L’optimisation offre la possibilité, dans de nombreux domaines, d’améliorer les performances d’unsystème donné, qu’il soit physique ou mathématique. Depuis quelques décennies, les méthodesd’optimisation metaheuristiques ont fait leurs preuves, notamment dans le domaine de la mécanique.Du grec meta signifiant "un niveau au dessus", les metaheuristiques permettent de s’affranchir ducalcul des sensibilités souvent problématique quant à la résolution de problèmes d’optimisationcomplexes et/ou NP difficiles. En outre, elles ont la capacité à analyser simultanément l’ensemble dudomaine des solutions, ce qui leur permet converger efficacement vers l’optimum global de la fonctionobjectif considérée. Notre travail propose le développement d’une nouvelle méthode metaheuristiqueintelligente, basée conjointement sur l’algorithme d’optimisation par essaim particulaire PSO, etl’algorithme PageRank développé par MM. Brin et Page, et utilisé par le moteur de rechercheGoogle. Cet algorithme, appelé Inverse-PageRank-PSO (I-PR-PSO), a été validé sur un benchmarkde fonctions mathématiques, puis en optimisation contrainte sur des treillis mécaniques. Interfacéeavec l’algorithme Evolutionary Structural Optimization (ESO), elle a été adaptée à l’optimisationtopologique et a permis de trouver des résultats dont les topologies sont régulières et les temps decalcul minimisés. Dans le domaine des metamatériaux, nous avons développé une cape d’invisibilitéélectromagnétique fréquentielle, c’est à dire un metamatériau dont les parties réelle et imaginaire dela perméabilité effective sont négatives. En appliquant notre algorithme I-PR-PSO aux metamatériauxmécaniques, nous avons montré qu’il est possible de développer un metamatériau constitué d’acierqui présente des grandes déformations à l’échelle macroscopique, dues notamment aux grandsdéplacements présents dans le Volume Elémentaire Représentatif à l’échelle microscopique. / Based on a recent research concerning the PageRank algorithm used by the famous search engineGoogle, a new Inverse-PageRank-Particle Swarm Optimizer (I-PR-PSO) is developed, in order toimprove the performances of classic PSO. After having been tested and validated on a benchmarkof classical mathematical functions, this algorithm has been validated on constrained optimization,applied on classical trusses of the literature. Interfaced with the Evolutionary Structural Optimizationalgorithm, this algorithm has shown its performances on topology optimization, applied to structuralmechanics. Finally, using the performances of our newly developed algorithm, we have developedmetamaterials. In electromagnetics, a frequantial cloaking device has been developed, minimizingthe effective permeability of the considered Representative Volume Element. In mechanics, we havedeveloped a metamaterial made of steel which exhibits hyper-elastic - or, at least, non linear -mechanical behaviour. Combining great displacements and rotations at microscale, the developedmetamaterial exhibits great deformations at the macroscale as well.

Ce travail présente nos principales contributions à la résolution de problèmes d'optimisation combinatoire en environnements déterministe et stochastique. Au niveau des métaheuristiques, une vue unifiée de la conception de métaheuristiques à solution unique et de métaheuristiques multi-objective est proposée. Cette unification a permis notamment de retravailler la plateforme ParadisEO afin d'offrir plus de flexibilité et de polyvalence. La synthèse des travaux présente également une vue unifiée des métaheuristiques coopératives. Nous montrons que cette vue convient aussi bien pour des coopérations entre métaheuristiques que des coopération entre des métaheuristiques et des méthodes exactes mais également des coopérations entre des métaheuristiques et des algorithmes d'extraction de connaissances. Différents exemples de coopérations réalisées dans mes travaux de recherche illustent ces coopérations et leur application à des problèmes d'optimisation combinatoire mono- et multi-objectif. Cette habilitation se termine par la présentation de travaux réalisés en optimisation stochastique notamment dans le cadre de l'optimisation sous incertitude et de l'optimisation dynamique. L'importance des critères de robustesse est discutée ainsi que l'intérêt et la mise en œuvre de méthodes coopératives dans un contexte dynamique. Les principales applications présentées on été réalisées sur des problèmes en transport et logistique ainsi qu'en biologie dans le cadre de l'ANR Dock.

Cette thèse, réalisée en collaboration avec Sagem Défense Sécurité, porte sur l'élaboration d'une stratégie de recherche efficace pour la résolution de problèmes de planification d'itinéraires de véhicules. Nous considérons ici en particulier les problèmes de planification avec contraintes de points de passage et de "capacité" (énergie, bande passante radio) appliquées au véhicule. Ce document propose une approche originale, hybridant un algorithme de colonies de fourmis avec un solveur de Programmation par Contraintes existant. Le premier est utilisé pour résoudre rapidement une version relaxée du problème. La solution partielle obtenue est alors employée pour guider la recherche du second, par le biais d'une méthode de sonde, vers les zones les plus prometteuses de l'espace d'état. Cette approche permet ainsi de combiner la rapidité des métaheuristiques et la complétude de la programmation par contraintes. Nous montrons expérimentalement que cette approche satisfait les exigences pour une utilisation du planificateur dans un cadre embarqué.

Dans la pratique, beaucoup de problèmes d'optimisation sont dynamiques : leur fonction objectif (ou fonction de coût) évolue au cours du temps. L'approche principalement adoptée dans la littérature consiste à adapter des algorithmes d'optimisation statique à l'optimisation dynamique, en compensant leurs défauts intrinsèques. Plutôt que d'emprunter cette voie, déjà largement explorée, l'objectif principal de cette thèse est d'élaborer un algorithme entièrement pensé pour l'optimisation dynamique. La première partie de cette thèse est ainsi consacrée à la mise au point d'un algorithme, qui doit non seulement se démarquer des algorithmes concurrents par son originalité, mais également être plus performant. Dans ce contexte, il s'agit de développer une métaheuristique d'optimisation dynamique. Deux algorithmes à base d'agents, MADO (MultiAgent algorithm for Dynamic Optimization) et MLSDO (Multiple Local Search algorithm for Dynamic Optimization), sont proposés et validés sur les deux principaux jeux de tests existant dans la littérature en optimisation dynamique : MPB (Moving Peaks Benchmark) et GDBG (Generalized Dynamic Benchmark Generator). Les résultats obtenus sur ces jeux de tests montrent l'efficacité des stratégies mises en oeuvre par ces algorithmes, en particulier : MLSDO est classé premier sur sept algorithmes évalués sur GDBG, et deuxième sur seize algorithmes évalués sur MPB. Ensuite, ces algorithmes sont appliqués à des problèmes pratiques en traitement de séquences d'images médicales (segmentation et recalage de séquences ciné-IRM cérébrales). A notre connaissance, ce travail est innovant, en ce sens que l'approche de l'optimisation dynamique n'avait jamais été explorée pour ces problèmes. Les gains de performance obtenus montrent l'intérêt d'utiliser les algorithmes d'optimisation dynamique proposés pour ce type d'applications

Dans la pratique, beaucoup de problèmes d'optimisation sont dynamiques : leur fonction objectif (ou fonction de coût) évolue au cours du temps. L'approche principalement adoptée dans la littérature consiste à adapter des algorithmes d'optimisation statique à l'optimisation dynamique, en compensant leurs défauts intrinsèques. Plutôt que d'emprunter cette voie, déjà largement explorée, l'objectif principal de cette thèse est d'élaborer un algorithme entièrement pensé pour l'optimisation dynamique. La première partie de cette thèse est ainsi consacrée à la mise au point d'un algorithme, qui doit non seulement se démarquer des algorithmes concurrents par son originalité, mais également être plus performant. Dans ce contexte, il s'agit de développer une métaheuristique d'optimisation dynamique. Deux algorithmes à base d'agents, MADO (MultiAgent algorithm for Dynamic Optimization) et MLSDO (Multiple Local Search algorithm for Dynamic Optimization), sont proposés et validés sur les deux principaux jeux de tests existant dans la littérature en optimisation dynamique : MPB (Moving Peaks Benchmark) et GDBG (Generalized Dynamic Benchmark Generator). Les résultats obtenus sur ces jeux de tests montrent l'efficacité des stratégies mises en oeuvre par ces algorithmes, en particulier : MLSDO est classé premier sur sept algorithmes évalués sur GDBG, et deuxième sur seize algorithmes évalués sur MPB. Ensuite, ces algorithmes sont appliqués à des problèmes pratiques en traitement de séquences d'images médicales (segmentation et recalage de séquences ciné-IRM cérébrales). A notre connaissance, ce travail est innovant, en ce sens que l'approche de l'optimisation dynamique n'avait jamais été explorée pour ces problèmes. Les gains de performance obtenus montrent l'intérêt d'utiliser les algorithmes d'optimisation dynamique proposés pour ce type d'applications / Many real-world problems are dynamic, i.e. their objective function (or cost function) changes over time. The main approach used in the literature is to adapt static optimization algorithms to dynamic optimization, compensating for their intrinsic defects. Rather than adopting this approach, already widely investigated, the main goal of this thesis is to develop an algorithm completely designed for dynamic optimization. The first part of this thesis is then devoted to the design of an algorithm, that should not only stand out from competing algorithms for its originality, but also perform better. In this context, our goal is to develop a dynamic optimization metaheuristic. Two agent-based algorithms, MADO (MultiAgent algorithm for Dynamic Optimization) and MLSDO (Multiple Local Search algorithm for Dynamic Optimization), are proposed and validated using the two main benchmarks available in dynamic environments : MPB (Moving Peaks Benchmark) and GDBG (Generalized Dynamic Benchmark Generator). The benchmark results obtained show the efficiency of the proposed algorithms, particularly : MLSDO is ranked at the first place among seven algorithms tested using GDBG, and at the second place among sixteen algorithms tested using MPB. Then, these algorithms are applied to real-world problems in medical image sequence processing (segmentation and registration of brain cine-MRI sequences). To our knowledge, this work is innovative in that the dynamic optimization approach had never been investigated for these problems. The performance gains obtained show the relevance of using the proposed dynamic optimization algorithms for this kind of applications

Cette thèse considère le problème de sélection et d'ordonnancement des prises de vue d'un satellite agile d'observation de la Terre. La mission d'un satellite d'observation est d'obtenir des photographies de la surface de la Terre afin de satisfaire des requêtes d'utilisateurs. Les demandes, émanant de différents utilisateurs, doivent faire l'objet d'un traitement avant transmission d'un ordre vers le satellite, correspondant à une séquence d'acquisitions sélectionnées. Cette séquence doit optimiser deux objectifs sous contraintes d'exploitation. Le premier objectif est de maximiser le profit global des acquisitions sélectionnées. Le second est d'assurer l'équité du partage des ressources en minimisant la différence maximale de profit entre les utilisateurs. Deux métaheuristiques, composées d'un algorithme génétique à clé aléatoire biaisées (biased random key genetic algorithm - BRKGA) et d'une recherche locale multi-objectif basée sur des indicateurs (indicator based multi-objective local search - IBMOLS), sont proposées pour résoudre le problème. Pour BRKGA, trois méthodes de sélection, empruntées à NSGA-II, SMS-EMOA, et IBEA, sont proposées pour choisir un ensemble de chromosomes préférés comme ensemble élite. Trois stratégies de décodage, parmi lesquelles deux sont des décodages uniques et la dernière un décodage hybride, sont appliquées pour décoder les chromosomes afin d'obtenir des solutions. Pour IBMOLS, plusieurs méthodes pour générer la population initiale sont testées et une structure de voisinage est également proposée. Des expériences sont menées sur des cas réalistes, issus d'instances modifiées du challenge ROADEF 2003. On obtient ainsi les fronts de Pareto approximés de BRKGA et IBMOLS dont on calcule les hypervolumes. Les résultats de ces deux algorithmes sont comparés.

Cette thèse, réalisée en collaboration avec Sagem Défense Sécurité, porte sur l'élaboration d'une stratégie de recherche efficace pour la résolution de problèmes de planification d'itinéraires de véhicules. Nous considérons ici en particulier les problèmes de planification avec contraintes de points de passage et de "capacité" (énergie, bande passante radio) appliquées au véhicule. Ce document propose une approche originale, hybridant un algorithme de colonies de fourmis avec un solveur de Programmation par Contraintes existant. Le premier est utilisé pour résoudre rapidement une version relaxée du problème. La solution partielle obtenue est alors employée pour guider la recherche du second, par le biais d'une méthode de sonde, vers les zones les plus prometteuses de l'espace d'état. Cette approche permet ainsi de combiner la rapidité des métaheuristiques et la complétude de la programmation par contraintes. Nous montrons expérimentalement que cette approche satisfait les exigences pour une utilisation du planificateur dans un cadre embarqué. / This thesis, led in collaboration with Sagem Defence & Security, focuses on defining an efficient search strategy to solve vehicle path planning problems. This work addresses more precisely planning problems in which waypoints and "capacity" constraints (energy, radio bandwidth) are applied to vehicles.This document proposes an original approach, mixing an Ant Colony algorithm with an existing Constraint Programming solver. The former is used to fastly solve a relaxed version of the problem. The partial solution returned is then employed to guide the search of the latter, through a Probe Backtrack mechanism, towards the most promising areas of the state space. This approach allows to combine the metaheuristics solving fastness and the Constraint Programming completeness. We experimentally show that this approach meets the requirements for an on-line use of the planner.

Cette thèse considère le problème de sélection et d’ordonnancement des prises de vue d’un satellite agile d’observation de la Terre. La mission d’un satellite d’observation est d’obtenir des photographies de la surface de la Terre afin de satisfaire des requêtes d’utilisateurs. Les demandes, émanant de différents utilisateurs, doivent faire l’objet d’un traitement avant transmission d’un ordre vers le satellite, correspondant à une séquence d’acquisitions sélectionnées. Cette séquence doit optimiser deux objectifs sous contraintes d’exploitation. Le premier objectif est de maximiser le profit global des acquisitions sélectionnées. Le second est d’assurer l’équité du partage des ressources en minimisant la différence maximale de profit entre les utilisateurs. Deux métaheuristiques, composées d’un algorithme génétique à clé aléatoire biaisées (biased random key genetic algorithm - BRKGA) et d’une recherche locale multi-objectif basée sur des indicateurs (indicator based multi-objective local search - IBMOLS), sont proposées pour résoudre le problème. Pour BRKGA, trois méthodes de sélection, empruntées à NSGA-II, SMS-EMOA, et IBEA, sont proposées pour choisir un ensemble de chromosomes préférés comme ensemble élite. Trois stratégies de décodage, parmi lesquelles deux sont des décodages uniques et la dernière un décodage hybride, sont appliquées pour décoder les chromosomes afin d’obtenir des solutions. Pour IBMOLS, plusieurs méthodes pour générer la population initiale sont testées et une structure de voisinage est également proposée. Des expériences sont menées sur des cas réalistes, issus d’instances modifiées du challenge ROADEF 2003. On obtient ainsi les fronts de Pareto approximés de BRKGA et IBMOLS dont on calcule les hypervolumes. Les résultats de ces deux algorithmes sont comparés / This thesis considers the selection and scheduling problem of observations for agile Earth observing satellites. The mission of Earth observing satellites is to obtain photographs of the Earth surface to satisfy user requirements. Requests from several users have to be managed before transmitting an order, which is a sequence of selected acquisitions, to the satellite. The obtained sequence must optimize two objectives under operation constraints. The first objective is to maximize the total profit of the selected acquisitions. The second one is to ensure the fairness of resource sharing by minimizing the maximum profit difference between users. Two metaheuristic algorithms, consisting of a biased random key genetic algorithm (BRKGA) and an indicator-based multi-objective local search (IBMOLS), are proposed to solve the problem. For BRKGA, three selection methods, borrowed from NSGA-II, SMS-EMOA, and IBEA, are proposed to select a set of preferred chromosomes to be the elite set. Three decoding strategies, which are two single decoding and a hybrid decoding, are applied to decode chromosomes to become solutions. For IBMOLS, several methods for generating the initial population are tested and the neighborhood structure according to the problem is also proposed. Experiments are conducted on realistic instances based on ROADEF 2003 challenge instances. Hypervolumes of the approximate Pareto fronts are computed and the results from the two algorithms are compared

In ant societies, and, more in general, in insect societies, the activities of the individuals, as well as of the society as a whole, are not regulated by any explicit form of centralized control. On the other hand, adaptive and robust behaviors transcending the behavioral repertoire of the single individual can be easily observed at society level. These complex global behaviors are the result of self-organizing dynamics driven by local interactions and communications among a number of relatively simple individuals. The simultaneous presence of these and other fascinating and unique characteristics have made ant societies an attractive and inspiring model for building new algorithms and new multi-agent systems. In the last decade, ant societies have been taken as a reference for an ever growing body of scientific work, mostly in the fields of robotics, operations research, and telecommunications. Among the different works inspired by ant colonies, the Ant Colony Optimization metaheuristic (ACO) is probably the most successful and popular one. The ACO metaheuristic is a multi-agent framework for combinatorial optimization whose main components are: a set of ant-like agents, the use of memory and of stochastic decisions, and strategies of collective and distributed learning. It finds its roots in the experimental observation of a specific foraging behavior of some ant colonies that, under appropriate conditions, are able to select the shortest path among few possible paths connecting their nest to a food site. The pheromone, a volatile chemical substance laid on the ground by the ants while walking and affecting in turn their moving decisions according to its local intensity, is the mediator of this behavior. All the elements playing an essential role in the ant colony foraging behavior were understood, thoroughly reverse-engineered and put to work to solve problems of combinatorial optimization by Marco Dorigo and his co-workers at the beginning of the 1990's. From that moment on it has been a flourishing of new combinatorial optimization algorithms designed after the first algorithms of Dorigo's et al., and of related scientific events. In 1999 the ACO metaheuristic was defined by Dorigo, Di Caro and Gambardella with the purpose of providing a common framework for describing and analyzing all these algorithms inspired by the same ant colony behavior and by the same common process of reverse-engineering of this behavior. Therefore, the ACO metaheuristic was defined a posteriori, as the result of a synthesis effort effectuated on the study of the characteristics of all these ant-inspired algorithms and on the abstraction of their common traits. The ACO's synthesis was also motivated by the usually good performance shown by the algorithms (e.g., for several important combinatorial problems like the quadratic assignment, vehicle routing and job shop scheduling, ACO implementations have outperformed state-of-the-art algorithms). The definition and study of the ACO metaheuristic is one of the two fundamental goals of the thesis. The other one, strictly related to this former one, consists in the design, implementation, and testing of ACO instances for problems of adaptive routing in telecommunication networks. This thesis is an in-depth journey through the ACO metaheuristic, during which we have (re)defined ACO and tried to get a clear understanding of its potentialities, limits, and relationships with other frameworks and with its biological background. The thesis takes into account all the developments that have followed the original 1999's definition, and provides a formal and comprehensive systematization of the subject, as well as an up-to-date and quite comprehensive review of current applications. We have also identified in dynamic problems in telecommunication networks the most appropriate domain of application for the ACO ideas. According to this understanding, in the most applicative part of the thesis we have focused on problems of adaptive routing in networks and we have developed and tested four new algorithms. Adopting an original point of view with respect to the way ACO was firstly defined (but maintaining full conceptual and terminological consistency), ACO is here defined and mainly discussed in the terms of sequential decision processes and Monte Carlo sampling and learning. More precisely, ACO is characterized as a policy search strategy aimed at learning the distributed parameters (called pheromone variables in accordance with the biological metaphor) of the stochastic decision policy which is used by so-called ant agents to generate solutions. Each ant represents in practice an independent sequential decision process aimed at constructing a possibly feasible solution for the optimization problem at hand by using only information local to the decision step. Ants are repeatedly and concurrently generated in order to sample the solution set according to the current policy. The outcomes of the generated solutions are used to partially evaluate the current policy, spot the most promising search areas, and update the policy parameters in order to possibly focus the search in those promising areas while keeping a satisfactory level of overall exploration. This way of looking at ACO has facilitated to disclose the strict relationships between ACO and other well-known frameworks, like dynamic programming, Markov and non-Markov decision processes, and reinforcement learning. In turn, this has favored reasoning on the general properties of ACO in terms of amount of complete state information which is used by the ACO's ants to take optimized decisions and to encode in pheromone variables memory of both the decisions that belonged to the sampled solutions and their quality. The ACO's biological context of inspiration is fully acknowledged in the thesis. We report with extensive discussions on the shortest path behaviors of ant colonies and on the identification and analysis of the few nonlinear dynamics that are at the very core of self-organized behaviors in both the ants and other societal organizations. We discuss these dynamics in the general framework of stigmergic modeling, based on asynchronous environment-mediated communication protocols, and (pheromone) variables priming coordinated responses of a number of ``cheap' and concurrent agents. The second half of the thesis is devoted to the study of the application of ACO to problems of online routing in telecommunication networks. This class of problems has been identified in the thesis as the most appropriate for the application of the multi-agent, distributed, and adaptive nature of the ACO architecture. Four novel ACO algorithms for problems of adaptive routing in telecommunication networks are throughly described. The four algorithms cover a wide spectrum of possible types of network: two of them deliver best-effort traffic in wired IP networks, one is intended for quality-of-service (QoS) traffic in ATM networks, and the fourth is for best-effort traffic in mobile ad hoc networks. The two algorithms for wired IP networks have been extensively tested by simulation studies and compared to state-of-the-art algorithms for a wide set of reference scenarios. The algorithm for mobile ad hoc networks is still under development, but quite extensive results and comparisons with a popular state-of-the-art algorithm are reported. No results are reported for the algorithm for QoS, which has not been fully tested. The observed experimental performance is excellent, especially for the case of wired IP networks: our algorithms always perform comparably or much better than the state-of-the-art competitors. In the thesis we try to understand the rationale behind the brilliant performance obtained and the good level of popularity reached by our algorithms. More in general, we discuss the reasons of the general efficacy of the ACO approach for network routing problems compared to the characteristics of more classical approaches. Moving further, we also informally define Ant Colony Routing (ACR), a multi-agent framework explicitly integrating learning components into the ACO's design in order to define a general and in a sense futuristic architecture for autonomic network control. Most of the material of the thesis comes from a re-elaboration of material co-authored and published in a number of books, journal papers, conference proceedings, and technical reports. The detailed list of references is provided in the Introduction.

In ant societies, and, more in general, in insect societies, the activities of the individuals, as well as of the society as a whole, are not regulated by any explicit form of centralized control. On the other hand, adaptive and robust behaviors transcending the behavioral repertoire of the single individual can be easily observed at society level. These complex global behaviors are the result of self-organizing dynamics driven by local interactions and communications among a number of relatively simple individuals.<p><p>The simultaneous presence of these and other fascinating and unique characteristics have made ant societies an attractive and inspiring model for building new algorithms and new multi-agent systems. In the last decade, ant societies have been taken as a reference for an ever growing body of scientific work, mostly in the fields of robotics, operations research, and telecommunications.<p><p>Among the different works inspired by ant colonies, the Ant Colony Optimization metaheuristic (ACO) is probably the most successful and popular one. The ACO metaheuristic is a multi-agent framework for combinatorial optimization whose main components are: a set of ant-like agents, the use of memory and of stochastic decisions, and strategies of collective and distributed learning.<p><p>It finds its roots in the experimental observation of a specific foraging behavior of some ant colonies that, under appropriate conditions, are able to select the shortest path among few possible paths connecting their nest to a food site. The pheromone, a volatile chemical substance laid on the ground by the ants while walking and affecting in turn their moving decisions according to its local intensity, is the mediator of this behavior.<p><p>All the elements playing an essential role in the ant colony foraging behavior were understood, thoroughly reverse-engineered and put to work to solve problems of combinatorial optimization by Marco Dorigo and his co-workers at the beginning of the 1990's.<p><p>From that moment on it has been a flourishing of new combinatorial optimization algorithms designed after the first algorithms of Dorigo's et al. and of related scientific events.<p><p>In 1999 the ACO metaheuristic was defined by Dorigo, Di Caro and Gambardella with the purpose of providing a common framework for describing and analyzing all these algorithms inspired by the same ant colony behavior and by the same common process of reverse-engineering of this behavior. Therefore, the ACO metaheuristic was defined a posteriori, as the result of a synthesis effort effectuated on the study of the characteristics of all these ant-inspired algorithms and on the abstraction of their common traits.<p><p>The ACO's synthesis was also motivated by the usually good performance shown by the algorithms (e.g. for several important combinatorial problems like the quadratic assignment, vehicle routing and job shop scheduling, ACO implementations have outperformed state-of-the-art algorithms).<p><p>The definition and study of the ACO metaheuristic is one of the two fundamental goals of the thesis. The other one, strictly related to this former one, consists in the design, implementation, and testing of ACO instances for problems of adaptive routing in telecommunication networks.<p><p>This thesis is an in-depth journey through the ACO metaheuristic, during which we have (re)defined ACO and tried to get a clear understanding of its potentialities, limits, and relationships with other frameworks and with its biological background. The thesis takes into account all the developments that have followed the original 1999's definition, and provides a formal and comprehensive systematization of the subject, as well as an up-to-date and quite comprehensive review of current applications. We have also identified in dynamic problems in telecommunication networks the most appropriate domain of application for the ACO ideas. According to this understanding, in the most applicative part of the thesis we have focused on problems of adaptive routing in networks and we have developed and tested four new algorithms.<p><p>Adopting an original point of view with respect to the way ACO was firstly defined (but maintaining full conceptual and terminological consistency), ACO is here defined and mainly discussed in the terms of sequential decision processes and Monte Carlo sampling and learning.<p><p>More precisely, ACO is characterized as a policy search strategy aimed at learning the distributed parameters (called pheromone variables in accordance with the biological metaphor) of the stochastic decision policy which is used by so-called ant agents to generate solutions. Each ant represents in practice an independent sequential decision process aimed at constructing a possibly feasible solution for the optimization problem at hand by using only information local to the decision step.<p>Ants are repeatedly and concurrently generated in order to sample the solution set according to the current policy. The outcomes of the generated solutions are used to partially evaluate the current policy, spot the most promising search areas, and update the policy parameters in order to possibly focus the search in those promising areas while keeping a satisfactory level of overall exploration.<p><p>This way of looking at ACO has facilitated to disclose the strict relationships between ACO and other well-known frameworks, like dynamic programming, Markov and non-Markov decision processes, and reinforcement learning. In turn, this has favored reasoning on the general properties of ACO in terms of amount of complete state information which is used by the ACO's ants to take optimized decisions and to encode in pheromone variables memory of both the decisions that belonged to the sampled solutions and their quality.<p><p>The ACO's biological context of inspiration is fully acknowledged in the thesis. We report with extensive discussions on the shortest path behaviors of ant colonies and on the identification and analysis of the few nonlinear dynamics that are at the very core of self-organized behaviors in both the ants and other societal organizations. We discuss these dynamics in the general framework of stigmergic modeling, based on asynchronous environment-mediated communication protocols, and (pheromone) variables priming coordinated responses of a number of ``cheap' and concurrent agents.<p><p>The second half of the thesis is devoted to the study of the application of ACO to problems of online routing in telecommunication networks. This class of problems has been identified in the thesis as the most appropriate for the application of the multi-agent, distributed, and adaptive nature of the ACO architecture.<p><p>Four novel ACO algorithms for problems of adaptive routing in telecommunication networks are throughly described. The four algorithms cover a wide spectrum of possible types of network: two of them deliver best-effort traffic in wired IP networks, one is intended for quality-of-service (QoS) traffic in ATM networks, and the fourth is for best-effort traffic in mobile ad hoc networks.<p><p>The two algorithms for wired IP networks have been extensively tested by simulation studies and compared to state-of-the-art algorithms for a wide set of reference scenarios. The algorithm for mobile ad hoc networks is still under development, but quite extensive results and comparisons with a popular state-of-the-art algorithm are reported. No results are reported for the algorithm for QoS, which has not been fully tested. The observed experimental performance is excellent, especially for the case of wired IP networks: our algorithms always perform comparably or much better than the state-of-the-art competitors.<p><p>In the thesis we try to understand the rationale behind the brilliant performance obtained and the good level of popularity reached by our algorithms. More in general, we discuss the reasons of the general efficacy of the ACO approach for network routing problems compared to the characteristics of more classical approaches. Moving further, we also informally define Ant Colony Routing (ACR), a multi-agent framework explicitly integrating learning components into the ACO's design in order to define a general and in a sense futuristic architecture for autonomic network control.<p><p>Most of the material of the thesis comes from a re-elaboration of material co-authored and published in a number of books, journal papers, conference proceedings, and technical reports. The detailed list of references is provided in the Introduction.<p><p><p> / Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished