Uma importante parte de um sistema multiagente é o seu mecanismo de coordenação que permite que os agentes possam agir de maneira coesa em direção aos seus objetivos, sejam eles individuais ou coletivos. Um agente pode optar por cooperar para atingir um determinado objetivo que seria inalcançável através de ações individuais, para realizar uma tarefa de maneira mais eficiente ou simplesmente porque ele foi projetado para tal. Em todos os casos, a formação de coalizões (grupos de agentes que concordam em coordenar suas ações em torno de um objetivo comum) é uma questão fundamental. O problema de geração de estruturas de coalizão entre agentes (conjunto de todas as combinações de coalizões) é um tópico de pesquisa que recebeu muita atenção principalmente na resolução do problema quando considerado como um jogo de função característica, onde o valor das coalizões independe dos agentes que não estão presentes nela. Essa abordagem, apesar de ser indicada para muitos tipos de problema, não cobre toda a área de pesquisa do assunto, visto que em muitos casos a criação de uma coalizão irá afetar os demais agentes do sistema. Quando o sistema possui agentes com objetivos sobrepostos ou contrários, uma coalizão cujos recursos são destinados a completar tais objetivos irá influenciar as demais coalizões desse sistema. Essa influência se chama externalidade e, nesses casos, o problema de formação de estruturas de coalizão deve ser tratado como um jogo de partição. Apesar das pesquisas na área de jogos de partição serem recentes, elas trazem resultados promissores e há alguns poucos algoritmos já desenvolvidos para buscar soluções a esse problema. A busca pela melhor estrutura de coalizão geralmente demanda que seja calculado o valor de todas possíveis coalizões, a fim de se encontrar aquele conjunto cuja soma dos valores das coalizões forneça o melhor resultado. Esse processo requer um alto número de computações e de memória, devido à natureza exponencial do problema. Assim, ao invés de apenas um agente central realizar todas as operações, é mais eficiente do ponto de vista do uso de recursos computacionais distribuir essas operações entres os diversos agentes presentes no sistema. Além dos benefícios computacionais, distribuir o processo de busca pela melhor estrutura de coalizão permitiria trabalhar com questões como privacidade e tolerância a falhas, tendo em vista que as informações não estão concentradas em um único agente. Apesar disso, não há na literatura qualquer algoritmo capaz de solucionar o problema de geração de estrutura de coalizão em ambientes distribuídos e que sejam modelados como jogos de partição. A proposta desse trabalho é utilizar a fundamentação teórica existente acerca do problema de formação de estruturas de coalizão (modelados tanto como jogos de função característica quanto como jogos de partição) para criar um algoritmo distribuído capaz de encontrar a estrutura de coalizão ótima em ambientes que possuam externalidade. Esse algoritmo utiliza como base a ordenação das coalizões e dos agentes para permitir a distribuição do cálculo dos limites superiores e inferiores de cada coalizão. Após, esses valores são utilizados para se encontrar o subespaço mais provável de conter a estrutura de coalizão ótima. Com base nos experimentos, percebe-se que o algoritmo encontrou a estrutura de coalizão ótima buscando em apenas uma pequena parte do espaço de busca. Para os experimentos com 16 agentes, o algoritmo foi capaz de encontrar a solução ótima procurando em apenas 0,01% do espaço de busca. Também, é demonstrado que em cenários com externalidade negativa os agentes necessitaram investigar um espaço de busca menor para encontrar a estrutura de coalizão ótima que em cenários com externalidade positiva. Experimentos também demonstram que o algoritmo não consegue encontrar a estrutura de coalizão ótima quando há falhas na comunicação entre os agentes. / An important part of a multi-agent system is its coordination mechanism that allows the agents to act cohesively towards their goals, whether individual or collective. An agent can choose to cooperate to achieve a certain goal that would be unattainable through individual actions, to perform a task more efficiently or simply because it was designed to do so. In all cases, the formation of coalitions (group of agents that agree to coordinate their actions around a common goal) is a key issue. The problem of generating coalition structures between agents (set of all combinations of coalitions) is a research topic that has received much attention mostly on solving the problem when considered as a characteristic function game, where the value of coalitions is independent of agents that are not part of it. This approach, although suitable for many types of problem, does not cover the whole area of research on the subject, since in many cases the creation of a coalition will affect the other agents of the system. When the system has agents with overlapping goals or opposing goals, a coalition whose resources are devoted to completing these objectives will influence the other coalitions of that system. This influence is called externality, and in these cases, the problem of formation of coalition structures should be treated as a partition function game. Although research in the area of partition games is recent, it brings promising results and there are few algorithms already developed to find solutions to this problem. The search for the best coalition structure generally requires computation of the value of all possible coalitions in order to find the set that the sum of the values of the coalitions provides the best result. This process requires a large number of computations and memory due to the exponential nature of the problem. Hence, instead of just one central agent performing all operations, it is more efficient to distribute those operations among several agents. Besides the computational benefits, distributing the search process for the best coalition structure would address issues such as privacy and fault tolerance, given that the information is not concentrated in a single agent. Nevertheless, in the literature there is not algorithm capable of solving the problem of coalition structure generation in decentralized environments and modeled as partition function game. The purpose of this work is to use the existing theoretical foundations for solving the coalition structure generation problem (modeled both as a characteristic function game and as a partition function game) to create a distributed algorithm capable of finding the optimal coalition structure in environments that have externality. This algorithm uses as a base the ordering of coalitions and agents to distribute the calculation of the upper and lower limits for each coalition. Afterwards, these values are used to find the subspace more likely to contain the optimal coalition structure. Based on experiments, the algorithm found the optimal coalition structure searching only a small part of the search space. For the experiments with 16 agents, the algorithm was able to find the solution looking at just 0.0001%of the search space. Also, it is shown that in scenarios with negative externality agents need to investigate a smaller search space to find the optimal coalition structure than in scenarios with positive externality. Experiments also show that the algorithm can not find the optimal coalition structure when there are failures in the communication among the agents.

Doctor of Philosophy / Electrical and Computer Engineering / Balasubramaniam Natarajan / A cyber-physical system (CPS) typically consists of networked computational elements that control physical processes. As an integral part of CPS, the widespread deployment of communicable sensors makes the task of monitoring and control quite challenging especially from the viewpoint of scalability and complexity. This research investigates two unique aspects of overcoming such barriers, making a CPS more robust against data explosion and network vulnerabilities. First, the correlated characteristics of high-resolution sensor data are exploited to significantly reduce the fused data volume. Specifically, spatial, temporal and spatiotemporal compressed sensing approaches are applied to sample the measurements in compressed form. Such aggregation can directly be used in centralized static state estimation even for a nonlinear system. This approach results in a remarkable reduction in communication overhead as well as memory/storage requirement. Secondly, an agent based architecture is proposed, where the communicable sensors (identified as agents) also perform local information processing. Based on the local and underdetermined observation space, each agent can monitor only a specific subset of global CPS states, necessitating neighborhood information exchange. In this framework, we propose an agent based static state estimation encompassing local consensus and least square solution. Necessary bounds for the consensus weights are obtained through the maximum eigenvalue based convergence analysis and are verified for a radial power distribution network. The agent based formulation is also applied for a linear dynamical system and the consensus approach is found to exhibit better and more robust performance compared to a diffusion filter. The agent based Kalman consensus filter (AKCF) is further investigated, when the agents can choose between measurements and/or consensus, allowing the economic allocation of sensing and communication tasks as well as the temporary omission of faulty agents. The filter stability is guaranteed by deriving necessary consensus bounds through Lyapunov stability analysis. The states dynamically estimated from AKCF can be used for state-feedback control in a model predictive fashion. The effect of lossy communication is investigated and critical bounds on the link failure rate and the degree of consensus that ensure stability of the agent based control are derived and verified via simulations.

Uma importante parte de um sistema multiagente é o seu mecanismo de coordenação que permite que os agentes possam agir de maneira coesa em direção aos seus objetivos, sejam eles individuais ou coletivos. Um agente pode optar por cooperar para atingir um determinado objetivo que seria inalcançável através de ações individuais, para realizar uma tarefa de maneira mais eficiente ou simplesmente porque ele foi projetado para tal. Em todos os casos, a formação de coalizões (grupos de agentes que concordam em coordenar suas ações em torno de um objetivo comum) é uma questão fundamental. O problema de geração de estruturas de coalizão entre agentes (conjunto de todas as combinações de coalizões) é um tópico de pesquisa que recebeu muita atenção principalmente na resolução do problema quando considerado como um jogo de função característica, onde o valor das coalizões independe dos agentes que não estão presentes nela. Essa abordagem, apesar de ser indicada para muitos tipos de problema, não cobre toda a área de pesquisa do assunto, visto que em muitos casos a criação de uma coalizão irá afetar os demais agentes do sistema. Quando o sistema possui agentes com objetivos sobrepostos ou contrários, uma coalizão cujos recursos são destinados a completar tais objetivos irá influenciar as demais coalizões desse sistema. Essa influência se chama externalidade e, nesses casos, o problema de formação de estruturas de coalizão deve ser tratado como um jogo de partição. Apesar das pesquisas na área de jogos de partição serem recentes, elas trazem resultados promissores e há alguns poucos algoritmos já desenvolvidos para buscar soluções a esse problema. A busca pela melhor estrutura de coalizão geralmente demanda que seja calculado o valor de todas possíveis coalizões, a fim de se encontrar aquele conjunto cuja soma dos valores das coalizões forneça o melhor resultado. Esse processo requer um alto número de computações e de memória, devido à natureza exponencial do problema. Assim, ao invés de apenas um agente central realizar todas as operações, é mais eficiente do ponto de vista do uso de recursos computacionais distribuir essas operações entres os diversos agentes presentes no sistema. Além dos benefícios computacionais, distribuir o processo de busca pela melhor estrutura de coalizão permitiria trabalhar com questões como privacidade e tolerância a falhas, tendo em vista que as informações não estão concentradas em um único agente. Apesar disso, não há na literatura qualquer algoritmo capaz de solucionar o problema de geração de estrutura de coalizão em ambientes distribuídos e que sejam modelados como jogos de partição. A proposta desse trabalho é utilizar a fundamentação teórica existente acerca do problema de formação de estruturas de coalizão (modelados tanto como jogos de função característica quanto como jogos de partição) para criar um algoritmo distribuído capaz de encontrar a estrutura de coalizão ótima em ambientes que possuam externalidade. Esse algoritmo utiliza como base a ordenação das coalizões e dos agentes para permitir a distribuição do cálculo dos limites superiores e inferiores de cada coalizão. Após, esses valores são utilizados para se encontrar o subespaço mais provável de conter a estrutura de coalizão ótima. Com base nos experimentos, percebe-se que o algoritmo encontrou a estrutura de coalizão ótima buscando em apenas uma pequena parte do espaço de busca. Para os experimentos com 16 agentes, o algoritmo foi capaz de encontrar a solução ótima procurando em apenas 0,01% do espaço de busca. Também, é demonstrado que em cenários com externalidade negativa os agentes necessitaram investigar um espaço de busca menor para encontrar a estrutura de coalizão ótima que em cenários com externalidade positiva. Experimentos também demonstram que o algoritmo não consegue encontrar a estrutura de coalizão ótima quando há falhas na comunicação entre os agentes. / An important part of a multi-agent system is its coordination mechanism that allows the agents to act cohesively towards their goals, whether individual or collective. An agent can choose to cooperate to achieve a certain goal that would be unattainable through individual actions, to perform a task more efficiently or simply because it was designed to do so. In all cases, the formation of coalitions (group of agents that agree to coordinate their actions around a common goal) is a key issue. The problem of generating coalition structures between agents (set of all combinations of coalitions) is a research topic that has received much attention mostly on solving the problem when considered as a characteristic function game, where the value of coalitions is independent of agents that are not part of it. This approach, although suitable for many types of problem, does not cover the whole area of research on the subject, since in many cases the creation of a coalition will affect the other agents of the system. When the system has agents with overlapping goals or opposing goals, a coalition whose resources are devoted to completing these objectives will influence the other coalitions of that system. This influence is called externality, and in these cases, the problem of formation of coalition structures should be treated as a partition function game. Although research in the area of partition games is recent, it brings promising results and there are few algorithms already developed to find solutions to this problem. The search for the best coalition structure generally requires computation of the value of all possible coalitions in order to find the set that the sum of the values of the coalitions provides the best result. This process requires a large number of computations and memory due to the exponential nature of the problem. Hence, instead of just one central agent performing all operations, it is more efficient to distribute those operations among several agents. Besides the computational benefits, distributing the search process for the best coalition structure would address issues such as privacy and fault tolerance, given that the information is not concentrated in a single agent. Nevertheless, in the literature there is not algorithm capable of solving the problem of coalition structure generation in decentralized environments and modeled as partition function game. The purpose of this work is to use the existing theoretical foundations for solving the coalition structure generation problem (modeled both as a characteristic function game and as a partition function game) to create a distributed algorithm capable of finding the optimal coalition structure in environments that have externality. This algorithm uses as a base the ordering of coalitions and agents to distribute the calculation of the upper and lower limits for each coalition. Afterwards, these values are used to find the subspace more likely to contain the optimal coalition structure. Based on experiments, the algorithm found the optimal coalition structure searching only a small part of the search space. For the experiments with 16 agents, the algorithm was able to find the solution looking at just 0.0001%of the search space. Also, it is shown that in scenarios with negative externality agents need to investigate a smaller search space to find the optimal coalition structure than in scenarios with positive externality. Experiments also show that the algorithm can not find the optimal coalition structure when there are failures in the communication among the agents.

Aujourd’hui, la compétition croissante, l'expansion des marchés et les progrès de la technologie engendrent un raccourcissement du cycle de vie des processus de développement des produits afin d’en améliorer les performances en termes de délai, coût et qualité. Ce raccourcissement du cycle de vie a engendré un accroissement des volumes de déchet généré et des conséquences que cela peut avoir sur l’environnement. Au niveau de l’Union européenne, des directives ont été introduites, tels que la gestion des déchets d'équipements électriques et électroniques (WEEE), la restriction de l'utilisation de certaines substances dangereuses dans les équipements électriques et électroniques (RoHS) et les directives pour le traitement des batteries usagées (Battery); ces directives permettent de limiter l'utilisation et le recyclage des substances dangereuses nocives pour la santé et pour l'environnement.Ces nouvelles réglementations et normes, permettant de gérer de manière efficace les retours et la fin de vie des produits (recovery process), ont été mises en place afin d'obliger les entreprises à assumer leurs responsabilités en termes de gestion des produits en fin de vie. Certaines entreprises ont montré que les produits recyclés ou réutilisés peuvent être une source supplémentaire de revenu (recyclage des matériaux, ou réutilisation des composants après démontage) dans le processus de fabrication.Ces connaissances liées à la performance environnementale (au niveau des processus de conception, de production, de transport, d’entreposage, de récupération…) devraient êtres saisies, évaluées et capitalisées dans des bases de connaissances afin d’être prisent en compte durant les différents phases du cycle de vie des produits.Nos travaux de recherche proposent donc de développer une architecture de gestion des connaissances (Knowledge Management Architecture) basée sur un Système Multi-Agents. L’objectif est de proposer un système qui met l'accent sur les concepts de « durabilité des produits et des cycles de vie », en établissant des liens entre des Agents Logiciels détenteurs de connaissances liées à la réglementation environnementale et les Systèmes d’Information de type PLM. Ces interconnexions permettront aux décideurs de prendre en compte les impacts environnementaux dans leurs décisions et ceci à chaque phase du cycle de vie des produits. / The increasing of competition, expanding markets and advanced technology create shorten lifecycle and the development process to improve product performance in terms of time, cost and quality. These shorten products lifecycle have led to increase volumes of waste generation and consequence impact to environment. EU directives have been introduced, such as Waste Electrical and Electronic Equipment (WEEE), Restriction of the use of certain Hazardous Substances in electrical and electronic equipment (RoHS) and guidelines for the treatment of waste batteries (Battery), these directives are used to handle the hazardous substances which are harmful to human health and the environment.These regulations and standards have been put in place to force companies take their responsibilities on managing their products when reach to the end of life. Some companies have found that the returned product can be recycled or reused as an additional source of income (material recycling, or reuse of components after disassembly) in the manufacturing process.Knowledge related to the environmental performance (in terms of process design, production, transportation, storage, etc.) should be captured, evaluated and stored in knowledge base in order to share between users in different phases of the product lifecycle.Therefore, this research proposes a Knowledge Management Architecture based on a Multi-Agent System approach. The objective of this work is to propose a system that focuses on the concept of "sustainability” of products lifecycle by establishing the link between agents, who hold knowledge related to the environmental performances, and PLM system. The connection encourages companies considering the environmental impacts in their decision making at every stage of product lifecycle.

Ces dernières années, les progrès en sciences de la gestion et de l’information ont transformé la Gestion de Processus d’Affaires (Business Process Management, BPM) en un sujet important, tant du côté de l’industrie que de celui de la recherche. Le BPM utilise des processus métiers pour améliorer la performance opérationnelle des organisations. Les processus métiers établissent un lien entre les personnes, les systèmes, et les différentes organisations, dans le but de créer de la valeur pour les parties prenantes. La cible de notre travail est la famille des systèmes BPM. Un système BPM est un système logiciel générique guidé par des modèles explicites de processus métier avec pour objectif d’exécuter et de gérer des processus opérationnels. Malgré le vaste éventail de sujets traités par ce domaine de recherche, il reste encore quelques questions qui méritent une étude plus approfondie. Un problème particulier concerne la médiation entre les systèmes BPM et les humains. L’interaction homme-machine dans ces systèmes repose sur des interfaces standard basées sur des listes de taches et des formulaires, ce qui est très contraignant pour les utilisateurs.Ceux-ci ont non seulement des difficultés à exécuter leurs processus métier, mais aussi a trouver le processus métier le mieux adapté à leurs besoins. Il serait beaucoup plus efficace d’utiliser des dialogues en langage naturel. Malheureusement les langages de modélisation de processus ne permettent pas de capturer ni de modéliser un domaine de discours. Le travail présent propose une approche originale de gestion du dialogue basée sur des systèmes multi-agents pour l’exécution des processus métier. La motivation globale pour ce travail fut de concevoir un modèle de dialogue extensible à différents domaines. Ce modèle s’appuie sur les ontologies de domaine, nécessitant un minimum d’effort d’adaptation pour améliorer l’interaction. Les résultats montrent tout le potentiel de notre approche multi-agent pour réaliser une médiation automatiquement, sans qu’il soit nécessaire de reconstruire les modèles de processus métier. / Over the last few years, the advances in management science and information technology have transformed the business process management (BPM) discipline into an important topic for both industry and academy. BPM uses business processes as the means for improving the operational performance of organizations, and setting processes are at the heart of BPM allows linking together people, systems, and different organizations to deliver value to stakeholders. The target of our work is the family of BPM systems. A BPM system is a generic software system that is driven by explicit process designs to enact and manage operational business processes. Despite the wide range of topics addressed by the academy on business processes, there are still aspects not addressed by prior research. A particular problem in this regard is the mediation between BPM systems and humans. Human interaction in those systems follows a standard user interface based predominantly on work item lists and forms. Thus, there is little room for creativity for users. They have not only difficulties in enacting their processes but also for searching the most suitable one for their needs. It would be more efficient to let humans interact in natural language. However, process modeling languages are an insufficient means of capturing and representing the domain of discourse. The present thesis develops an original approach to agent dialog management for the problem of business process enactment. The overarching motivation for this work was to design a dialog model scalable to different domains. The model relies on domain and business process ontologies, and necessitates a minimum effort of adaptation on ontologies to improve the interaction. Results indicate the potential of our agent-based approach to generate natural language.

Real-Time Optimization (RTO) is a family of techniques that pursue to improve the performance of chemical processes. As general scheme, the method reevaluates the process conditions in a frequent basis and tries to adjust some selected variables, taking into account the plant state, actual operational constraints and optimization objectives. Several RTO approaches have born from the academy research and industrial practices, at the same time that more applications have been implemented in real facilities. Between the main motivations to apply RTO are the dynamic of markets, the seek for quality in the process results and environmental sustainability. That is why the interest on deeply understand the phases and steps involved in an RTO application has increased in recent years. Nevertheless, the fact that most of the existing RTO systems have been developed by commercial organizations makes it difficult to meet that understanding. This work studies the nature of RTO systems from a software point of view. Software requirements for a generic system are identied. Based on that, a software architecture is proposed that could be adapted for specfic cases. Benefits of the designed architecture are listed. At the same time, the work proposes a new approach to implement that architecture as a Multi-Agent System (MAS). Two RTO system prototypes were developed then, one for a well-know academic case study and the other oriented to be used in a real unit. The benefits of the MAS approach and the architecture, for researching on the RTO field and implementation on real plants, are analyzed in the text. A sub-product of the development, a software framework covering main concepts from the RTO ontology, is proposed as well. As the framework was designed to be generic, it can be used in new applications development and extended to very specific scenarios. / Otimização em Tempo Real (OTR) é uma família de técnicas que buscam melhorar o desempenho dos processos químicos. Como esquema geral, o método reavalia frequentemente as condições do processo e tenta ajustar algumas variáveis selecionadas, levando em considera ção o estado da planta, restrições operacionais e os objetivos da otimização. Várias abordagens para OTR t^em surgido da pesquisa acadêmica e das práticas industriais, ao mesmo tempo em que mais aplicações têm sido implementadas em plantas reais. As principais motivações para aplicar OTR são: a dinâmica dos mercados, a busca de qualidade nos resultados dos processos e a sustentabilidade ambiental. É por isso que o interesse em entender as fases e etapas envolvidas em uma aplicação OTR cresceu nos últimos anos. No entanto, o fato de que a maioria dos sistemas OTR em operação foram desenvolvidos por organizações comerciais dificulta o caminho para chegar nesse entendimento. Este trabalho analisa a natureza dos sistemas OTR desde o ponto de vista do software. Os requerimentos para um sistema genérico são levantados. Baseado nisso, é proposta uma arquitetura de software que pode ser adaptada para casos específicos. Os benefícios da arquitetura projetada foram listados. Ao mesmo tempo, o trabalho propõe uma nova abordagem para implementar essa arquitetura: Sistema Multi-Agentes (SMA). Dois protótipos de sistema OTR foram desenvolvidos. O primeiro aplicado num estudo de caso bem conhecido na literatura acadêmica. O segundo voltado para ser usado em uma unidade industrial. Os benefícios da abordagem SMA e da arquitetura, tanto na pesquisa relacionada com OTR, quanto na implementação em plantas reais, são analisados no texto. Um arcabouço de software que abrange os principais conceitos da ontologia OTR é proposto como resultado derivado do desenvolvimento. O arcabouço foi projetado para ser genérico, possibilitando seu uso no desenvolvimento de novas aplicações OTR e sua extensão a cenários muito específicos.

Uma importante parte de um sistema multiagente é o seu mecanismo de coordenação que permite que os agentes possam agir de maneira coesa em direção aos seus objetivos, sejam eles individuais ou coletivos. Um agente pode optar por cooperar para atingir um determinado objetivo que seria inalcançável através de ações individuais, para realizar uma tarefa de maneira mais eficiente ou simplesmente porque ele foi projetado para tal. Em todos os casos, a formação de coalizões (grupos de agentes que concordam em coordenar suas ações em torno de um objetivo comum) é uma questão fundamental. O problema de geração de estruturas de coalizão entre agentes (conjunto de todas as combinações de coalizões) é um tópico de pesquisa que recebeu muita atenção principalmente na resolução do problema quando considerado como um jogo de função característica, onde o valor das coalizões independe dos agentes que não estão presentes nela. Essa abordagem, apesar de ser indicada para muitos tipos de problema, não cobre toda a área de pesquisa do assunto, visto que em muitos casos a criação de uma coalizão irá afetar os demais agentes do sistema. Quando o sistema possui agentes com objetivos sobrepostos ou contrários, uma coalizão cujos recursos são destinados a completar tais objetivos irá influenciar as demais coalizões desse sistema. Essa influência se chama externalidade e, nesses casos, o problema de formação de estruturas de coalizão deve ser tratado como um jogo de partição. Apesar das pesquisas na área de jogos de partição serem recentes, elas trazem resultados promissores e há alguns poucos algoritmos já desenvolvidos para buscar soluções a esse problema. A busca pela melhor estrutura de coalizão geralmente demanda que seja calculado o valor de todas possíveis coalizões, a fim de se encontrar aquele conjunto cuja soma dos valores das coalizões forneça o melhor resultado. Esse processo requer um alto número de computações e de memória, devido à natureza exponencial do problema. Assim, ao invés de apenas um agente central realizar todas as operações, é mais eficiente do ponto de vista do uso de recursos computacionais distribuir essas operações entres os diversos agentes presentes no sistema. Além dos benefícios computacionais, distribuir o processo de busca pela melhor estrutura de coalizão permitiria trabalhar com questões como privacidade e tolerância a falhas, tendo em vista que as informações não estão concentradas em um único agente. Apesar disso, não há na literatura qualquer algoritmo capaz de solucionar o problema de geração de estrutura de coalizão em ambientes distribuídos e que sejam modelados como jogos de partição. A proposta desse trabalho é utilizar a fundamentação teórica existente acerca do problema de formação de estruturas de coalizão (modelados tanto como jogos de função característica quanto como jogos de partição) para criar um algoritmo distribuído capaz de encontrar a estrutura de coalizão ótima em ambientes que possuam externalidade. Esse algoritmo utiliza como base a ordenação das coalizões e dos agentes para permitir a distribuição do cálculo dos limites superiores e inferiores de cada coalizão. Após, esses valores são utilizados para se encontrar o subespaço mais provável de conter a estrutura de coalizão ótima. Com base nos experimentos, percebe-se que o algoritmo encontrou a estrutura de coalizão ótima buscando em apenas uma pequena parte do espaço de busca. Para os experimentos com 16 agentes, o algoritmo foi capaz de encontrar a solução ótima procurando em apenas 0,01% do espaço de busca. Também, é demonstrado que em cenários com externalidade negativa os agentes necessitaram investigar um espaço de busca menor para encontrar a estrutura de coalizão ótima que em cenários com externalidade positiva. Experimentos também demonstram que o algoritmo não consegue encontrar a estrutura de coalizão ótima quando há falhas na comunicação entre os agentes. / An important part of a multi-agent system is its coordination mechanism that allows the agents to act cohesively towards their goals, whether individual or collective. An agent can choose to cooperate to achieve a certain goal that would be unattainable through individual actions, to perform a task more efficiently or simply because it was designed to do so. In all cases, the formation of coalitions (group of agents that agree to coordinate their actions around a common goal) is a key issue. The problem of generating coalition structures between agents (set of all combinations of coalitions) is a research topic that has received much attention mostly on solving the problem when considered as a characteristic function game, where the value of coalitions is independent of agents that are not part of it. This approach, although suitable for many types of problem, does not cover the whole area of research on the subject, since in many cases the creation of a coalition will affect the other agents of the system. When the system has agents with overlapping goals or opposing goals, a coalition whose resources are devoted to completing these objectives will influence the other coalitions of that system. This influence is called externality, and in these cases, the problem of formation of coalition structures should be treated as a partition function game. Although research in the area of partition games is recent, it brings promising results and there are few algorithms already developed to find solutions to this problem. The search for the best coalition structure generally requires computation of the value of all possible coalitions in order to find the set that the sum of the values of the coalitions provides the best result. This process requires a large number of computations and memory due to the exponential nature of the problem. Hence, instead of just one central agent performing all operations, it is more efficient to distribute those operations among several agents. Besides the computational benefits, distributing the search process for the best coalition structure would address issues such as privacy and fault tolerance, given that the information is not concentrated in a single agent. Nevertheless, in the literature there is not algorithm capable of solving the problem of coalition structure generation in decentralized environments and modeled as partition function game. The purpose of this work is to use the existing theoretical foundations for solving the coalition structure generation problem (modeled both as a characteristic function game and as a partition function game) to create a distributed algorithm capable of finding the optimal coalition structure in environments that have externality. This algorithm uses as a base the ordering of coalitions and agents to distribute the calculation of the upper and lower limits for each coalition. Afterwards, these values are used to find the subspace more likely to contain the optimal coalition structure. Based on experiments, the algorithm found the optimal coalition structure searching only a small part of the search space. For the experiments with 16 agents, the algorithm was able to find the solution looking at just 0.0001%of the search space. Also, it is shown that in scenarios with negative externality agents need to investigate a smaller search space to find the optimal coalition structure than in scenarios with positive externality. Experiments also show that the algorithm can not find the optimal coalition structure when there are failures in the communication among the agents.

Un certain nombre de travaux basés sur les réseaux de capteurs sans fil s'intéressent à la gestion de l'énergie de ces capteurs. Cette énergie est, de fait, un facteur critique dans le fonctionnement de ces réseaux. Une construction adéquate des clusters de capteurs est un très bon moyen pour minimiser la consommation de cette énergie. La problématique liée à ces réseaux réside ainsi souvent dans leur durée de vie mais aussi dans le nécessaire maintien de la connectivité entre tous les capteurs. Ces deux aspects sont étroitement liés. Dans cette thèse, nous nous sommes focalisés sur ces deux volets, dans le contexte de réseaux de capteurs statiques mais aussi celui de capteurs mobiles.Nous proposons, dans un premier temps, un algorithme hybride pour la mise en place des clusters et la gestions de ces clusters. L'originalité de cette solution réside dans la mise en place de zones géographiques de désignation des cluster heads mais aussi dans la transmission, dans les messages échangés, de la quantité d'énergie restante sur les capteurs. Ainsi, les données sur les capteurs permettront de désigner les cluster heads et leurs successeurs qui détermineront les seuils pour les autres capteurs et pour leur fonctionnement. L'algorithme est testé à travers de nombreuses simulations. La seconde partie du travail consiste à adapter notre premier algorithme pour les réseaux de capteurs mobiles. Nous in_uons sur la trajectoire des capteurs pour maintenir la connectivité et limiter la consommation d'énergie. Pour cela, nous nous inspirons de l'écho-localisation pratiquée par les chauvessouris. Nous nous sommes donc intéressés à la topologie changeante et dynamique dans les réseaux de capteurs. Nous avons analysé la perte d'énergie en fonction de la distance et de la puissance de transmission entre les n÷uds et le cluster head. Nous évaluons également notre algorithme sur des capteurs qui ont un déplacement aléatoire. Nous appliquons ces algorithmes à une simulation de _otte de drones de surveillance. / A number of works based on wireless sensor networks are interested in the energy management of these sensors. This energy is in fact a critical factor in the operation of these networks. Proper construction of sensor clusters is a great way to minimize the consumption of this energy. The problems related to these networks and often lies in their lifetime but also in the need to maintain connectivity between all transducers. These two aspects are closely linked. In this thesis, we focused on these two aspects in the context of static sensor networks but also of mobile sensors.We propose, as a _rst step, a hybrid algorithm for setting up clusters and the management of theseclusters. The uniqueness of this solution lies in the establishment of geographic areas for designation fcluster heads but also in transmission, in the exchanged messages, the amount of remaining energy on the sensors. Thus, the sensor data will designate the cluster heads and their successors will determine the thresholds for other sensors and for their operation. The algorithm is tested through many simulations. The second part of the work is to adapt our _rst algorithm for mobile sensor networks. We a_ect the trajectory of sensors to maintain connectivity and reduce energy consumption. For this, we are guided echo-location practiced by bats. We're interested in changing and dynamic topology in sensor networks. We analyzed the loss of energy as a function of the distance and the power transmission between the nodes and the head cluster. We also evaluate our algorithm on sensors that have a random move. We apply these algorithms to a _eet of surveillance drones simulation.

La complexité des systèmes de transport rend difficile la maîtrise de leur conception dans la mesure où ils intègrent des composantes technologiques, mais aussi sociologiques et politiques. Il est alors opportun de proposer un système destiné à aider à la définition d’une politique de transport urbain. L’objectif principal de notre recherche est de proposer l’architecture logicielle d’un outil de simulation visant à aider un décideur, chargé de la régulation d’un système de transport urbain, pour son travail d’analyse et d’évaluation des impacts des stratégies de régulation. Notre système est basé sur un simulateur à base d’agents intégrant des informations géographiques et temporelles pour évaluer des scénarii de régulation. En termes d’architecture du système, nous avons adopté une approche «système de systèmes», principalement structurée par couches, afin de modéliser les principaux éléments du système. La validation de notre outil de simulation a pu être effectuée à partir d’une étude de cas de taille et de complexité significative puisque nous disposons des enquêtes de déplacement, de recensement, et des mesures de trafic. Nous avons réalisé un prototype pour les déplacements des usagers dans la ville de La Rochelle à partir des données statistiques de l’INSEE et de la BD TOPO 2 de l’IGN en utilisant la plate-forme de simulation GAMA. / Transport systems are becoming more complex and must incorporate not only technological components, but also sociological and political ones. In particular, they should be easy to adapt in order to incorporate the goals set by decision makers, such as the integration of sustainable development settings. The main objective of our research is to propose software architecture of a simulation tool to help a decision maker, responsible for the regulation of an urban transportation system to analyze and evaluate the impacts of regulatory strategies. We propose a system to assist in the definition of an urban transportation policy. Our system is based on an agent-based simulation integrating spatial and temporal information to evaluate regulatory scenarios. In terms of system architecture, we adopted a “system of systems” approach, mainly structured in layers, in order to model the main elements of the system. The validation of our simulation tool could be done from a case study of significant size and complexity because we have travel surveys, census, and traffic measurements. We made a prototype for the movement of people in the city of La Rochelle from statistical data of INSEE and the BD TOPO 2 of IGN using the GAMA platform.

After the historic chess match between Deep Blue and Garry Kasparov, many researchers considered the game of chess solved and moved on to the more complex game of soccer. Artificial intelligence research has shifted focus to creating artificial players capable of mimicking the task of playing soccer. A new training algorithm is presented in this thesis for training teams of players from zero knowledge, evaluated on a simplified version of the game of soccer. The new algorithm makes use of the charged particle swarm optimiser as a neural network trainer in a coevolutionary training environment. To counter the lack of domain information a new relative fitness measure based on the FIFA league-ranking system was developed. The function provides a granular relative performance measure for competitive training. Gameplay strategies that resulted from the trained players are evaluated. It was found that the algorithm successfully trains teams of agents to play in a cooperative manner. Techniques developed in this study may also be widely applied to various other artificial intelligence fields. / Dissertation (MSc)--University of Pretoria, 2013. / Computer Science / unrestricted