Généralement, les problèmes de conception de réseaux consistent à sélectionner les arcs et les sommets d’un graphe G de sorte que la fonction coût est optimisée et l’ensemble de contraintes impliquant les liens et les sommets dans G sont respectées. Une modification dans le critère d’optimisation et/ou dans l’ensemble de contraintes mène à une nouvelle représentation d’un problème différent. Dans cette thèse, nous nous intéressons au problème de conception d’infrastructure de réseaux maillés sans fil (WMN- Wireless Mesh Network en Anglais) où nous montrons que la conception de tels réseaux se transforme d’un problème d’optimisation standard (la fonction coût est optimisée) à un problème d’optimisation à plusieurs objectifs, pour tenir en compte de nombreux aspects, souvent contradictoires, mais néanmoins incontournables dans la réalité. Cette thèse, composée de trois volets, propose de nouveaux modèles et algorithmes pour la conception de WMNs où rien n’est connu à l’ avance. Le premiervolet est consacré à l’optimisation simultanée de deux objectifs équitablement importants : le coût et la performance du réseau en termes de débit. Trois modèles bi-objectifs qui se différent principalement par l’approche utilisée pour maximiser la performance du réseau sont proposés, résolus et comparés. Le deuxième volet traite le problème de placement de passerelles vu son impact sur la performance et l’extensibilité du réseau. La notion de contraintes de sauts (hop constraints) est introduite dans la conception du réseau pour limiter le délai de transmission. Un nouvel algorithme basé sur une approche de groupage est proposé afin de trouver les positions stratégiques des passerelles qui favorisent l’extensibilité du réseau et augmentent sa performance sans augmenter considérablement le coût total de son installation. Le dernier volet adresse le problème de fiabilité du réseau dans la présence de pannes simples. Prévoir l’installation des composants redondants lors de la phase de conception peut garantir des communications fiables, mais au détriment du coût et de la performance du réseau. Un nouvel algorithme, basé sur l’approche théorique de décomposition en oreilles afin d’installer le minimum nombre de routeurs additionnels pour tolérer les pannes simples, est développé. Afin de résoudre les modèles proposés pour des réseaux de taille réelle, un algorithme évolutionnaire (méta-heuristique), inspiré de la nature, est développé. Finalement, les méthodes et modèles proposés on été évalués par des simulations empiriques et d’événements discrets. / Generally, network design problems consist of selecting links and vertices of a graph G so that a cost function is optimized and all constraints involving links and the vertices in G are met. A change in the criterion of optimization and/or the set of constraints leads to a new representation of a different problem. In this thesis, we consider the problem of designing infrastructure Wireless Mesh Networks (WMNs) where we show that the design of such networks becomes an optimization problem with multiple objectives instead of a standard optimization problem (a cost function is optimized) to take into account many aspects, often contradictory, but nevertheless essential in the reality. This thesis, composed of three parts, introduces new models and algorithms for designing WMNs from scratch. The first part is devoted to the simultaneous optimization of two equally important objectives: cost and network performance in terms of throughput. Three bi-objective models which differ mainly by the approach used to maximize network performance are proposed, solved and compared. The second part deals with the problem of gateways placement, given its impact on network performance and scalability. The concept of hop constraints is introduced into the network design to reduce the transmission delay. A novel algorithm based on a clustering approach is also proposed to find the strategic positions of gateways that support network scalability and increase its performance without significantly increasing the cost of installation. The final section addresses the problem of reliability in the presence of single failures. Allowing the installation of redundant components in the design phase can ensure reliable communications, but at the expense of cost and network performance. A new algorithm is developed based on the theoretical approach of "ear decomposition" to install the minimum number of additional routers to tolerate single failures. In order to solve the proposed models for real-size networks, an evolutionary algorithm (meta-heuristics), inspired from nature, is developed. Finally, the proposed models and methods have been evaluated through empirical and discrete events based simulations.

Dans cette thèse, nous considérons le problème du séquencement sur une ligne de montage multi-modèles de véhicules industriels. Pour équilibrer au mieux la charge dynamique des opérateurs, la minimisation de la somme des retards à l’issue de chaque véhicule est proposée.Deux approches peuvent être utilisées pour optimiser le lissage de charge dans un problème de séquencement : l’utilisation directe des temps opératoires ou le respect de règles. La plupart des travaux appliqués à l’industrie automobile utilisent l’approche de respect de règles. Une originalité de ce travail est d’utiliser l’approche de la prise en compte directe des temps opératoires.L’étude de la littérature de ce problème a dévoilé deux lacunes dans les travaux précédents : l’essentiel des travaux modélisent un seul type d’opérateurs d’une part, et proposent des heuristiques ou des métaheuristiques pour résoudre ces problèmes, d’autre part. L’originalité de ce travail est de tester des méthodes exactes pour des instances industrielles et de modéliser le fonctionnement de trois différents types d’opérateurs spécifiques au cas industriel.Deux méthodes exactes sont développées : la programmation linéaire mixte et la programmation dynamique. Une étude expérimentale des facteurs de complexité sur des instances académiques des deux modèles est développée. Les modèles sont aussi testés sur des instances du cas d’étude.Par ailleurs, le problème est traité par deux méthodes approchées : une heuristique basée sur la programmation dynamique d’une part, et des métaheuristiques (algorithme génétique, recuit simulé et un couplage des deux) d’autre part. Les deux approches sont testées sur des instances académiques et des instances du cas d’étude.Ce travail a permis d’apporter une solution intéressante d’un point de vue industriel puisqu’il prend en compte les caractéristiques de la ligne de montage (opérateurs spécifiques) et améliore significativement la qualité du séquencement en un temps de calcul raisonnable. / In this thesis, the problem of sequencing mixed model assembly lines (MMAL) is considered. Our goal is to determine the sequence of products to minimize the work overload. This problem is known as the mixed model assembly line sequencing problem with work overload minimization (MMSP-W). This work is based on an industrial case study of a truck assembly line.Two approaches can be used to minimize the work overload: the use of task operation times or the respect of sequencing rules. Most of the earlier works applied in car industry use the latter approach. The originality of this work is to employ the task operation times for the generation of the product sequence in a MMAL.The literature review has highlighted two main gaps in previous works: most of the papers consider a single type of operators, and propose heuristics or metaheuristics to solve the problem. The originality of this work is to test exact methods for industrial case instances and to model three different types of operators.Two exact methods are developed: the mixed integer linear programming and dynamic programming. The models are tested on industrial case study instances. An experimental study is developed for both approaches in order to understand the complexity factors.Moreover, the problem is treated by two approximate methods: a heuristic based on dynamic programming and metaheuristics (genetic algorithm, simulated annealing and a hybrid method based on both genetic algorithm and simulated annealing). All approaches are tested on academic instances and on real data from the industrial case study.

Submitted by FABRÍCIO MACIEL GOMES null (fabricio@dequi.eel.usp.br) on 2016-01-04T00:06:19Z No. of bitstreams: 1 Tese_Fabricio_Maciel_Gomes.pdf: 1836829 bytes, checksum: 3ec7860a9d87ebfeaef21b25dc157d25 (MD5) / Approved for entry into archive by Juliano Benedito Ferreira (julianoferreira@reitoria.unesp.br) on 2016-01-06T16:12:19Z (GMT) No. of bitstreams: 1 gomes_fm_dr_guara.pdf: 1836829 bytes, checksum: 3ec7860a9d87ebfeaef21b25dc157d25 (MD5) / Made available in DSpace on 2016-01-06T16:12:19Z (GMT). No. of bitstreams: 1 gomes_fm_dr_guara.pdf: 1836829 bytes, checksum: 3ec7860a9d87ebfeaef21b25dc157d25 (MD5) Previous issue date: 2015-12-15 / Empresas não medem esforços para aperfeiçoar seus processos e produtos de acordo com diferentes critérios para satisfazer as exigências e necessidades dos clientes em busca de um padrão de competitividade superior ao de suas concorrentes. Neste cenário é muito comum a necessidade de se estabelecer condições que resultem na melhoria de mais de um critério de forma simultânea. Neste trabalho foi realizada uma avaliação da utilização de quatro métodos que utilizam as Meta-heurísticas Recozimento Simulado, Algoritmo Genético, Recozimento Simulado combinado com o método Nelder Mead Simplex e algoritmo genético combinado com o método Nelde-Mead simplex para o estabelecimento de melhoria das condições de processos com múltiplas respostas. Para a avaliação dos métodos propostos foram utilizados problemas-teste criteriosamente selecionados na literatura de forma a serem analisados casos com diferente número de variáveis, número de respostas e tipos de resposta. A aglutinação das respostas foi realizada por quatro métodos diferentes: Desirability, Desvio Médio Percentual, Programação por Compromisso e Programação por Compromisso normalizada pela distância euclidiana. A avaliação dos métodos foi realizada por meio de comparação entre os resultados obtidos na utilização de um mesmo método de aglutinação, determinando assim a eficiência do método de busca. Os resultados obtidos na avaliação dos métodos sugerem a aplicação do método do algoritmo genético quando se pretende estabelecer parâmetros que resultem na melhoria de processos com múltiplas respostas, em particular quando essas respostas são modeladas por equações com termos cúbicos, independentemente do número de termos que possam conter, do tipo de respostas e do número de variáveis. / Companies go to great lengths to improve its processes and products according to different criteria to meet the demands and needs of customers looking for a higher standard of competitiveness to that of their competitors. This scenario is very common the need to establish conditions that result in the improvement of more than one criterion simultaneously. This work was carried out an evaluation of the use of four methods that use Metaheuristics Simulated Annealing, Genetic Algorithms, Simulated Annealing combined with the Nelder Mead Simplex method and genetic algorithm combined with Nelde Mead simplex method for the improvement of establishing the conditions of processes with multiple answers. For the evaluation of the proposed test methods were used in the literature problems carefully selected in order to be analyzed cases with different numbers of variables, response numbers and types of responses. In this research we used the average percentage deviation function as a way to bring together the answers. The agglutination of the answers was performed by four different methods: Desirability, Average Percentage Deviation, Compromise Programming and Compromise Programming normalized by Euclidean distance. The evaluation method was performed by comparison between the results obtained in using the same bonding method, thereby determining the efficiency of the search method. The results obtained in the evaluation of the methods suggest the application of the genetic algorithm method when you want to set parameters that result in the improvement of processes with multiple answers, particularly when these responses are modeled by equations with cubic terms, regardless of the number of terms that can contain the type of responses and the number of variables.

Made available in DSpace on 2014-06-11T19:22:35Z (GMT). No. of bitstreams: 0 Previous issue date: 2008-06-06Bitstream added on 2014-06-13T20:09:51Z : No. of bitstreams: 1 sanchez_frl_me_ilha.pdf: 660422 bytes, checksum: f8ab299d7cef18ca3a218acf27a94f43 (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Neste trabalho foram implementados diversos agentes heuristicos construtivos, baseados no modelo híbrido linear, que fazem parte de um time assíncrono que tem como objetivo gerar configurações de boa qualidade para inicializar as metaheuríticas que resolvem o problema do planejamento da expansão da transmissão dos sistemas de energia elétrica. A teoria de times assíncronos foi aplicada para reunir as qualidades individuais dos métodos heurísticos, de uma maneira que, partindo de uma configuração base (sem adições) e utilizando um fluxo de dados cíclico, os agentes construtivos adicionassem circuitos a esta configuração de maneira sistemática e aleatória até que esta atenda as demandas de carga solicitadas pelo sistema elétrico em um horizonte futuro. Estas configurações foram então utilizadas por um algoritmo genético no intuito de validar a qualidade das mesmas. Os algoritmos foram implementados em Fortran, utilizando as rotinas de trocas de mensagens do LAM-MPI e simulados para sistemas teste de pequeno, médio e grande porte em ambiente de processamento distribuido. Os resultados comprovam que os times ass´ıncronos de vários metodos heurísticos são mais eficazes comparados com uma única heurística. / In this study, it was implemented several constructive heuristic algorithms, based on hybrid linear model, which are part of a asynchronous team that aims to generate initial solutions with good quality for meta-heuristics that solve the transmission expansion planning problem of electric power systems. The theory of asynchronous team was applied to meet the individual qualities of each heuristic method, in a way that, starting from a base network configuration and using a cyclical flow of data, heuristic agents add circuits to is configuration in a systematic and random way until they meet the load demands requested by the electrical system on a future horizon. Then these configurations are utilized by a genetic algorithm in order to validate the quality of them. The algorithms were implemented in Fortran, using exchanging messages routines from LAM-MPI and simulated for small, medium and large size test-systems in distributed processing environment. The results show that the solutions obtained with asynchronous teams of several heuristic methods are more effective than the solutions with a single heuristic algorithm.

Esta pesquisa trata da aplicação de meta-heurísticas baseadas em busca em vizinhança variável em problemas de operação de transportes. Desta forma, buscou-se encontrar problemas complexos durante a operação de sistemas de transportes, nas grandes cidades, que possam ser resolvidos com a aplicação de meta-heurística baseada em busca em vizinhança variável. Este trabalho aborda dois diferentes problemas de planejamento e operação de transportes. O primeiro problema abordado neste trabalho é o Problema de Programação da Tabela de Horários, de Veículos e de Tripulantes de Ônibus, no qual as viagens que comporão a tabela de horários, os veículos que executarão as viagens e as tripulações que operarão os veículos são alocadas simultaneamente e de maneira integrada. O segundo problema a ser abordado é o problema de distribuição física, o qual envolve o agrupamento e a alocação de entregas a uma frota de veículos visando minimizar o frete total. Uma abordagem para a modelagem matemática deste problema é modelar como um problema de bin-packing, com bins de tamanho variável unidimensional (do inglês Variable Sized Bin-Packing Problem - VSBPP), ou seja, uma generalização do tradicional problema de bin-packing no qual bins (veículos) de diferentes capacidades e custos estão disponíveis para a alocação de um conjunto de objetos (cargas), de modo que o custo total dos bins (veículos) utilizados seja mínimo. A outra abordagem proposta para o problema de distribuição física é modelar o problema como um problema de bin-packing, com bins de tamanho variável bidimensional (do inglês Bidimensional Variable Sized Bin-Packing Problem BiD-VSBPP). Assim sendo, trata-se de uma expansão do problema de bin-packing com bins de tamanho variável unidimensional (VSBPP), no qual bins (veículos) de diferentes capacidades (capacidade volumétrica e capacidade de carga) e custos estão disponíveis para a alocação de um conjunto de objetos (cargas), os quais possuem as dimensões peso e volume, de modo que o custo total dos bins (veículos) utilizados seja mínimo. Durante a realização deste trabalho, foi desenvolvido um programa computacional em C++, o qual implementa a meta-heurística Busca em Vizinhança Variável (VNS) e duas meta-heurísticas baseadas em VNS. São apresentados resultados de experimentos computacionais com dados reais e dados benchmarking. Os resultados obtidos comprovam a eficácia das meta-heurísticas propostas. / This work approaches variable neighborhood search meta-heuristic applicate on transport operation problems. This way, we sought find complex transport operation problems in large cities that can be solved with the variable neighborhood search meta-heuristic application. This work approaches two different transport planning and operation problems. The first problem approached in this paper is the Bus Timetable Vehicle Crew Scheduling Problem, in which timetabling, bus and crew schedules are simultaneously determined in an integrated approach. The second problem to be approached is the physical distribution problem which comprises grouping and assigning deliveries to a heterogeneous fleet of vehicles aiming to minimize the total freight cost. The problem can be mathematical modeled as one-dimensional Variable Sized Bin-Packing Problem (VSBPP), a generalization of the traditional bin-packing problem, in which bins (vehicles) with different sizes and costs are available for the assignment of the objects (deliveries) such that the total cost of the used bins (vehicles) is minimized. Another proposed approach to the problem of physical distribution is model as two dimensional Variable Sized Bin-Packing Problem (BiD-VSBPP). Therefore, it is an expansion of the bin-packing problem with bins variable-length-dimensional (VSBPP), in which bins (vehicle) of different capacity (capacity and load carrying capacity) and costs are available for allocation a set of objects (loads), which have the dimensions weight and volume, so that minimized the total cost of bins (vehicle). In this work, was developed a C++ software implemented, which was implemented a meta-heuristic Variable Neighborhood Search (VNS) and two others meta-heuristics based on VNS. Computational results for real-world problems and benchmarking problems are presented, showing the effectiveness of these proposed meta-heuristics.

Le Cloud est devenu une plate-forme très répandue pour le déploiement d'applications distribuées. Beaucoup d'entreprises peuvent sous-traiter leurs infrastructures d'hébergement et, ainsi, éviter des dépenses provenant d'investissements initiaux en infrastructure et de maintenance.Des petites et moyennes entreprises, en particulier, attirés par le modèle de coûts sur demande du Cloud, ont désormais accès à des fonctionnalités comme le passage à l'échelle, la disponibilité et la fiabilité, qui avant le Cloud étaient presque réservées à de grandes entreprises.Les services du Cloud peuvent être offerts aux utilisateurs de plusieurs façons. Dans cette thèse, nous nous concentrons sur le modèle d'Infrastructure sous Forme de Service. Ce modèle permet aux utilisateurs d’accéder à des ressources de calcul virtualisés sous forme de machine virtuelles (MVs).Pour installer une application distribuée, un client du Cloud doit d'abord définir l'association entre son application et l'infrastructure. Il est nécessaire de prendre en considération des contraintesde coût, de ressource et de communication pour pouvoir choisir un ensemble de MVs provenant d'opérateurs de Cloud publiques et privés le plus adaptés. Cependant, étant donné la quantité exponentiel de configurations, la définition manuelle de l'association entre application et infrastructure peut être un challenge dans des scénarios à large échelle ou ayant des contraintes importantes de temps. En effet, ce problème est une généralisation du problème de calcul de homomorphisme de graphes, qui est NP-complet.Dans cette thèse, nous adressons le problème de calculer des placements initiaux et de reconfiguration pour des applications distribuées sur potentiellement de multiples Clouds. L'objectif est de minimiser les coûts de location et de migration en satisfaisant des contraintes de ressources et communications. Pour cela, nous proposons des heuristiques performantes capables de calculer des placements de bonne qualité très rapidement pour des scénarios à petite et large échelles. Ces heuristiques, qui sont basées sur des algorithmes de partition de graphes et de vector packing, ont été évaluées en les comparant avec des approches de l'état de l'art comme des solveurs exactes et des méta-heuristiques. Nous montrons en utilisant des simulations que les heuristiques proposées arrivent à calculer des solutions de bonne qualité en quelques secondes tandis que des autres approches prennent des heures ou jours pour les calculer. / The Cloud has become a very popular platform for deploying distributed applications. Today, virtually any credit card holder can have access to Cloud services. There are many different ways of offering Cloud services to customers. In this thesis we especially focus on theInfrastructure as a Service (IaaS), a model that, usually, proposes virtualized computing resources to costumers in the form of virtual machines (VMs). Thanks to its attractive pay-as-you-use cost model, it is easier for customers, specially small and medium companies, to outsource hosting infrastructures and benefit of savings related to upfront investments and maintenance costs. Also, customers can have access to features such as scalability, availability, and reliability, which previously were almost exclusive for large companies. To deploy a distributed application, a Cloud customer must first consider the mapping between her application (or its parts) to the target infrastructure. She needs to take into consideration cost, resource, and communication constraints to select the most suitable set of VMs, from private and public Cloud providers. However, defining a mapping manually may be a challenge in large-scale or time constrained scenarios since the number of possible configuration explodes. Furthermore, when automating this process, scalability issues must be taken into account given that this mapping problem is a generalization of the graph homomorphism problem, which is NP-complete.In this thesis we address the problem of calculating initial and reconfiguration placements for distributed applications over possibly multiple Clouds. Our objective is to minimize renting and migration costs while satisfying applications' resource and communication constraints. We concentrate on the mapping between applications and Cloud infrastructure. Using an incremental approach, we split the problem into three different parts and propose efficient heuristics that can compute good quality placements very quickly for small and large scenarios. These heuristics are based on graph partition and vector packing heuristics and have been extensively evaluated against state of the art approaches such as MIP solvers and meta-heuristics. We show through simulations that the proposed heuristics manage to compute solutions in a few seconds that would take many hours or days for other approaches to compute.

Les techniques émergentes de l’ingénierie dirigée par les modèles et de la programmation générative ont permis la création de plusieurs générateurs (générateurs de code et compilateurs). Ceux-ci sont souvent utilisés afin de faciliter le développement logiciel et automatiser le processus de génération de code à partir des spécifications abstraites. De plus, les générateurs modernes comme les compilateurs C, sont devenus hautement configurables, offrant de nombreuses options de configuration à l'utilisateur de manière à personnaliser facilement le code généré pour la plateforme matérielle cible. Par conséquent, la qualité logicielle est devenue fortement corrélée aux paramètres de configuration ainsi qu'au générateur lui-même. Dans ce contexte, il est devenu indispensable de vérifier le bon comportement des générateurs. Cette thèse établit trois contributions principales : Contribution I: détection automatique des inconsistances dans les familles de générateurs de code : Dans cette contribution, nous abordons le problème de l'oracle dans le domaine du test non-fonctionnel des générateurs de code. La disponibilité de multiples générateurs de code avec des fonctionnalités comparables (c.-à-d. familles de générateurs de code) nous permet d'appliquer l'idée du test métamorphique en définissant des oracles de test de haut-niveau (c.-à-d. relation métamorphique) pour détecter des inconsistances. Une inconsistance est détectée lorsque le code généré présente un comportement inattendu par rapport à toutes les implémentations équivalentes de la même famille. Nous évaluons notre approche en analysant la performance de Haxe, un langage de programmation de haut niveau impliquant un ensemble de générateurs de code multi-plateformes. Les résultats expérimentaux montrent que notre approche est capable de détecter plusieurs inconsistances qui révèlent des problèmes réels dans cette famille de générateurs de code. Contribution II: une approche pour l'auto-configuration des compilateurs. Le grand nombre d'options de compilation des compilateurs nécessite une méthode efficace pour explorer l'espace d’optimisation. Ainsi, nous appliquons, dans cette contribution, une méta-heuristique appelée Novelty Search pour l'exploration de cet espace de recherche. Cette approche aide les utilisateurs à paramétrer automatiquement les compilateurs pour une architecture matérielle cible et pour une métrique non-fonctionnelle spécifique tel que la performance et l'utilisation des ressources. Nous évaluons l'efficacité de notre approche en vérifiant les optimisations fournies par le compilateur GCC. Nos résultats expérimentaux montrent que notre approche permet d'auto-configurer les compilateurs en fonction des besoins de l'utilisateur et de construire des optimisations qui surpassent les niveaux d'optimisation standard. Nous démontrons également que notre approche peut être utilisée pour construire automatiquement des niveaux d'optimisation qui représentent des compromis optimaux entre plusieurs propriétés non-fonctionnelles telles que le temps d'exécution et la consommation des ressources. Contribution III: Un environnement d'exécution léger pour le test et la surveillance de la consommation des ressources des logiciels. Enfin, nous proposons une infrastructure basée sur les micro-services pour assurer le déploiement et la surveillance de la consommation des ressources des différentes variantes du code généré. Cette contribution traite le problème de l'hétérogénéité des plateformes logicielles et matérielles. Nous décrivons une approche qui automatise le processus de génération, compilation, et exécution du code dans le but de faciliter le test et l'auto-configuration des générateurs. Cet environnement isolé repose sur des conteneurs système, comme plateformes d'exécution, pour une surveillance et analyse fine des propriétés liées à l'utilisation des ressources (CPU et mémoire). / Generative software development has paved the way for the creation of multiple generators (code generators and compilers) that serve as a basis for automatically producing code to a broad range of software and hardware platforms. With full automatic code generation, users are able to rapidly synthesize software artifacts for various software platforms. In addition, they can easily customize the generated code for the target hardware platform since modern generators (i.e., C compilers) become highly configurable, offering numerous configuration options that the user can apply. Consequently, the quality of generated software becomes highly correlated to the configuration settings as well as to the generator itself. In this context, it is crucial to verify the correct behavior of generators. Numerous approaches have been proposed to verify the functional outcome of generated code but few of them evaluate the non-functional properties of automatically generated code, namely the performance and resource usage properties. This thesis addresses three problems : (1) Non-functional testing of generators: We benefit from the existence of multiple code generators with comparable functionality (i.e., code generator families) to automatically test the generated code. We leverage the metamorphic testing approach to detect non-functional inconsistencies in code generator families by defining metamorphic relations as test oracles. We define the metamorphic relation as a comparison between the variations of performance and resource usage of code, generated from the same code generator family. We evaluate our approach by analyzing the performance of HAXE, a popular code generator family. Experimental results show that our approach is able to automatically detect several inconsistencies that reveal real issues in this family of code generators. (2) Generators auto-tuning: We exploit the recent advances in search-based software engineering in order to provide an effective approach to tune generators (i.e., through optimizations) according to user's non-functional requirements (i.e., performance and resource usage). We also demonstrate that our approach can be used to automatically construct optimization levels that represent optimal trade-offs between multiple non-functional properties such as execution time and resource usage requirements. We evaluate our approach by verifying the optimizations performed by the GCC compiler. Our experimental results show that our approach is able to auto-tune compilers and construct optimizations that yield to better performance results than standard optimization levels. (3) Handling the diversity of software and hardware platforms in software testing: Running tests and evaluating the resource usage in heterogeneous environments is tedious. To handle this problem, we benefit from the recent advances in lightweight system virtualization, in particular container-based virtualization, in order to offer effective support for automatically deploying, executing, and monitoring code in heterogeneous environment, and collect non-functional metrics (e.g., memory and CPU consumptions). This testing infrastructure serves as a basis for evaluating the experiments conducted in the two first contributions.

[pt] A otimização combinatória (OC) está presente em inúmeras aplicações práticas (por exemplo, planejamento de produção, logística, etc.). Ao longo dos anos, OC e aprendizado de máquina (AM) surgiram, juntas, como uma área prospectiva de pesquisa para melhorar processos de tomada de decisão. Nesse contexto, há interesse em utilizar algoritmos de AM para melhorar métodos de OC. Por outro lado, como muitas tarefas de AM podem ser reformuladas como problemas de otimização, há um amplo interesse em utilizar métodos de OC para resolver esses problemas. Nesta tese, três estudos que conectam OC e AM em torno de duas aplicações importantes são conduzidos: o problema de roteamento de veículos capacitado (PRVC) e máquinas de vetores de suporte com perda em margem rígida (SVM-HML – do inglês support vector machines with hard-margin loss). No primeiro estudo, uma estratégia para explorar vizinhanças de busca local de alta ordem por mineração de padrões em duas meta-heurísticas estado da arte para o PRVC é proposta. Em um segundo estudo, também no contexto do PRVC, critérios de relacionamento para nós de clientes baseados em saídas de redes neurais em grafos são explorados. Com base nessas saídas, medidas de relação podem ser exploradas para orientar a busca local e estender operadores de cruzamento em um algoritmo genético estado da arte. Por fim, no terceiro estudo, uma abordagem eficiente de programação inteira mista baseada em cortes combinatórios de Benders e estratégias de amostragem são utilizadas para treinar modelos de SVM-HML de maneira mais eficiente. / [en] Combinatorial optimization (CO) is ubiquitous in myriad practical applications (e.g., production planning, scheduling, logistics, etc.). Over the years, CO and machine learning (ML) have emerged, together, as a prospective area of research for improving decision-making processes. There is interest to harness ML algorithms to improve existing CO methods. Conversely, since many ML tasks can be reformulated as optimization problems, there is broad interest in leveraging state-of-the-art CO methods for them. In this thesis, we conduct three studies that connect CO and ML around two important applications: the capacitated vehicle routing problem (CVRP) and support vector machines with hard-margin loss (SVM-HML). Our first study proposes a strategy to explore high-order local-search neighborhoods by pattern mining into two state-of-the-art metaheuristics for the CVRP. In a second study, also in the context of the CVRP, we exploit relatedness criteria for customer nodes using predictions from graph neural networks. We show that relatedness measures can be exploited to steer local search and extend crossover operators in a stateof- the-art genetic algorithm. Lastly, in a third study, we propose an efficient mixed-integer programming approach based on Combinatorial Benders cuts and sampling strategies for optimally training the SVM-HML.

[pt] Esta pesquisa trata de duas variantes do conhecido Problema de Roteirização de Veículos com Estoque (do inglês Inventory Routing Problem – IRP). O problema nasce num contexto de um sistema de Vendor Managed Inventory (VMI) no qual o fornecedor é responsável pela gestão de estoques do cliente. Tal problema é a junção dos problemas de transporte e gestão de estoques, que correspondem aos maiores custos em uma operação logística. Destarte este trabalho apresenta um modelo matemático para uma variante do IRP que considera que o fornecedor tem clientes dentro e fora do sistema de VMI. Este caso surge quando para alguns clientes não é interessante a realização do controle de seus estoques dentro do sistema de VMI, somente o atendimento de suas demandas. Além disto, o modelo contempla três diferentes tipos de políticas de gestão de estoques e é capaz de lidar com casos contendo vários períodos e vários veículos. Após sua elaboração, o modelo foi validado em instâncias do IRP, do Problema de Roteamento de Veículos (do inglês Capacitated Vehicle Routing Problem - CVRP) e instâncias próprias para a variante. Foram realizados também estudos sobre os impactos das diferentes políticas de gestão de estoques. Além do modelo matemático, foi desenvolvida uma meta-heurística híbrida que resolve uma variante do IRP considerando vários períodos e vários veículos. Cada movimento considerado durante a meta-heurística é divido em duas etapas, a primeira sendo a modificação da posição de um ou mais clientes nos veículos e períodos e uma segunda etapa que resolve de forma exata um Problema de Fluxo Máximo a Custo Mínimo para a atribuição ótima do volume de carga transportada para cada cliente por cada veículo em cada período. Esta abordagem é então testada em instâncias clássicas para esta variante do IRP, obtendo resultados que comprovam a eficiência do algoritmo. / [en] This research deals with two variants of the Inventory Routing Problem (IRP). This problem comes from the context of a Vendor Managed Inventory (VMI) system in which the vendor is responsible for managing the customer s inventory. It is the combination of transportation and inventory management problems, which correspond to the higher costs in a logistics operation. Hence, this paper presents a mathematical model for an IRP variant, in which the vendor has customers inside and outside the VMI system. This situation is presented when it is not interesting to manage the inventories of some clients within the VMI system, resulting only in meeting their demands. In addition, the model considers three different types of stock management policies and it can comprehend multiple periods and multiple vehicles. After its modelling, the model was validated using IRP instaces, the Vehicle Routing Problem (CVRP) and specific instances for this variant. The impacts of different inventory management policies were also analyzed. In addition to the mathematical model, a hybrid meta-heuristic was developed, which solves an IRP variant considering several periods and several vehicles. Each iteration of the metaheuristic is divided into two stages: the first is modifying the position of one or more customers attended by the vehicles and periods, and a second step that solves a Maximum Flow at Minimum Cost problem, to optimally assign the load volumes transported to each customer in each vehicle in each period. Then, this approach is tested in classical instances for this IRP variant, obtaining results that prove the efficiency of the algorithm.

La gestion des ressources, équipements, équipes de travail, et autres, devrait être prise en compte lors de la conception de tout plan réalisable pour le problème de conception de réseaux de services. Cependant, les travaux de recherche portant sur la gestion des ressources et la conception de réseaux de services restent limités. La présente thèse a pour objectif de combler cette lacune en faisant l’examen de problèmes de conception de réseaux de services prenant en compte la gestion des ressources. Pour ce faire, cette thèse se décline en trois études portant sur la conception de réseaux. La première étude considère le problème de capacitated multi-commodity fixed cost network design with design-balance constraints(DBCMND). La structure multi-produits avec capacité sur les arcs du DBCMND, de même que ses contraintes design-balance, font qu’il apparaît comme sous-problème dans de nombreux problèmes reliés à la conception de réseaux de services, d’où l’intérêt d’étudier le DBCMND dans le contexte de cette thèse. Nous proposons une nouvelle approche pour résoudre ce problème combinant la recherche tabou, la recomposition de chemin, et une procédure d’intensification de la recherche dans une région particulière de l’espace de solutions. Dans un premier temps la recherche tabou identifie de bonnes solutions réalisables. Ensuite la recomposition de chemin est utilisée pour augmenter le nombre de solutions réalisables. Les solutions trouvées par ces deux méta-heuristiques permettent d’identifier un sous-ensemble d’arcs qui ont de bonnes chances d’avoir un statut ouvert ou fermé dans une solution optimale. Le statut de ces arcs est alors fixé selon la valeur qui prédomine dans les solutions trouvées préalablement. Enfin, nous utilisons la puissance d’un solveur de programmation mixte en nombres entiers pour intensifier la recherche sur le problème restreint par le statut fixé ouvert/fermé de certains arcs. Les tests montrent que cette approche est capable de trouver de bonnes solutions aux problèmes de grandes tailles dans des temps raisonnables. Cette recherche est publiée dans la revue scientifique Journal of heuristics. La deuxième étude introduit la gestion des ressources au niveau de la conception de réseaux de services en prenant en compte explicitement le nombre fini de véhicules utilisés à chaque terminal pour le transport de produits. Une approche de solution faisant appel au slope-scaling, la génération de colonnes et des heuristiques basées sur une formulation en cycles est ainsi proposée. La génération de colonnes résout une relaxation linéaire du problème de conception de réseaux, générant des colonnes qui sont ensuite utilisées par le slope-scaling. Le slope-scaling résout une approximation linéaire du problème de conception de réseaux, d’où l’utilisation d’une heuristique pour convertir les solutions obtenues par le slope-scaling en solutions réalisables pour le problème original. L’algorithme se termine avec une procédure de perturbation qui améliore les solutions réalisables. Les tests montrent que l’algorithme proposé est capable de trouver de bonnes solutions au problème de conception de réseaux de services avec un nombre fixe des ressources à chaque terminal. Les résultats de cette recherche seront publiés dans la revue scientifique Transportation Science. La troisième étude élargie nos considérations sur la gestion des ressources en prenant en compte l’achat ou la location de nouvelles ressources de même que le repositionnement de ressources existantes. Nous faisons les hypothèses suivantes: une unité de ressource est nécessaire pour faire fonctionner un service, chaque ressource doit retourner à son terminal d’origine, il existe un nombre fixe de ressources à chaque terminal, et la longueur du circuit des ressources est limitée. Nous considérons les alternatives suivantes dans la gestion des ressources: 1) repositionnement de ressources entre les terminaux pour tenir compte des changements de la demande, 2) achat et/ou location de nouvelles ressources et leur distribution à différents terminaux, 3) externalisation de certains services. Nous présentons une formulation intégrée combinant les décisions reliées à la gestion des ressources avec les décisions reliées à la conception des réseaux de services. Nous présentons également une méthode de résolution matheuristique combinant le slope-scaling et la génération de colonnes. Nous discutons des performances de cette méthode de résolution, et nous faisons une analyse de l’impact de différentes décisions de gestion des ressources dans le contexte de la conception de réseaux de services. Cette étude sera présentée au XII International Symposium On Locational Decision, en conjonction avec XXI Meeting of EURO Working Group on Locational Analysis, Naples/Capri (Italy), 2014. En résumé, trois études différentes sont considérées dans la présente thèse. La première porte sur une nouvelle méthode de solution pour le "capacitated multi-commodity fixed cost network design with design-balance constraints". Nous y proposons une matheuristique comprenant la recherche tabou, la recomposition de chemin, et l’optimisation exacte. Dans la deuxième étude, nous présentons un nouveau modèle de conception de réseaux de services prenant en compte un nombre fini de ressources à chaque terminal. Nous y proposons une matheuristique avancée basée sur la formulation en cycles comprenant le slope-scaling, la génération de colonnes, des heuristiques et l’optimisation exacte. Enfin, nous étudions l’allocation des ressources dans la conception de réseaux de services en introduisant des formulations qui modèlent le repositionnement, l’acquisition et la location de ressources, et l’externalisation de certains services. À cet égard, un cadre de solution slope-scaling développé à partir d’une formulation en cycles est proposé. Ce dernier comporte la génération de colonnes et une heuristique. Les méthodes proposées dans ces trois études ont montré leur capacité à trouver de bonnes solutions. / Resource management in freight transportation service network design is an important issue that has been studied extensively in recent years. Resources such as vehicles, crews, etc. are factors that can not be ignored when designing a feasible plan for any service network design problem. However, contributions related to resource management issues and service network design are still limited. The goal of the thesis is to fill this gap by taking into account service network design problems with resource management issues. In this thesis, we propose and address three service network design problems that consider resource management. In the first study, we consider the capacitated multi-commodity fixed cost network design with design-balance constraints which is a basic sub-problem for many service design problems because of the capacitated multi-commodity structure as well as its design-balance property. We propose a three-phase matheuristic that combines tabu-search, path-relinking and an exactbased intensification procedure to find high quality solutions. Tabu-search identifies feasible solutions while path-relinking extends the set of feasible solutions. The solutions found by these two meta-heuristics are used to fix arcs as open or close. An exact solver intensifies the search on a restricted problem derived from fixing arcs. The experiments on benchmark instances show that the solution approach finds good solutions to large-scale problems in a reasonable amount of time. The contribution with regard to this study has been accepted in the Journal of Heuristics. In the second study, together with the consideration of the design of routes to transport a set of commodities by vehicles, we extend resources management by explicitly taking account of the number of available vehicles at each terminal. We introduce a matheuristic solution framework based on a cycle-based formulation that includes column generation, slope-scaling, heuristic and exact optimization techniques. As far as we know, this is the first matheuristic procedure developed for a cycle-based formulation. The column generation solves the linear relaxation model and provides a set of cycles to define the approximation model used in slopescaling loop. A heuristic is used to convert each solution to the approximation problem into a feasible solution. Memory-based perturbation procedure is used to enhance the performance of the algorithm. Experiments show that the proposed algorithm is able to find good feasible solutions for the problem. The contribution with regard to this study has been accepted for publication in Transportation Science. In the third study, we examine resources allocation issues in service network design. We aim to address a number of fleet utilization issues which usually appear at the beginning of the season because of the change of demand patterns: 1) reposition resources among terminals to account for shifts in demand patterns; 2) acquire (buy or long-term rent) new resources and as sign them to terminals; 3) outsource particular services. We present an integrated formulation combining these selection-location and scheduled service design decisions. The mixed-integer formulation is defined over a time-space network, the initial period modeling the location de cisions on resource acquisition and positioning, while the decisions on service selection and scheduling, resource assignment and cycling routing, and demand satisfaction being modeled on the rest of the network. We also present a matheuristic solution method combining slope scaling and column generation, discuss its algorithmic performance, and explore the impact of combining the location and design decisions in the context of consolidation carrier service design. This study will be presented at XII International Symposium On Locational Deci sion, in conjunction with the XXI Meeting of EURO Working Group on Locational Analysis, Naples/Capri (Italy), 2014. In summary, three studies are considered in this thesis. The first one considers the capaciated multi-commodity fixed cost network design with design-balance constraints, a basic problem in many service network design problems with design-balance constraints. We propose an ef ficient three-phase matheuristic solution method that includes tabu search, path relinking and exact optimization. In the second study, we propose a new service network design model that takes into account resources limitations at each terminal. We also propose an advanced matheuristic framework solution method based on a cycle-based formulation which includes slope-scaling, column generation, heuristics and exact optimization for this problem. The last study addresses resources allocation issues in service network design. We introduce formula tions that model the reposition, acquisition/renting of resources and outsourcing of services. A solution framework based on the slope-scaling approach on cycle-based formulations is pro posed. Tests indicate that these proposed algorithms are able to find good feasible solutions for each of threse problems.