tree, or an r-tree, of G is either the empty graph, or a tree containing r. If a node-capacity vector c is given, then a subgraph of G is said to be bounded if the degree of each node in the subgraph does not exceed its capacity. Let w be an edge-weight vector and p a node-price vector. The Maximum Bounded r-Tree (MBrT) problem consists of finding a bounded r-tree T = (U, F) of G such that its weight is maximized. If the capacity constraint from the MBrT problem is relaxed, we then obtain the Maximum r-Tree (MrT) problem. This dissertation contributes to the study of the MBrT problem and the MrT problem.First we introduce the problems with their definitions and complexities. We define the associated polytopes along with a formulation for each of them. We present several polynomial-time combinatorial algorithms for both the MBrT problem (and thus the MrT problem) on trees, cycles and cactus graphs. Particularly, a dynamic-programming-based algorithm is used to solve the MBrT problem on trees, whereas on cycles we reduce it to some polynomially solvable problems in three different cases. For cactus graphs, we first show that the MBrT problem can be solved in polynomial time on a so-called cactus basis, then break down the problem on any cactus graph into a series of subproblems on trees and on cactus basis.The second part of this work investigates the polyhedral structure of three polytopes associated with the MBrT problem and the MrT problem, namely Bxy(G, r, c), Bx(G, r, c) and Rx(G, r). Bxy(G, r, c) and Bx(G, r, c) are polytopes associated with the MBrT problem, where Bxy(G, r, c) considers both edge- and node-indexed variables and Bx(G, r, c) considers only edge-indexed variables. Rx(G, r) is the polytope associated with the MrT problem that only considers edge-indexed variables. For each of the three polytopes, we study their dimensions, facets as well as possible ways of decomposition. We introduce some newly discovered constraints for each polytope, and show that these new constraints allow us to characterize them on several graph classes. Specifically, we provide characterization for Bxy (G, r, c) on cactus graphs with the help of a decomposition through 1-sum. On the other hand, a TDI-system that characterizes Bx(G,r,c) is given in each case of trees and cycles. The characterization of Rx(G,r) on trees and cycles then follows as an immediate result.Finally, we discuss the separation problems for all the inequalities we have found so far, and present algorithms or cut-generation heuristics accordingly. A couple of branch-and-cut frameworks are implemented to solve the MBrT problem together with a greedy-based matheuristic. We compare the performances of the enhanced formulations with the original formulations through intensive computational test, where the results demonstrate convincingly the strength of the enhanced formulations.

This thesis deals with the exact solution of large-scale minimum bisection problems via a semidefinite relaxation in a branch-and-cut framework. After reviewing known results on the underlying bisection cut polytope a study of new facet-defining inequalities is presented. They are derived from the known knapsack tree inequalities. We investigate strengthenings based on the new cluster weight polytope and present polynomial separation algorithms for special cases. The dual of the semidefinite relaxation of the minimum bisection problem is tackled in its equivalent form as an eigenvalue optimisation problem with the spectral bundle method. Implementational details regarding primal heuristics, branching rules, so-called support extensions for cutting planes and warm start are presented. We conclude with a computational study in which we show that our approach is competetive to state-of-the-art implementations using linear programming or semidefinite programming relaxations. / Diese Dissertation befasst sich mit der exakten Lösung großer Minimum Bisection Probleme über eine semidefinite Relaxierung in einem Branch-and-Cut Zugang. Nachdem bekannte Resultate zum zugrundeliegenden Bisection Cut Polytop dargestellt wurden, wird eine Studie neuer facettendefinierender Ungleichungen präsentiert. Diese werden von den bekannten Knapsack Tree Ungleichungen abgeleitet. Wir untersuchen Verstärkungen basierend auf dem neuen Cluster Weight Polytop und zeigen polynomiale Separierungsalgorithmen für Spezialfälle. Die Duale der semidefiniten Relaxierung des Minumum Bisection Problems wird in ihrer äquivalenten Form als Eigenwertoptimierungsproblem mit dem Spektralen Bündelverfahren bearbeitet. Details der Implementierung bezüglich primaler Heuristiken, Branchingregeln, sogenannter Supporterweiterungen für die Schnittebenen und Warmstart werden präsentiert. Wir beenden die Arbeit mit einer numerischen Studie, in der wir zeigen, dass unser Zugang konkurrenzfähig zu aktuellen Implementationen basierend auf linearen und semidefiniten Relaxierungen ist.

NP-hard problems of higher-dimensional orthogonal packing are considered. We look closer at their logical structure and show that they can be decomposed into problems of a smaller dimension with a special contiguous structure. This decomposition influences the modeling of the packing process, which results in three new solution approaches. Keeping this decomposition in mind, we model the smaller-dimensional problems in a single position-indexed formulation with non-overlapping inequalities serving as binding constraints. Thus, we come up with a new integer linear programming model, which we subject to polyhedral analysis. Furthermore, we establish general non-overlapping and density inequalities and prove under appropriate assumptions their facet-defining property for the convex hull of the integer solutions. Based on the proposed model and the strong inequalities, we develop a new branch-and-cut algorithm. Being a relaxation of the higher-dimensional problem, each of the smaller-dimensional problems is also relevant for different areas, e.g. for scheduling. To tackle any of these smaller-dimensional problems, we use a Gilmore-Gomory model, which is a Dantzig-Wolfe decomposition of the position-indexed formulation. In order to obtain a contiguous structure for the optimal solution, its basis matrix must have a consecutive 1's property. For construction of such matrices, we develop new branch-and-price algorithms which are distinguished by various strategies for the enumeration of partial solutions. We also prove some characteristics of partial solutions, which tighten the slave problem of column generation. For a nonlinear modeling of the higher-dimensional packing problems, we investigate state-of-the-art constraint programming approaches, modify them, and propose new dichotomy and intersection branching strategies. To tighten the constraint propagation, we introduce new pruning rules. For that, we apply 1D relaxation with intervals and forbidden pairs, an advanced bar relaxation, 2D slice relaxation, and 1D slice-bar relaxation with forbidden pairs. The new rules are based on the relaxation by the smaller-dimensional problems which, in turn, are replaced by a linear programming relaxation of the Gilmore-Gomory model. We conclude with a discussion of implementation issues and numerical studies of all proposed approaches. / Es werden NP-schwere höherdimensionale orthogonale Packungsprobleme betrachtet. Wir untersuchen ihre logische Struktur genauer und zeigen, dass sie sich in Probleme kleinerer Dimension mit einer speziellen Nachbarschaftsstruktur zerlegen lassen. Dies beeinflusst die Modellierung des Packungsprozesses, die ihreseits zu drei neuen Lösungsansätzen führt. Unter Beachtung dieser Zerlegung modellieren wir die Probleme kleinerer Dimension in einer einzigen positionsindizierten Formulierung mit Nichtüberlappungsungleichungen, die als Bindungsbedingungen dienen. Damit entwickeln wir ein neues Modell der ganzzahligen linearen Optimierung und unterziehen dies einer Polyederanalyse. Weiterhin geben wir allgemeine Nichtüberlappungs- und Dichtheitsungleichungen an und beweisen unter geeigneten Annahmen ihre facettendefinierende Eigenschaft für die konvexe Hülle der ganzzahligen Lösungen. Basierend auf dem vorgeschlagenen Modell und den starken Ungleichungen entwickeln wir einen neuen Branch-and-Cut-Algorithmus. Jedes Problem kleinerer Dimension ist eine Relaxation des höherdimensionalen Problems. Darüber hinaus besitzt es Anwendungen in verschiedenen Bereichen, wie zum Beispiel im Scheduling. Für die Behandlung der Probleme kleinerer Dimension setzen wir das Gilmore-Gomory-Modell ein, das eine Dantzig-Wolfe-Dekomposition der positionsindizierten Formulierung ist. Um eine Nachbarschaftsstruktur zu erhalten, muss die Basismatrix der optimalen Lösung die consecutive-1’s-Eigenschaft erfüllen. Für die Konstruktion solcher Matrizen entwickeln wir neue Branch-and-Price-Algorithmen, die sich durch Strategien zur Enumeration von partiellen Lösungen unterscheiden. Wir beweisen auch einige Charakteristiken von partiellen Lösungen, die das Hilfsproblem der Spaltengenerierung verschärfen. Für die nichtlineare Modellierung der höherdimensionalen Packungsprobleme untersuchen wir moderne Ansätze des Constraint Programming, modifizieren diese und schlagen neue Dichotomie- und Überschneidungsstrategien für die Verzweigung vor. Für die Verstärkung der Constraint Propagation stellen wir neue Ablehnungskriterien vor. Wir nutzen dabei 1D Relaxationen mit Intervallen und verbotenen Paaren, erweiterte Streifen-Relaxation, 2D Scheiben-Relaxation und 1D Scheiben-Streifen-Relaxation mit verbotenen Paaren. Alle vorgestellten Kriterien basieren auf Relaxationen durch Probleme kleinerer Dimension, die wir weiter durch die LP-Relaxation des Gilmore-Gomory-Modells abschwächen. Wir schließen mit Umsetzungsfragen und numerischen Experimenten aller vorgeschlagenen Ansätze.

NP-hard problems of higher-dimensional orthogonal packing are considered. We look closer at their logical structure and show that they can be decomposed into problems of a smaller dimension with a special contiguous structure. This decomposition influences the modeling of the packing process, which results in three new solution approaches. Keeping this decomposition in mind, we model the smaller-dimensional problems in a single position-indexed formulation with non-overlapping inequalities serving as binding constraints. Thus, we come up with a new integer linear programming model, which we subject to polyhedral analysis. Furthermore, we establish general non-overlapping and density inequalities and prove under appropriate assumptions their facet-defining property for the convex hull of the integer solutions. Based on the proposed model and the strong inequalities, we develop a new branch-and-cut algorithm. Being a relaxation of the higher-dimensional problem, each of the smaller-dimensional problems is also relevant for different areas, e.g. for scheduling. To tackle any of these smaller-dimensional problems, we use a Gilmore-Gomory model, which is a Dantzig-Wolfe decomposition of the position-indexed formulation. In order to obtain a contiguous structure for the optimal solution, its basis matrix must have a consecutive 1's property. For construction of such matrices, we develop new branch-and-price algorithms which are distinguished by various strategies for the enumeration of partial solutions. We also prove some characteristics of partial solutions, which tighten the slave problem of column generation. For a nonlinear modeling of the higher-dimensional packing problems, we investigate state-of-the-art constraint programming approaches, modify them, and propose new dichotomy and intersection branching strategies. To tighten the constraint propagation, we introduce new pruning rules. For that, we apply 1D relaxation with intervals and forbidden pairs, an advanced bar relaxation, 2D slice relaxation, and 1D slice-bar relaxation with forbidden pairs. The new rules are based on the relaxation by the smaller-dimensional problems which, in turn, are replaced by a linear programming relaxation of the Gilmore-Gomory model. We conclude with a discussion of implementation issues and numerical studies of all proposed approaches. / Es werden NP-schwere höherdimensionale orthogonale Packungsprobleme betrachtet. Wir untersuchen ihre logische Struktur genauer und zeigen, dass sie sich in Probleme kleinerer Dimension mit einer speziellen Nachbarschaftsstruktur zerlegen lassen. Dies beeinflusst die Modellierung des Packungsprozesses, die ihreseits zu drei neuen Lösungsansätzen führt. Unter Beachtung dieser Zerlegung modellieren wir die Probleme kleinerer Dimension in einer einzigen positionsindizierten Formulierung mit Nichtüberlappungsungleichungen, die als Bindungsbedingungen dienen. Damit entwickeln wir ein neues Modell der ganzzahligen linearen Optimierung und unterziehen dies einer Polyederanalyse. Weiterhin geben wir allgemeine Nichtüberlappungs- und Dichtheitsungleichungen an und beweisen unter geeigneten Annahmen ihre facettendefinierende Eigenschaft für die konvexe Hülle der ganzzahligen Lösungen. Basierend auf dem vorgeschlagenen Modell und den starken Ungleichungen entwickeln wir einen neuen Branch-and-Cut-Algorithmus. Jedes Problem kleinerer Dimension ist eine Relaxation des höherdimensionalen Problems. Darüber hinaus besitzt es Anwendungen in verschiedenen Bereichen, wie zum Beispiel im Scheduling. Für die Behandlung der Probleme kleinerer Dimension setzen wir das Gilmore-Gomory-Modell ein, das eine Dantzig-Wolfe-Dekomposition der positionsindizierten Formulierung ist. Um eine Nachbarschaftsstruktur zu erhalten, muss die Basismatrix der optimalen Lösung die consecutive-1’s-Eigenschaft erfüllen. Für die Konstruktion solcher Matrizen entwickeln wir neue Branch-and-Price-Algorithmen, die sich durch Strategien zur Enumeration von partiellen Lösungen unterscheiden. Wir beweisen auch einige Charakteristiken von partiellen Lösungen, die das Hilfsproblem der Spaltengenerierung verschärfen. Für die nichtlineare Modellierung der höherdimensionalen Packungsprobleme untersuchen wir moderne Ansätze des Constraint Programming, modifizieren diese und schlagen neue Dichotomie- und Überschneidungsstrategien für die Verzweigung vor. Für die Verstärkung der Constraint Propagation stellen wir neue Ablehnungskriterien vor. Wir nutzen dabei 1D Relaxationen mit Intervallen und verbotenen Paaren, erweiterte Streifen-Relaxation, 2D Scheiben-Relaxation und 1D Scheiben-Streifen-Relaxation mit verbotenen Paaren. Alle vorgestellten Kriterien basieren auf Relaxationen durch Probleme kleinerer Dimension, die wir weiter durch die LP-Relaxation des Gilmore-Gomory-Modells abschwächen. Wir schließen mit Umsetzungsfragen und numerischen Experimenten aller vorgeschlagenen Ansätze.

Le problème de localisation-routage avec capacités (PLRC) apparaît comme un problème clé dans la conception de réseaux de distribution de marchandises. Il généralisele problème de localisation avec capacités (PLC) ainsi que le problème de tournées de véhicules à multiples dépôts (PTVMD), le premier en ajoutant des décisions liées au routage et le deuxième en ajoutant des décisions liées à la localisation des dépôts. Dans cette thèse on dévelope des outils pour résoudre le PLRC à l’aide de la programmation mathématique. Dans le chapitre 3, on introduit trois nouveaux modèles pour le PLRC basés sur des flots de véhicules et des flots de commodités, et on montre comment ceux-ci dominent, en termes de la qualité de la borne inférieure, la formulation originale à deux indices [19]. Des nouvelles inégalités valides ont été dévelopées et ajoutées aux modèles, de même que des inégalités connues. De nouveaux algorithmes de séparation ont aussi été dévelopés qui dans la plupart de cas généralisent ceux trouvés dans la litterature. Les résultats numériques montrent que ces modèles de flot sont en fait utiles pour résoudre des instances de petite à moyenne taille. Dans le chapitre 4, on présente une nouvelle méthode de génération de colonnes basée sur une formulation de partition d’ensemble. Le sous-problème consiste en un problème de plus court chemin avec capacités (PCCC). En particulier, on utilise une relaxation de ce problème dans laquelle il est possible de produire des routes avec des cycles de longueur trois ou plus. Ceci est complété par des nouvelles coupes qui permettent de réduire encore davantage le saut d’intégralité en même temps que de défavoriser l’apparition de cycles dans les routes. Ces résultats suggèrent que cette méthode fournit la meilleure méthode exacte pour le PLRC. Dans le chapitre 5, on introduit une nouvelle méthode heuristique pour le PLRC. Premièrement, on démarre une méthode randomisée de type GRASP pour trouver un premier ensemble de solutions de bonne qualité. Les solutions de cet ensemble sont alors combinées de façon à les améliorer. Finalement, on démarre une méthode de type détruir et réparer basée sur la résolution d’un nouveau modèle de localisation et réaffectation qui généralise le problème de réaffectaction [48]. / The capacitated location-routing problem (CLRP) arises as a key problem in the design of distribution networks. It generalizes both the capacitated facility location problem (CFLP) and the multiple depot vehicle routing problem (MDVRP), the first by considering additional routing decisions and the second by adding the location decision variables. In this thesis we use different mathematical programming tools to develop and specialize new models and algorithms for solving the CLRP. In Chapter 3, three new models are presented for the CLRP based on vehicle-flow and commodity-flow formulations, all of which are shown to dominate, in terms of the linear relaxation lower bound, the original two-index vehicle-flow formulation [19]. Known valid inequalities are complemented with some new ones and included using separation algorithms that in many cases generalize extisting ones found in the literature. Computational experiments suggest that flow models can be efficient for dealing with small or medium size instances of the CLRP (50 customers or less). In Chapter 4, a new branch-and-cut-and-price exact algorithm is introduced for the CLRP based on a set-partitioning formulation. The pricing problem is a shortest path problem with resource constraints (SPPRC). In particular, we consider a relaxation of such problem in which routes are allowed to contain cycles of length three or more. This is complemented with the development of new valid inequalities that are shown to be effective for closing the optimality gap as well as to restrict the appearance of cycles. Computational experience supports the fact that this method is now the best exact method for the CLRP. In Chapter 5, we introduce a new metaheuristic with the aim of finding good quality solutions in short or moderate computing times. First, a bundle of good solutions is generated with the help of a greedy randomized adaptive search procedure (GRASP). Following this, a blending procedure is applied with the aim of producing a better upper bound as a combination of all the others in the bundle. An iterative destroy-and-repair method is then applied using a location-reallocation model that generalizes the reallocation model due to de Franceschi et al. [48].