A Inteligência Artificial Distribuída traz uma série de novas perspectivas para a computação quando considera sistemas heterogêneos, adaptativos, evolutivos, continuamente em funcionamento e abertos. Estes sistemas, chamados de sociedades, apresentam tais características por permitirem que seus componentes, chamados de agentes, migrem entre sociedades, isto é, agentes podem sair e entrar em sociedades. Sociedades abertas permitem a migração dos agentes e coloca dois tipos de problemas para o agente que está migrando: problemas de linguagem e interação, que concernem ao uso de expressões usadas e à maneira como as interações são organizadas na nova sociedade; e, problemas de conhecimento e atuação, que se referem à como um agente irá se comportar a fim de realizar justamente aquilo que a sociedade espera dele. Este trabalho se atem aos problemas de conhecimento e atuação. Para que os agentes da sociedade possam cooperar e coordenar suas ações, é necessário que tenham conhecimento das capacidades, habilidades, desejos e planos dos outros agentes. Grande parte do conhecimento a respeito dos outros pode ser extraído dos papéis que estes podem assumir na sociedade. Assim sendo, o problema colocado para este trabalho é como os agentes da sociedade que receberam o agente imigrante e o próprio agente imigrante conhecerão/aprenderão os papéis uns dos outros. São desenvolvidos três mecanismos de identificação de papéis, bem como a comparação entre eles e sua adequação a tipos de migração. Os três mecanismos são os seguintes: i) Identificação de papéis por protocolo de apresentação: é proposta uma linguagem de descrição de protocolos (LDP) e uma especificação de protocolo de apresentação nesta LDP. Os agentes que utilizam este mecanismo conseguem se identificar com rapidez, porém necessitam conhecer várias informações “locais” da sociedade, o que pode ser muito restritivo para um agente migrante. ii) Identificação de papéis por observação e classificação: esta solução procura classificar o agente observado em um papel de um conjunto prédefinido de papéis. Neste conjunto, os papéis são descritos por meio de processos de interação (PI). Para isto, desenvolveu-se a noção de PI. Para isto, desenvolveu-se a noção de PI. Foram desenvolvidas duas formas de proceder a classificação: construir uma especificação do agente a partir da observação das suas ações e verificar se esta pertence ao conjunto pré-definido de papéis; e, verificar se o comportamento do agente confere com as execuções possíveis para algum dos papéis pré-definidos. Este mecanismo é mais adequado para sociedades abertas e tem boa precisão no resultado apresentado, porém, a identificação do papel de um agente pode ser demorada. iii) Identificação de papéis por reconhecimento de intenções em planos: este mecanismo baseia-se na existência de uma relação entre intenções e papéis. A partir das ações observadas para o agente, procura-se saber qual seu plano, sua intenção e, conseqüentemente, seu papel. Para isto foi implementado um procedimento de indução de planos. Este mecanismo também é adequado para sociedades abertas, no entanto, a identificação, embora satisfatória, nem sempre é completamente correta. Estes três mecanismos foram testados em simulações numa implementação da sociedade Produtor - Consumidor, onde puderam ser comprovadas as características de cada um. / Distributed Artificial Intelligence brings a number of new perspectives to Computing Science when heterogeneous, adaptative and evolutive systems, those under functioning and open, are taken into consideration. These systems, named societies, present these characteristics because they allow their components, named agents, to migrate within societies, that is, agents are allowed to enter and to leave societies. Agents' migration brings two kinds of problems to the migrating agent: language and interaction problems both related to the use of used expressions and to the way the interactions are organized in the new society; and, knowledge and performance problems refering to the way the agent will behave in order to accomplish exactly what society expects him to do. This work is limited to knowledge and performance problems. In order to cooperate and coordinate their actions, the society's agents need to learn about the capabilities, abilities wishes and plans of other agents. A great part of knowledge of others can be extracted from the social roles these agents can play. Therefore, the problem posed in this work is how social agents who has received an immigrating agent and the immigrating agent himself will know and learn one another's roles. Three role identification mechanisms, and the comparison between them and their adaptation to migration types as well are developed. The three mechanisms are the following: i) Role Identification by means of presentation protocol:a language of protocol description (LPD) and a specification of presentation protocol in this LPD are proposed. The agents who use this mechanism can rapidly identify each other, however they need know a number of 'local' social information, which can be very restrictive to the migrating agent. ii) Role identification by means of observation and classification: this solution tries to classify the observed agent as a role out of set of definite roles. In this set, the roles are described by means of interactional processes (IP). Therefore, the notion of IP was developed. Two ways to proceed the role classification were developed: to build the agent's especification departing from the observation of their actions and to check whether this especification belongs to a set of pre-defined roles; and to check whether the agent's behavior fits the possible executions to some predefined roles. This mechanism is more adequate to open societies and has good precision in the result presented, but, the agent's role identification can last longer. iii) Role Identification by means of intention and plans recognition: this mechanism is based on the existence of a relationship between intentions and roles. By departing from the agent's observed actions, his plan, intention, consequently, his role is recognized. Therefore an induced plan procedure was implemented. This mechanism is also adequate to open societies, however, the identification, though satisfactory, is not always totally correct. These three mechanisms were tested in simulated situations in a kind of Producer- Consumer Society implementation in which each one's characteristics could be verified.

A Inteligência Artificial Distribuída traz uma série de novas perspectivas para a computação quando considera sistemas heterogêneos, adaptativos, evolutivos, continuamente em funcionamento e abertos. Estes sistemas, chamados de sociedades, apresentam tais características por permitirem que seus componentes, chamados de agentes, migrem entre sociedades, isto é, agentes podem sair e entrar em sociedades. Sociedades abertas permitem a migração dos agentes e coloca dois tipos de problemas para o agente que está migrando: problemas de linguagem e interação, que concernem ao uso de expressões usadas e à maneira como as interações são organizadas na nova sociedade; e, problemas de conhecimento e atuação, que se referem à como um agente irá se comportar a fim de realizar justamente aquilo que a sociedade espera dele. Este trabalho se atem aos problemas de conhecimento e atuação. Para que os agentes da sociedade possam cooperar e coordenar suas ações, é necessário que tenham conhecimento das capacidades, habilidades, desejos e planos dos outros agentes. Grande parte do conhecimento a respeito dos outros pode ser extraído dos papéis que estes podem assumir na sociedade. Assim sendo, o problema colocado para este trabalho é como os agentes da sociedade que receberam o agente imigrante e o próprio agente imigrante conhecerão/aprenderão os papéis uns dos outros. São desenvolvidos três mecanismos de identificação de papéis, bem como a comparação entre eles e sua adequação a tipos de migração. Os três mecanismos são os seguintes: i) Identificação de papéis por protocolo de apresentação: é proposta uma linguagem de descrição de protocolos (LDP) e uma especificação de protocolo de apresentação nesta LDP. Os agentes que utilizam este mecanismo conseguem se identificar com rapidez, porém necessitam conhecer várias informações “locais” da sociedade, o que pode ser muito restritivo para um agente migrante. ii) Identificação de papéis por observação e classificação: esta solução procura classificar o agente observado em um papel de um conjunto prédefinido de papéis. Neste conjunto, os papéis são descritos por meio de processos de interação (PI). Para isto, desenvolveu-se a noção de PI. Para isto, desenvolveu-se a noção de PI. Foram desenvolvidas duas formas de proceder a classificação: construir uma especificação do agente a partir da observação das suas ações e verificar se esta pertence ao conjunto pré-definido de papéis; e, verificar se o comportamento do agente confere com as execuções possíveis para algum dos papéis pré-definidos. Este mecanismo é mais adequado para sociedades abertas e tem boa precisão no resultado apresentado, porém, a identificação do papel de um agente pode ser demorada. iii) Identificação de papéis por reconhecimento de intenções em planos: este mecanismo baseia-se na existência de uma relação entre intenções e papéis. A partir das ações observadas para o agente, procura-se saber qual seu plano, sua intenção e, conseqüentemente, seu papel. Para isto foi implementado um procedimento de indução de planos. Este mecanismo também é adequado para sociedades abertas, no entanto, a identificação, embora satisfatória, nem sempre é completamente correta. Estes três mecanismos foram testados em simulações numa implementação da sociedade Produtor - Consumidor, onde puderam ser comprovadas as características de cada um. / Distributed Artificial Intelligence brings a number of new perspectives to Computing Science when heterogeneous, adaptative and evolutive systems, those under functioning and open, are taken into consideration. These systems, named societies, present these characteristics because they allow their components, named agents, to migrate within societies, that is, agents are allowed to enter and to leave societies. Agents' migration brings two kinds of problems to the migrating agent: language and interaction problems both related to the use of used expressions and to the way the interactions are organized in the new society; and, knowledge and performance problems refering to the way the agent will behave in order to accomplish exactly what society expects him to do. This work is limited to knowledge and performance problems. In order to cooperate and coordinate their actions, the society's agents need to learn about the capabilities, abilities wishes and plans of other agents. A great part of knowledge of others can be extracted from the social roles these agents can play. Therefore, the problem posed in this work is how social agents who has received an immigrating agent and the immigrating agent himself will know and learn one another's roles. Three role identification mechanisms, and the comparison between them and their adaptation to migration types as well are developed. The three mechanisms are the following: i) Role Identification by means of presentation protocol:a language of protocol description (LPD) and a specification of presentation protocol in this LPD are proposed. The agents who use this mechanism can rapidly identify each other, however they need know a number of 'local' social information, which can be very restrictive to the migrating agent. ii) Role identification by means of observation and classification: this solution tries to classify the observed agent as a role out of set of definite roles. In this set, the roles are described by means of interactional processes (IP). Therefore, the notion of IP was developed. Two ways to proceed the role classification were developed: to build the agent's especification departing from the observation of their actions and to check whether this especification belongs to a set of pre-defined roles; and to check whether the agent's behavior fits the possible executions to some predefined roles. This mechanism is more adequate to open societies and has good precision in the result presented, but, the agent's role identification can last longer. iii) Role Identification by means of intention and plans recognition: this mechanism is based on the existence of a relationship between intentions and roles. By departing from the agent's observed actions, his plan, intention, consequently, his role is recognized. Therefore an induced plan procedure was implemented. This mechanism is also adequate to open societies, however, the identification, though satisfactory, is not always totally correct. These three mechanisms were tested in simulated situations in a kind of Producer- Consumer Society implementation in which each one's characteristics could be verified.

La prise de décision collective est un processus dans lequel un groupe d'individus, ayant des intérêts différents, se réunit pour trouver une solution collective à un problème. Ce processus est inhérent aux activités de toute organisation politique, économique ou sociale. Le développement de l'Intelligence Artificielle notamment les Systèmes Multi-Agents a permis la modélisation et l'automatisation des processus de prise de décision afin de mieux comprendre et d'analyser leur fonctionnement. Une décision collective peut être prise par vote ou par négociation. Dans le cadre de cette thèse, nous abordons les mécanismes de négociation multilatérale pour la prise de décision collective basés sur des approches heuristiques. Les agents construisent la solution à leur problème à travers leurs interactions à la différence des modèles basés sur la théorie des jeux dont l'espace des solutions est supposé connu par tous les agents. Le problème des négociations heuristiques réside dans les mécanismes de raisonnement des agents dont la complexité augmente lorsque le nombre d'agents et d'attributs à négocier devient important. L'objectif de cette thèse est ainsi de proposer des mécanismes de négociation décentralisés (sans médiateur) et distribués en mettant en exergue l'aspect organisationnel des agents. Notre approche s'inspire du concept diviser pour régner et permet aux agents de négocier de façon incrémentale. Le but est de faciliter la recherche d'accords et de limiter la complexité du raisonnement des agents. Les travaux de cette thèse ont abouti à trois contributions abordant la négociation multi-agents sous différents angles tels que l'organisation des agents, le protocole d'interaction et les stratégies de concession et de choix de solutions équitables et justes. Pour valider nos propositions, nous avons implémenté sous JavaJade les mécanismes de négociation proposés. Les critères de performance que nous avons évalués sont, notamment, la convergence, le temps de négociation et la qualité de la solution. Nous avons comparé nos modèles avec ceux existants et les résultats obtenus montrent leur efficacité pour l'obtention des accords entre les agents / Collective decision making is a process in which many participants with different interests interact in order to build a solution to their problem. It is inherent to many organisations and companies. Nowadays, the advances in Artificial Intelligence, notably, Multi-Agents Systems enabled the automation of decision-making processes in order to analyse and to better understand how these mechanisms work. A collective decision may be made by using a voting system or by using negotiation. In this thesis, we focus on multilateral negotiation for collective decision making by proposing negotiation models. The proposed models based on heuristic approach. The agents interact with them in order to build a solution to their problem. This context is different from models based on game theory where the set of possible solutions are supposed to be known by all agents. So heuristic negotiation issue is that agents' reasoning may be very complex. This complexity grows where the number of agents and issues to be negotiated are important. The goal of this research work consists of devising negotiation mechanisms where agents'interaction are fully decentralized. We focus on organisation aspect of the multi-agent system by using divide and conquer approach in order to reduce the negotiation complexity and hence to facilitate research of agreements. Our works tackle negotiation under different contexts which lead us to bring three contributions which focus on agents' organization, interaction protocols, negotiation object, concession strategies and effective and fair solution concept. The proposed mechanisms are implemented in JavaJade. We analyse the convergence of the negotiation, negotiation time and quality of the solution. Our models are compared with a centralized approach where all of the agents are gathered around one group to negotiate. Our empirical analyses show that our propositions allow the agents to reach collectives agreements

Si la question de l'origine du langage reste d'un abord compliqué, celle de l'origine des formes du langage semble plus susceptible de se confronter à la démarche expérimentale. Malgré leur infinie variété, d'évidentes régularités y sont présentes~: les universaux du langage. Nous les étudions par des raisonnements plus généraux sur l'émergence du langage, notamment sur la recherche de précurseurs onto- et phylogénétiques. Nous abordons trois thèmes principaux~: la situation de communication parlée, les architectures cognitives des agents et l'émergence des universaux du langage dans des sociétés d'agents. Notre première contribution est un modèle conceptuel des agents communicants en interaction, issu de notre analyse bibliographique. Nous en proposons ensuite une formalisation mathématique Bayésienne~: le modèle d'un agent est une distribution de probabilités, et la production et la perception sont des inférences bayésiennes. Cela permet la comparaison formelle des différents courants théoriques en perception et en production de la parole. Enfin, nos simulations informatiques de société d'agents identifient les conditions qui favorisent l'apparition des universaux du langage. / If the origin of language is difficult to properly study, the origin of its forms appears to be accessible to the experimental method. Languages, despite their large variety, display obvious regularities, the linguistic universals. We study them through more general reasoning about language emergence, in particular in the search of its precursors, both in ontogeny and phylogeny. We study three main themes: the communication situation, the agent's cognitive architectures and the emergence of linguistic universals in agent societies. Our first contribution is a conceptual model of communicating agents in interaction, emanating from our bibliographic survey. We then cast it into the Bayesian mathematical formalism: an agent model is a probability distribution, and production and perception are defined by Bayesian inference. This allows a formal comparison of speech perception and production theoretical trends. Finally, computer simulations of agent societies help identify the conditions that favor the appearance of linguistic universals.

Le développement futur des performances sportives est un sujet de mythe et de désaccord entre les experts. Un article, publié en 2004, a donné lieu à un vif débat dans le domaine universitaire [1]. Il suggère que les modèles linéaires peuvent être utilisés pour prédire -sur le long terme- la performance humaine dans les courses de sprint. Des arguments en faveur et en défaveur de cette méthodologie ont été avancés par différent scientifiques et d’autres travaux ont montré que le développement des performances est non linéaire au cours du siècle passé [2, 3]. Une autre étude a également souligné que la performance est liée au contexte économique et géopolitique [4]. Dans ce travail, nous avons étudié les frontières suivantes : le développement temporel des performances dans des disciplines Olympiques et non Olympiques, avec le vieillissement chez les humains et d’autres espèces (lévriers, pur sangs, souris). Nous avons également étudié le développement des performances d’un point de vue plus large en analysant la relation entre performance, durée devie et consommation d’énergie primaire. Nous montrons que tous ces dévelopments sont limitées dans le temps [5, 6, 7] et que les modèles linéaires introduits précédemment sont de mauvais prédicteurs des phénomènes biologiques et physiologiques étudiés. Trois facteurs principaux et directs de la performance sportive sont l’âge [8, 9], la technologie [10, 11] et les conditions climatiques (température) [12]. Cependant, toutes les évolutions observées sont liées au contexte international et à l’utilisation des énergies primaires, ce dernier étant un paramètre indirect du développement de la performance. Nous montrons que lorsque les indicateurs des performances physiologiques et sociétales -tels que la durée de vie et la densité de population- dépendent des énergies primaires, la source d’énergie, la compétitioninter-individuelle et la mobilité sont des paramètres favorisant la réalisation de trajectoires durables sur le long terme. Dans le cas contraire, la grande majorité (98,7%) des trajectoires étudiées atteint une densité de population égale à 0 avant 15 générations, en raison de la dégradation des conditions environnementales et un faible taux de mobilité. Ceci nous a conduit à considérer que, dans le contexte économique turbulent actuel et compte tenu de la crise énergétique à venir, les performances sociétales et physiques ne devraient pas croître continuellement. / The development of sport performances in the future is a subject of myth and disagreement among experts. In particular, an article in 2004 [1] gave rise to a lively debate in the academic field. It stated that linear models can be used to predict human performance in sprint races in a far future. As arguments favoring and opposing such methodology were discussed, other publications empirically showed that the past development of performances followed a non linear trend [2, 3]. Other works, while deeply exploring the conditions leading to world records, highlighted that performance is tied to the economical and geopolitical context [4]. Here we investigated the following human boundaries : development of performances withtime in Olympic and non-Olympic events, development of sport performances with aging among humans and others species (greyhounds, thoroughbreds, mice). Development of performances from a broader point of view (demography & lifespan) in a specific sub-system centered on primary energy were also investigated. We show that all these developments are limited with time [5, 6, 7] and that previously introduced linear models are poor predictors of biological and physiological phenomena. Three major and direct determinants of sport performance are age [8, 9], technology [10, 11] and climatic conditions (temperature) [12]. However, all observed developments are related to the international context including the efficient use of primary energies. This last parameter is a major indirect propeller of performance development. We show that when physiological and societal performance indicators such as lifespan and population density depend on primary energies, the energy source, competition and mobility are key parameters for achieving longterm sustainable trajectories. Otherwise, the vast majority (98.7%) of the studied trajectories reaches 0 before 15 generations, due to the consumption of fossil energy and a low mobility rate. This led us to consider that in the present turbulent economical context and given the upcoming energy crisis, societal and physical performances are not expected to grow continuously.

Avec la convergence de plusieurs tendances profondes du secteur énergétique, lesréseaux électriques intelligents (smart grids) émergent comme le paradigme principal pourla modernisation des réseaux électriques. Les smart grids doivent notamment permettred’intégrer de larges proportions d’énergie renouvelable intermittente, de stockage et devéhicules électriques, ainsi que donner aux consommateurs plus de contrôle sur leur consommationénergétique. L’atteinte de ces objectifs repose sur l’adoption de nombreusestechnologies, et en particulier des technologies de l’information et de la communication.Ces changements transforment les réseaux en des systèmes de plus en plus complexes,nécessitant des outils adaptés pour modéliser, contrôler et simuler leur comportement.Dans cette thèse, l’utilisation des systèmes multi-agents (SMA) permet une approchesystémique de la gestion de l’énergie, ainsi que la définition d’architectures et d’algorithmesbénéficiant des propriétés des SMA. Cette approche permet de prendre en compte lacomplexité d’un tel système cyber-physique, en intégrant de multiples aspects commele réseau en lui-même, les infrastructures de communication, les marchés ou encore lecomportement des utilisateurs. L’approche est mise en valeur à travers deux applications.Dans une première application, un système de gestion de l’énergie pour centrales àturbines à gaz est conçu avec l’objectif de minimiser les coûts de fonctionnement et lesémissions de gaz à effet de serre pour des profils de charge variables. Un modèle de turbineà gaz basé sur des données réelles est proposé et utilisé dans un simulateur spécifiquementdéveloppé. Une métaheuristique optimise dynamiquement le dispatching entre les turbinesen fonction de leurs caractéristiques propres. Les résultats montrent que le systèmeest capable d’atteindre ses objectifs initiaux. Les besoins en puissance de calcul et encommunication sont également évalués.Avec d’autres mesures de gestion de la demande, l’effacement diffus permet de réduiretemporairement la charge électrique, par exemple dans la cas d’une congestion du réseaude transport. Dans cette seconde application, un système d’effacement diffus est proposéet utilise les ressources disponibles chez les particuliers (véhicules électriques, climatisation,chauffe-eau) pour maintenir la demande sous une valeur limite. Des aggrégateursde capacité de réduction de charge servent d’interface entre les opérateurs du réseau etun marché de l’effacement. Un simulateur est également développé pour évaluer la performancedu système. Les résultats de simulations montrent que le système réussit àatteindre ses objectifs sans compromettre la stabilité du réseau de distribution en régimecontinu. / Due to the convergence of several profound trends in the energy sector, smart gridsare emerging as the main paradigm for the modernization of the electric grid. Smartgrids hold many promises, including the ability to integrate large shares of distributedand intermittent renewable energy sources, energy storage and electric vehicles, as wellas the promise to give consumers more control on their energy consumption. Such goalsare expected to be achieved through the use of multiple technologies, and especially ofinformation and communication technologies, supported by intelligent algorithms.These changes are transforming power grids into even more complex systems, thatrequire suitable tools to model, simulate and control their behaviors. In this dissertation,properties of multi-agent systems are used to enable a new systemic approach to energymanagement, and allow for agent-based architectures and algorithms to be defined. Thisnew approach helps tackle the complexity of a cyber-physical system such as the smart gridby enabling the simultaneous consideration of multiple aspects such as power systems, thecommunication infrastructure, energy markets, and consumer behaviors. The approach istested in two applications: a “smart” energy management system for a gas turbine powerplant, and a residential demand response system.An energy management system for gas turbine power plants is designed with the objectiveto minimize operational costs and emissions, in the smart power generation paradigm.A gas turbine model based on actual data is proposed, and used to run simulations witha simulator specifically developed for this problem. A metaheuristic achieves dynamicdispatch among gas turbines according to their individual characteristics. Results showthat the system is capable of operating the system properly while reducing costs and emissions.The computing and communication requirements of the system, resulting from theselected architecture, are also evaluated.With other demand-side management techniques, demand response enables reducingload during a given duration, for example in case of a congestion on the transmissionsystem. A demand response system is proposed and relies on the use of the assets ofresidential customers to curtail and shift local loads (hybrid electric vehicles, air conditioning,and water heaters) so that the total system load remains under a given threshold.Aggregators act as interfaces between grid operators and a demand response market. Asimulator is also developed to evaluate the performance of the proposed system. Resultsshow that the system manages to maintain the total load under a threshold by usingavailable resources, without compromising the steady-state stability of the distributionsystem.

Ces travaux abordent la problématique du contrôle endogène dans les Systèmes Multi-Agents (SMA) pour la résolution de problèmes complexes, que nous explorons grâce au problème de partage de ressources critiques. Les problèmes dits complexes que nous abordons se caractérisent par une explosion combinatoire du nombre de solutions avec la taille des problèmes, une forte dynamique des données du problème induite par un environnement ouvert dans lequel de nombreux événements peuvent avoir lieu, une grande complexité systémique issue des interdépendances entre les nombreuses variables du problèmes et enfin une décentralisation du processus de résolution imposée par une distribution physique et fonctionnelle des variables incompatible avec une vision centralisée du problème. Un parcours complet des espaces de recherche associés à de tels problèmes est irréaliste en un temps acceptable, il est alors nécessaire d'employer des méthodes de résolution dite incomplètes. Quelque soit l'approche incomplète considérée, le parcours incomplet de l'espace de recherche requiert un contrôle afin de maximiser la probabilité de converger vers une solution satisfaisante. Nous identifions trois niveaux de contrôle du parcours de l'espace de recherche indépendamment de l'approche utilisée : un contrôle statique (définition a priori du comportement du système), un contrôle dynamique (évoluant en cours de résolution selon des mécanismes préétablis) et un contrôle adaptatif (évoluant dynamiquement en cours de résolution). Nous montrons qu'un contrôle endogène de l'activité du système, c.-à-d. un contrôle adaptatif issu de l'activité même des agents, est nécessaire au guidage du parcours de l'espace de recherche dans le contexte de résolution de problèmes complexes. Ces travaux ayant été réalisés dans un contexte de collaboration industrielle, ils s'appuient sur une approche développée au cours de précédents travaux : CESNA (Complex Exchanges between Stigmergic Negotiating Agents). CESNA est une approche multi-agents auto-organisationnelle exploitant des agents situés dans un environnement matérialisant le problème et exploité par un processus de résolution basé sur une négociation stigmergique entre les agents. Le cas applicatif utilisé par l'approche CESNA et permettant d'illustrer ces travaux est le problème de partage de ressources critiques, caractérisé par un ensemble restreint de ressources exploitées par un grand nombre de consommateurs. Nos contributions sont de deux types : nous avons dans un premier temps proposé des évolutions de la représentation du problème exploitée par l'approche initiale (CESNA) afin d'en supprimer les limitations interdisant un passage à l'échelle, et dans un second temps, nous avons défini un nouveau modèle (MANA : Multi-level bAlancing Negotiating Agents) exploitant cette nouvelle représentation à l'aide d'un nouveau processus de résolution basé sur des mécanismes endogènes de contrôle de l'activité du système. Ces mécanismes reposent sur la matérialisation des effets microscopiques du phénomène macroscopique à orienter (le parcours de l'espace de recherche) afin de le rendre perceptible localement par les agents. Nos mesures montrent que ce nouveau modèle permet le passage à l'échelle (la résolution de problèmes industriels de grande taille) et une amélioration significative des performances de résolution par rapport à l'approche initiale montrant ainsi l'efficacité du guidage permis par les mécanismes utilisés. / This work addresses the issue of the endogenous control of Multi-Agents Systems (MAS) for solving complex problems, which we explore through the critical resources sharing problem. The complex problems we address are characterized by a combinatorial explosion of number of solutions with the size of the problems, a strong dynamic of the problem's data caused by an open environment in which many events can take place, a huge systemic complexity caused by the interdependencies between the many variables of the problem and a decentralization of the resolution process imposed by a physical and functional distribution of the variables incompatible with a centralized view of the problem. A complete course of the search space associated with such problems is unrealistic in an acceptable time, it is necessary to employ resolution methods known as incomplete. Whatever the incomplete approach considered, the incomplete course of the search space requires a control to maximize the probability of converging to a satisfactory solution. We identify three levels of control of the course of the search space regardless of the used approach : a static control (textit a priori definition of the behavior of the system), a dynamic control (evolving during the resolution according to pre-established mechanisms) and adaptive control (dynamically evolving during resolution). We show that an endogenous control of the system activity, ie. an adaptive control from the agents activity, is necessary to guide the course of the search space in the context of solving complex problems. This work was made in a context of industrial collaboration, they rely on an approach developed in previous work : CESNA (Complex Exchanges Between Stigmergic Negotiating Agents). CESNA is a multi-agent self-organizational approach using agents situated in an environment embodying the problem and used by a resolution process based on a stigmergic negotiation between agents. The application used by the CESNA approach allowing to illustrate this work is the critical resources sharing problem, characterized by a limited set of resources exploited by many consumers. Our contributions are of two kinds : we initially proposed changes in the representation of the problem used by the initial approach (CESNA) to remove restrictions prohibiting scalability, and in a second time we defined a new model (MANA : Multi-level balancing Negotiating Agents) using this new representation with a new resolution process based on endogenous control mechanisms of the system activity. These mechanisms are based on the materialization of the microscopic effects of the macroscopic phenomenon to direct (the path in the search space) to make it noticeable by agents. Our measurements show that this new model allows the scaling (the resolution of industrial problems) and a significant performances improvement of the resolution showing the effectiveness of the control allowed by the mechanisms used.

The development of sport performances in the future is a subject of myth and disagreement among experts. In particular, an article in 2004 [1] gave rise to a lively debate in the academic field. It stated that linear models can be used to predict human performance in sprint races in a far future. As arguments favoring and opposing such methodology were discussed, other publications empirically showed that the past development of performances followed a non linear trend [2, 3]. Other works, while deeply exploring the conditions leading to world records, highlighted that performance is tied to the economical and geopolitical context [4]. Here we investigated the following human boundaries : development of performances withtime in Olympic and non-Olympic events, development of sport performances with aging among humans and others species (greyhounds, thoroughbreds, mice). Development of performances from a broader point of view (demography & lifespan) in a specific sub-system centered on primary energy were also investigated. We show that all these developments are limited with time [5, 6, 7] and that previously introduced linear models are poor predictors of biological and physiological phenomena. Three major and direct determinants of sport performance are age [8, 9], technology [10, 11] and climatic conditions (temperature) [12]. However, all observed developments are related to the international context including the efficient use of primary energies. This last parameter is a major indirect propeller of performance development. We show that when physiological and societal performance indicators such as lifespan and population density depend on primary energies, the energy source, competition and mobility are key parameters for achieving longterm sustainable trajectories. Otherwise, the vast majority (98.7%) of the studied trajectories reaches 0 before 15 generations, due to the consumption of fossil energy and a low mobility rate. This led us to consider that in the present turbulent economical context and given the upcoming energy crisis, societal and physical performances are not expected to grow continuously.

Le travail que nous présentons dans cette thèse s'articule autour du problème de la formation de coalitions entre des agents égoïstes qui planifient leurs activités, dans les systèmes multi-agents (SMA). Nous avons proposé, dans un premier temps, un mécanisme qui se fonde sur l’analyse des actions des agents dans leurs plans et le raisonnement sur les plans des autres, grâce notamment au calcul d’un degré de croyance sur les actions. Nous nous sommes, par ailleurs, intéressés au problème de la formation de coalitions avec des contraintes dynamiques et des préférences que les agents révèlent et communiquent aux autres lors de leurs négociations. Enfin, nous avons affiné notre mécanisme de formation des coalitions en permettant une recherche des coalitions guidée par la construction d'un arbre de contraintes et d'un arbre de coalitions, qui sont ensuite exploré par le biais de l'algorithme Monte-Carlo. / The work we present, in this thesis, focuses on the coalition formation problem for self-interested agents which plan their activities in multi-agents systems. As a first step, we have proposed, a mechanism that is based on the analysis of the agents' actions in their plans and reasoning about the plans of others. Additionally, we have addressed the problem of coalition formation with dynamic constraints and preferences that agents reveal and communicate to others during their negotiations. Finally, we have refined our coalition formation mechanism allowing a guided search of the coalitions by building a tree of constraints and a tree of coalitions. Each tree is explored by means of the Monte-Carlo algorithm.

Cette thèse présente comment l’intégration de systèmes multi-agents au sein de systèmes embarqués peut permettre d’optimiser la gestion des tâches. Nous relevons un manque de flexibilité pour les systèmes embarqués et posons l’hypothèse d’une solution multi-agents permettant la prise en compte dynamique du contexte d’évolution du système. Les systèmes embarqués, intégrés à l’environnement utilisateur, sont contraints en termes d’espace physique et donc de ressources matérielles. Ces limites impliquent un besoin d’optimisation des ressources. Nous proposons d’expérimenter des algorithmes de négociation multi-agents pour déléguer des tâches entre les différentes ressources. Pour valider nos hypothèses, nous détaillons les caractéristiques des systèmes multi-agents, leurs comportements, leurs modèles,les plateformes sur lesquelles ils évoluent, leurs standards de communication et leurs algorithmes sociaux. Nous avons constaté que la majorité des travaux du domaine agent se concentraient sur d’autres problématiques. Par conséquent, nous avons proposé la formalisation de systèmes multiagents embarqués et d’une plateforme multi-agents adaptée. Nous avons ensuite expérimenté cette plateforme au sein de systèmes embarqués avec le cas d’étude du traitement d’images, notamment avec le calcul d’une interpolation de pixels. Nous avons mené des études de performances pour estimer le coût administratif d’une solution multiagents,puis avons considéré ces résultats au regard des gains de capacité de nos systèmes embarqués. Nos dernières expérimentations mettent à l’épreuve notre solution de délégation de tâches entre plusieurs cartes embarquées dans un contexte hétérogène. / This thesis shows how the integration of multi-agents systems within embedded systems can optimize tasks management. We notice a lack of flexibility for embedded systems and hypothesize that a multi-agents solution will allow the dynamic consideration of the system context of evolution. Embedded systems, being integrated into the user environment, are limited in terms of physical space and thus hardware resources. These limits involve the necessity to optimize the resources. We suggest experimenting multi-agents negotiation algorithms to delegate tasks between several resources. To validate our hypotheses, we detail the characteristics of multi-agents systems, their behavior, their models, the platforms on which they evolve, their communication standards and their social algorithms.We observed that the majority of the works of the multi-agents domain concentrated on other problems. Therefore, we proposed the formalization of embedded multi-agents systems and of anadapted multi-agents platform. We then experimented this platform within embedded systems with the case study of image processing, especially the calculation of a pixels interpolation.We led performance studies to estimate the administrative cost of a multi-agents solution, then considered these results in relation to the capacity earnings of our embedded systems. Our last experiments put to the test our solution of tasks delegation between several embedded cards within a heterogeneous context.