[pt] O Brasil é um dos principais produtores e exportadores de celulose e papel no mundo, beneficiando-se de condições climáticas e de solo favoráveis, além de investimentos substanciais em pesquisa. Um desafio significativo nesse setor é o Problema de Planejamento de Colheita Florestal (PPCF), semelhante a um derivado do Problema de Roteamento de Veículos (VRP), com uma frota heterogênea, demanda periódica e ganho de volume de madeira. Este estudo aborda o PPCF utilizando um modelo matemático de Programação Linear Inteira Mista (MILP) e a metaheurística Greedy Randomized Adaptive Search Procedure (GRASP) em cenários simulados e reais para otimizar o sequenciamento dos times de colheita entre as unidades produtivas. O objetivo é reduzir os custos operacionais e aumentar o crescimento do volume ao longo de um horizonte de planejamento de 12 meses, considerando também as restrições de janelas de tempo. Um total de 12 instâncias foram testadas para avaliar o desempenho do GRASP, sendo que a metaheurística superou o resultado do modelo MILP em nove casos. Além disso, três instâncias refletem cenários reais de uma grande empresa brasileira de celulose e papel. Quando comparado aos resultados da equipe de planejamento da empresa, o GRASP alcançou uma redução de até 61,9 por cento nos custos totais. Além disso, o GRASP fornece planos de colheita detalhados em um curto tempo de execução, reduzindo a carga de trabalho da equipe de planejamento e aumentando a flexibilidade na tomada de decisões. / [en] Brazil is one of the world s leading producers and exporters of pulp and paper, benefiting from favorable climatic and soil conditions, coupled with substantial investments in research. A significant challenge in this sector is the Forest Harvesting Planning Problem (FHPP), akin to a derivative of the Vehicle Routing Problem (VRP) featuring a heterogeneous fleet, periodic demand, and wood volume gain. This study addresses FHPP by employing Mixed Integer Linear Programming (MILP) modeling and the Greedy Randomized Adaptive Search Procedure (GRASP) metaheuristic across real and simulated scenarios to optimize the sequencing of harvesting teams among stands. The objective is to reduce operational costs and enhance volume growth over a 12- month planning horizon, while also considering time windows and scheduling constraints. A total of 12 instances were tested to evaluate GRASP s performance, with the metaheuristic matching or outperforming the MILP model in nine cases. Additionally, three instances reflect real scenarios from a major Brazilian pulp and paper company. When compared against the company s planning team results, GRASP achieved up to a 61.9 percent reduction in total costs. Furthermore, GRASP provides detailed harvesting plans within a short execution time, reducing planning team workload and enhancing decision-making flexibility.

O ambiente de produção abordado neste trabalho é o Job Shop Flexível (JSF), uma generalização do Job Shop (JS). O problema de programação de tarefas, ou jobs, no ambiente JS é classificado por Garey; Johnson e Sethi (1976) como NP-Difícil e o JSF é, no mínimo, tão difícil quanto o JS. O JSF é composto por um conjunto de jobs, cada qual constituído por operações. Cada operação deve ser processada individualmente, sem interrupção, em uma única máquina de um subconjunto de máquinas habilitadas. O principal critério de desempenho considerado é a minimização dos atrasos dos jobs. São apresentados modelos de Programação Linear Inteira Mista (PLIM) para minimizar o atraso total e o instante de término da última operação, o makespan. São propostas novas regras de prioridade dos jobs, além de adaptações de regras da literatura. Tais regras são utilizadas por heurísticas construtivas e são aliadas a estratégias cujo objetivo é explorar características específicas do JSF. Visando aprimorar as soluções inicialmente obtidas, são propostas buscas locais e outros mecanismos de melhoria utilizados no desenvolvimento de três meta-heurísticas de diferentes categorias. Essas meta-heurísticas são: Iterated Local Search (ILS), classificada como meta-heurística de trajetória; Greedy Randomized Adaptive Search (GRASP), meta-heurística construtiva; e Artificial Bee Colony (ABC), meta-heurística populacional recentemente proposta. Esses métodos foram selecionados por alcançarem bons resultados para diversos problemas de otimização da literatura. São realizados experimentos computacionais com 600 instâncias do JSF, permitindo comparações entre os métodos de resolução. Os resultados mostram que explorar as características do problema permite que uma das regras de prioridade propostas supere a melhor regra da literatura em 81% das instâncias. As meta-heurísticas ILS, GRASP e ABC chegam a conseguir mais de 31% de melhoria sobre as soluções iniciais e a obter atrasos, em média, somente 2,24% superiores aos das soluções ótimas. Também são propostas modificações nas meta-heurísticas que permitem obter melhorias ainda mais expressivas sem aumento do tempo de execução. Adicionalmente é estudada uma versão do JSF com operações de Montagem e Desmontagem (JSFMD) e os experimentos realizados com um conjunto de 150 instâncias também indicam o bom desempenho dos métodos desenvolvidos. / The production environment addressed herein is the Flexible Job Shop (FJS), a generalization of the Job Shop (JS). In the JS environment, the jobs scheduling problem is classified by Garey; Johnson and Sethi (1976) as NP-Hard and the FJS is at least as difficult as the JS. FJS is composed of a set of jobs, each consisting of operations. Each operation must be processed individually, without interruption, in a single machine of a subset of enabled machines. The main performance criterion is minimizing the jobs tardiness. Mixed Integer Linear Programming (MILP) models are presented. These models minimize the total tardiness and the completion time of the last operation, makespan. New priority rules of jobs are proposed, as well as adaptations of rules from the literature. These rules are used by constructive heuristics and are combined with strategies aimed at exploiting specific characteristics of FSJ. In order to improve the solutions initially obtained, local searches and other improvement mechanisms are proposed and used in the development of metaheuristics of three different categories. These metaheuristics are: Iterated Local Search (ILS), classified as trajectory metaheuristic; Greedy Randomized Adaptive Search (GRASP), constructive metaheuristic, and Artificial Bee Colony (ABC), recently proposed population metaheuristic. These methods were selected owing to their good results for various optimization problems in the literature. Computational experiments using 600 FJS instances are carried out to allow comparisons between the resolution methods. The results show that exploiting the characteristics of the problem allows one of the proposed priority rules to exceed the best literature rule in about 81% of instances. Metaheuristics ILS, GRASP and ABC achieve more than 31% improvement over the initial solutions and obtain an average tardiness only 2.24% higher than the optimal solutions. Modifications in metaheuristics are proposed to obtain even more significant improvements without increased execution time. Additionally, a version called Disassembly and Assembly FSJ (DAFJS) is studied and the experiments performed with a set of 150 instances also indicate good performance of the methods developed.

Orientador: Jose Sanches Mantovani / Resumo: Neste trabalho propõem-se metodologias para realizar o planejamento de sistemas de distribuição de energia elétrica de média tensão (MT), baixa tensão (BT) e o planejamento integrado de sistemas de média e baixa tensão (MT/BT). Nos modelos de funções objetivos considerados minimizam-se os custos associados à construção, expansão, operação e confiabilidade das redes de MT, BT e MT/BT, considerando a presença de geradores distribuídos e variáveis de natureza estocástica. A geração distribuída presente no planejamento de sistemas de distribuição é avaliada através de metodologias de geração de cenários e análise de risco. Como método de busca de soluções para o problema de planejamento dos sistemas de distribuição é proposta a meta-heurística de busca em vizinhança variável GVNS (General Variable Neighborhood Search). A meta-heurística GVNS trabalha com uma série de estruturas de vizinhanças que permitem explorar o espaço de busca de forma eficiente através dos critérios de diversificação e intensificação, aumentando a probabilidade de obter soluções que não sejam ótimos locais. No planejamento integrado MT/BT, as soluções obtidas são analisadas através de um fluxo de potência integrado entre as redes de MT e BT, considerando as relações de conexão dos transformadores entre estes sistemas. Os algoritmos propostos relacionados ao planejamento de sistemas MT, BT e MT/BT são implementados em linguagem de programação C++ e testado em diferentes sistemas testes de MT (54 barras e 182... (Resumo completo, clicar acesso eletrônico abaixo) / Doutor

Submitted by DIOGO RUPOLO null (rupolo.diogo@gmail.com) on 2017-09-08T14:50:14Z No. of bitstreams: 1 Tese Final Diogo Rupolo.pdf: 4125842 bytes, checksum: 60d451cc4d13afb90cca7e443d7436b1 (MD5) / Approved for entry into archive by Monique Sasaki (sayumi_sasaki@hotmail.com) on 2017-09-11T20:51:49Z (GMT) No. of bitstreams: 1 rupolo_d_dr_ilha.pdf: 4125842 bytes, checksum: 60d451cc4d13afb90cca7e443d7436b1 (MD5) / Made available in DSpace on 2017-09-11T20:51:49Z (GMT). No. of bitstreams: 1 rupolo_d_dr_ilha.pdf: 4125842 bytes, checksum: 60d451cc4d13afb90cca7e443d7436b1 (MD5) Previous issue date: 2017-08-18 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Neste trabalho propõem-se metodologias para realizar o planejamento de sistemas de distribuição de energia elétrica de média tensão (MT), baixa tensão (BT) e o planejamento integrado de sistemas de média e baixa tensão (MT/BT). Nos modelos de funções objetivos considerados minimizam-se os custos associados à construção, expansão, operação e confiabilidade das redes de MT, BT e MT/BT, considerando a presença de geradores distribuídos e variáveis de natureza estocástica. A geração distribuída presente no planejamento de sistemas de distribuição é avaliada através de metodologias de geração de cenários e análise de risco. Como método de busca de soluções para o problema de planejamento dos sistemas de distribuição é proposta a meta-heurística de busca em vizinhança variável GVNS (General Variable Neighborhood Search). A meta-heurística GVNS trabalha com uma série de estruturas de vizinhanças que permitem explorar o espaço de busca de forma eficiente através dos critérios de diversificação e intensificação, aumentando a probabilidade de obter soluções que não sejam ótimos locais. No planejamento integrado MT/BT, as soluções obtidas são analisadas através de um fluxo de potência integrado entre as redes de MT e BT, considerando as relações de conexão dos transformadores entre estes sistemas. Os algoritmos propostos relacionados ao planejamento de sistemas MT, BT e MT/BT são implementados em linguagem de programação C++ e testado em diferentes sistemas testes de MT (54 barras e 182 barras), BT (76 barras) e MT/BT (172 e 412 barras), sob diferentes cenários operacionais. / This work proposes methodologies for the planning of medium voltage (MV), low voltage (LV), and integrated planning of medium and low voltage (MV/LV) systems. In the objective function models are considered the costs associated with the construction, expansion, operation and reliability of MV, LV and integrated MV/LV networks. Distributed generators and stochastic variables are also considered in the models. The distributed generation present in the planning of distribution systems is evaluated through methodologies of scenario generation and risk analysis. As a method of finding solutions to the problem of distribution system planning, the GVNS metaheuristic (General Variable Neighborhood Search) is proposed. The metaheuristic GVNS works with a series of neighborhood structures that allow to explore the search space efficiently through diversification and intensification criteria, increasing the probability of obtaining solutions that are not local optimum. In integrated MV/LV planning, the solutions obtained are analyzed through an integrated power flow between the MV and LV networks, considering the connection ratios of the transformers between these systems. The proposed algorithms related to the planning of MV, LV and integrated MV/LV systems are implemented in C ++ programming language and tested in different distribution systems, MV (54 and 182 bus), LV (76 bus) and MV/LV (172 and 412 bus), under different operating scenarios.

O ambiente de produção abordado neste trabalho é o Job Shop Flexível (JSF), uma generalização do Job Shop (JS). O problema de programação de tarefas, ou jobs, no ambiente JS é classificado por Garey; Johnson e Sethi (1976) como NP-Difícil e o JSF é, no mínimo, tão difícil quanto o JS. O JSF é composto por um conjunto de jobs, cada qual constituído por operações. Cada operação deve ser processada individualmente, sem interrupção, em uma única máquina de um subconjunto de máquinas habilitadas. O principal critério de desempenho considerado é a minimização dos atrasos dos jobs. São apresentados modelos de Programação Linear Inteira Mista (PLIM) para minimizar o atraso total e o instante de término da última operação, o makespan. São propostas novas regras de prioridade dos jobs, além de adaptações de regras da literatura. Tais regras são utilizadas por heurísticas construtivas e são aliadas a estratégias cujo objetivo é explorar características específicas do JSF. Visando aprimorar as soluções inicialmente obtidas, são propostas buscas locais e outros mecanismos de melhoria utilizados no desenvolvimento de três meta-heurísticas de diferentes categorias. Essas meta-heurísticas são: Iterated Local Search (ILS), classificada como meta-heurística de trajetória; Greedy Randomized Adaptive Search (GRASP), meta-heurística construtiva; e Artificial Bee Colony (ABC), meta-heurística populacional recentemente proposta. Esses métodos foram selecionados por alcançarem bons resultados para diversos problemas de otimização da literatura. São realizados experimentos computacionais com 600 instâncias do JSF, permitindo comparações entre os métodos de resolução. Os resultados mostram que explorar as características do problema permite que uma das regras de prioridade propostas supere a melhor regra da literatura em 81% das instâncias. As meta-heurísticas ILS, GRASP e ABC chegam a conseguir mais de 31% de melhoria sobre as soluções iniciais e a obter atrasos, em média, somente 2,24% superiores aos das soluções ótimas. Também são propostas modificações nas meta-heurísticas que permitem obter melhorias ainda mais expressivas sem aumento do tempo de execução. Adicionalmente é estudada uma versão do JSF com operações de Montagem e Desmontagem (JSFMD) e os experimentos realizados com um conjunto de 150 instâncias também indicam o bom desempenho dos métodos desenvolvidos. / The production environment addressed herein is the Flexible Job Shop (FJS), a generalization of the Job Shop (JS). In the JS environment, the jobs scheduling problem is classified by Garey; Johnson and Sethi (1976) as NP-Hard and the FJS is at least as difficult as the JS. FJS is composed of a set of jobs, each consisting of operations. Each operation must be processed individually, without interruption, in a single machine of a subset of enabled machines. The main performance criterion is minimizing the jobs tardiness. Mixed Integer Linear Programming (MILP) models are presented. These models minimize the total tardiness and the completion time of the last operation, makespan. New priority rules of jobs are proposed, as well as adaptations of rules from the literature. These rules are used by constructive heuristics and are combined with strategies aimed at exploiting specific characteristics of FSJ. In order to improve the solutions initially obtained, local searches and other improvement mechanisms are proposed and used in the development of metaheuristics of three different categories. These metaheuristics are: Iterated Local Search (ILS), classified as trajectory metaheuristic; Greedy Randomized Adaptive Search (GRASP), constructive metaheuristic, and Artificial Bee Colony (ABC), recently proposed population metaheuristic. These methods were selected owing to their good results for various optimization problems in the literature. Computational experiments using 600 FJS instances are carried out to allow comparisons between the resolution methods. The results show that exploiting the characteristics of the problem allows one of the proposed priority rules to exceed the best literature rule in about 81% of instances. Metaheuristics ILS, GRASP and ABC achieve more than 31% improvement over the initial solutions and obtain an average tardiness only 2.24% higher than the optimal solutions. Modifications in metaheuristics are proposed to obtain even more significant improvements without increased execution time. Additionally, a version called Disassembly and Assembly FSJ (DAFJS) is studied and the experiments performed with a set of 150 instances also indicate good performance of the methods developed.

Étant donné leur faible impact environnemental, l’utilisation des véhicules électriques dans les activités de service a beaucoup augmenté depuis quelques années. Cependant, leur déploiement est freiné par des contraintes techniques telles qu’une autonomie limitée et de longs temps de charge des batteries. La prise en compte de ces contraintes a mené à l’apparition de nouveaux problèmes de tournées de véhicules pour lesquels, en plus d’organiser les tournées,il faut décider où et de combien charger les batteries. Dans cette thèse nous nous intéressons à ces problèmes au travers de quatre études. La première concerne le développement d’une métaheuristique en deux phases simple mais performante pour résoudre un problème particulier appelé "Green VRP”. Dans la seconde, nous nous concentrons sur la modélisation d’un aspect essentiel dans ces problèmes : le processus de chargement des batteries. Nous étudions différentes stratégies pour modéliser ce processus et montrons l’importance de considérer la nature non linéaire des fonctions de chargement. Dans la troisième étude nous proposons une recherche locale itérative pour résoudre des problèmes avec des fonctions de chargement non linéaires. Nous introduisons un voisinage dédié aux décisions de chargement basé sur un nouveau problème de chargement sur une tournée fixée. Dans la dernière étude, nous traitons un problème réel de tournées de techniciens avec des véhicules classiques et électriques. Ce problème est résolu par une métaheuristique qui décompose le problème en plusieurs sous-problèmes plus simples résolus en parallèle, puis qui assemble des parties des solutions trouvées pour construire la solution finale. / Electric vehicles (evs) are one of the most promising technologies to reduce the greenhouse gas emissions. For this reason, the use of evs in service operations has dramatically increased in recent years. Despite their environmental benefits, evs still face technical constraints such as short autonomy and long charging times. Taking into account these constraints when planning ev operations leads to a new breed of vehicle routing problems (vrps), known as electricVrps (evrps). In addition, to the standard routing decisions, evrps are concerned with charging decisions: where and how much to charge. In this ph. D thesis, we address evrps through 4 different studies. In the first study, we tackle the green vehicle routing problem. To solve the problem, we propose a simple, yet effective, two-phase matheuristic. In the second study, we focus a key modelling aspects in evrps: the battery charging process. We study different strategies to model this process and show the importance of considering the nonlinear nature of the battery charging functions. InThe third study, we propose an iterated local search to tackle evrp with non-linear charging functions. We introduce a particular local search operator for the charging decisions based on a new fixedroute charging problem. The fourth and last study considers a real technician routing problem with conventional and electric vehicles (trp-cev). To tackle this problem, we propose a parallel matheuristic that decomposes the problem into a set of easier-to-solve subproblemsThat are solved in parallel processors. Then the approach uses parts of the solutions found to the subproblems to assemble final solution to the trp-cev.

Cette thèse comporte des résultats théoriques et pratiques sur deux métaheuristiques, la Recherche Dispersée et la Recherche Voisinage variable (RVV), ainsi que sur des Matheuristiques. Au niveau théorique, la contribution principale de cette thèse est la proposition d’un algorithme de recherche dispersée avec l’arrondi directionnel convergent pour les programmes en nombres entiers mixtes (0-1 MIP), avec une preuve de cette convergence en un nombre fini d’itérations. En se basant sur cet algorithme convergeant, deux implémentations et plusieurs heuristiques sont proposées et testées sur des instances de 0-1 MIP. Les versions testées reposent sur des implémentations non optimisées pour mettre en évidence la puissance des approches dans une forme simplifiée. Nos résultats démontrent l’efficacité de ces approches initiales, ce qui les rend attractives lorsque des solutions de très haute qualité sont recherchées avec un investissement approprié en termes d’effort de calcul. Cette thèse inclut également quelques nouvelles variantes de la métaheuristique Recherche Voisinage Variable telles qu’une recherche voisinage variable deux niveaux, une recherche voisinage variable imbriquée, une descente voisinage variable cyclique et une heuristique de plongée voisinage variable. En outre, plusieurs implémentations efficaces de ces algorithmes basés sur la recherche voisinage variable ont été appliquées avec succès à des problèmes NP-Difficiles apparaissant en transport, logistique, production d’énergie, ordonnancement, et segmentation. Les heuristiques proposées se sont avérées être les nouvelles heuristiques de référence sur tous les problèmes considérés. La dernière contribution de cette thèse repose sur la proposition de plusieurs matheuristiques pour résoudre le problème de Conception de Réseau Multi-flots avec Coût fixe (CRMC). Les performances de ces matheuristiques ont été évaluées sur un ensemble d’instances de référence du CRMC. Les résultats obtenus démontrent la compétitivité des approches proposées par rapport aux approches existantes de la littérature. / This thesis consists of results obtained studying Scatter Search, Variable Neighbourhood Search (VNS), and Matheuristics in both theoretical and practical context. Regarding theoretical results, one of the main contribution of this thesis is a convergent scatter search with directional rounding algorithm for 0-1 Mixed Integer Programs (MIP) with the proof of its finite convergence. Besides this, a convergent scatter search algorithm is accompanied by two variants of its implementation. Additionally, several scatter search based heuristics, stemming from a convergent scatter search algorithm have been proposed and tested on some instances of 0-1 MIP. The versions of the methods tested are first stage implementations to establish the power of the methods in a simplified form. Our findings demonstrate the efficacy of these first stage methods, which makes them attractive for use in situations where very high quality solutions are sought with an efficient investment of computational effort.This thesis also includes new variants of Variable Neighborhood Search metaheuristic such as a two-level variable neighborhood search, a nested variable neighborhood search, a cyclic variable neighborhood descent and a variable neighborhood diving. Additionally, several efficient implementation of those variable neighborhood search algorithms have been successfully applied for solving NP-Hard problems appearing in transportation, logistics, power generation, scheduling and clustering. On all tested problems, the proposed VNS heuristics turned out to be a new state-of-the art heuristics. The last contribution of this thesis consists of proposing several matheuristics for solving Fixed-Charge Multicommodity Network Design (MCND) problem. The performances of these matheuristics have been disclosed on benchmark instances for MCND. The obtained results demonstrate the competitiveness of the proposed matheuristics with other existing approaches in the literature.

Cette thèse s’intéresse au problème classique de tournées de véhicules avec contraintes de capacité (CVRP pour Capacitated Vehicle Routing Problem) ainsi qu’une variante beaucoup plus complexe, soit le problème de tournées de véhicules dépendant du temps avec fenêtres de temps et points de transfert défini sur un réseau routier (TDVRPTWTP-RN pour Time-Dependent Vehicle Routing Problem with Time Windows and Transfer Points on a Road Network). Dans le premier article, le TDVRPTWTP-RN est résolu en adaptant une métaheuristique qui représente l’état de l’art pour le CVRP, appelé Slack Induction for String Removals (SISR). Cette métaheuristique fait appel au principe “détruire et reconstruire” en retirant des séquences de clients consécutifs dans les routes de la solution courante et en réinsérant ensuite ces clients de façon à créer une nouvelle solution. Le problème est défini sur un réseau routier où différents chemins alternatifs peuvent être utilisés pour se déplacer d’un client à l’autre. De plus, le temps de parcours sur chacun des arcs du réseau n’est pas fixe, mais dépend du moment où le véhicule quitte le sommet origine. S’inspirant de problèmes rencontrés en logistique urbaine, nous considérons également deux types de véhicules, de petite et grande capacité, où les grands véhicules sont interdits de passage au centre-ville. Ainsi, les clients du centre-ville ne peuvent être servis que suite au transfert de leur demande d’un grand à un petit véhicule à un point de transfert. Comme un point de transfert n’a pas de capacité, une problématique de synchronisation apparaît quand un grand véhicule doit y rencontrer un ou plusieurs petits véhicules pour leur transférer une partie de son contenu. Contrairement aux problèmes stricts de tournées de véhicules à deux échelons, les grands véhicules peuvent aussi servir des clients localisés à l’extérieur du centre-ville. Comme le problème abordé est beaucoup plus complexe que le CVRP, des modifications importantes ont dû être apportées à la métaheuristique SISR originale. Pour évaluer la performance de notre algorithme, un ensemble d’instances tests a été généré à partir d’instances existantes pour le TDVRPTW-RN. Les réseaux omt été divisés en trois régions : centre-ville, frontière et extérieur. Le centre-ville et l’extérieur sont respectivemnt les royaumes des petits et grands véhicules, tandis que la frontière (où l’on retrouve les points de transfert) peut être visité par les deux types de véhicules. Les résultats numériques montrent que la métaheuristique proposée exploite les opportunités d’optimiser une solution en déplaçant autant que possible les clients neutres, soit ceux qui peuvent être servis indifféremment par un petit ou un grand véhicule, des routes des petits véhicules vers les routes des grands véhicules, réduisant ainsi les coûteuses visites aux points de transfert. Les deuxième et troisième article s’intéressent à des concepts plus fondamentaux et font appel au problème plus simple du CVRP pour les évaluer. Dans le second article, un étude expérimentale est conçue afin d’examiner l’impact de données (distances) imprécises sur la performance de différents types d’heuristiques, ainsi qu’une méthode exacte, pour le CVRP. À cette fin, différents niveaux d’imprécision ont été introduits dans des instances tests classiques pour le CVRP avec 100 à 1 000 clients. Nous avons observé que les meilleures métaheuristiques demeurent les meilleures, même en présence de hauts niveaux d’imprécision, et qu’elles ne sont pas affectées autant par les imprécisions qu’une heuristique simple. Des expériences avec des instances réelles ont mené aux mêmes conclusions. Le troisième article s’intéresse à l’intégration de l’apprentissage automatique dans la métaheuristique SISR qui représente l’état de l’art pour le CVRP. Dans ce travail, le principe “détruire et reconstruire” au coeur de SISR est hybridé avec une méthode d’apprentissage par renforcement qui s’inspire des systèmes de colonies de fourmis. L’ap- prentissage automatique a pour but d’identifier les arêtes les plus intéressantes, soit celles qui se retrouvent le plus fréquemment dans les solutions de grande qualité précédemment rencontrées au cours de la recherche. L’inclusion de telles arêtes est alors favorisé lors de la réinsertion des clients ayant été retirés de la solution par le mécanisme de destruction. Les instances utilisées pour tester notre approche hybride sont les mêmes que celles du second article. Nous avons observé que notre algorithme ne peut produire que des solutions lé- gèrement meilleures que la métaheuristique SISR originale, celle-ci étant déjà quasi-optimale. / This thesis is concerned both with the classical Capacitated Vehicle Routing Problem (CVRP) and a much more complex variant called the Time-Dependent Vehicle Routing Problem with Time Windows and Transfer Points on a Road Network (TDVRPTWTP-RN ). In the first paper, the TDVRPTWTP RN is solved by adapting a state-of-the-art metaheuris- tic for the CVRP, called Slack Induction for String Removals (SISR). This metaheuristic is based on the ruin and recreate principle and removes strings of consecutive customers in the routes of the current solution and then reinserts the removed customers to create a new solution. The problem is formulated in a full road network where different alternative paths can be used to go from one customer to the next. Also, the travel time on each arc of the road network is not fixed, but depends on the departure time from the origin node. Motivated from city logistics applications, we also consider two types of vehicles, large and small, with large vehicles being forbidden from the downtown area. Thus, downtown customers can only be served through a transfer of their goods from large to small vehicles at designated transfer points. Since transfer points have no capacity, synchronization issues arise when a large vehicle must meet one or more small vehicles to transfer goods. As opposed to strict two-echelon VRPs, large vehicles can also directly serve customers that are outside of the downtown area. Given that the TDVRPTWTP-RN is much more complex than the CVRP, important modifications to the original SISR metaheuristic were required. To evaluate the performance of our algorithm, we generated a set of test instances by extending existing instances of the TDVRPTW-RN . The road networks are divided into three regions: downtown, boundary and outside. The downtown and outside areas are the realm of small and large vehicles, respectively, while the boundary area that contains the transfer points can be visited by both small and large vehicles. The results show that the proposed metaheuristic exploits optimization opportunities by moving as much as possible neutral customers (which can be served by either small or large vehicles) from the routes of small vehicles to those of large vehicles, thus avoiding costly visits to transfer points. The second and third papers examine more fundamental issues, using the classical CVRP as a testbed. In the second paper, an experimental study is designed to examine the impact of inaccurate data (distances) on the performance of different types of heuristics, as well as one exact method, for the CVRP. For this purpose, different levels of distance inaccuracies were introduced into well-known benchmark instances for the CVRP with 100 to 1,000 customers. We observed that the best state-of-the-art metaheuristics remain the best, even in the presence of high inaccuracy levels, and that they are not as much affected by inaccuracies when compared to a simple heuristic. Some experiments performed on real-world instances led to the same conclusions. The third paper focuses on the integration of learning into the state-of-the-art SISR for the CVRP. In this work, the ruin and recreate mechanism at the core of SISR is enhanced by a reinforcement learning technique inspired from ant colony systems. The learning component is aimed at identifying promising edges, namely those that are often found in previously encountered high-quality solutions. The inclusion of these promising edges is then favored during the reinsertion of removed customers. The benchmark instances of the second paper were also used here to test the new hybrid algorithm. We observed that the latter can produce only slightly better solutions than the original SISR, due to the quasi-optimality of the original solutions.