L'objectif de cette thèse est de proposer une démarche méthodologique de conception d'un modèle de simulation de systèmes de production qui intègre une représentation des comportements distribués et cognitifs des acteurs décisionnels, et dont l'intérêt est de permettre d'étudier la pertinence de changements organisationnels dans ces systèmes. En effet, l'analyse des conditions d'évolution des systèmes de production et de leur contexte a permis de mettre en relief le besoin de " réactivité " pour faire face à un environnement de moins en moins stable et prévisible et devant lequel l'organisation doit chercher des alternatives à un fonctionnèment planifié, prédictif, figé et séquentiel. Les réponses organisationnelles des entreprises à ces sollicitations s'expriment par un mouvement vers la fleXIbilité et l'intégration, et passent par un rapprochement entre la structure décisionnelle et la structure physique. Pour mettre en évidence cette nouvelle réalité, nous proposons un processus de modélisation qui débute par la construction d'un modèle conceptuel de compréhension des organisations, basé sur une représentation des processus de décision, et sur une formalisation des phénomènes d'orientation des comportements décisionnels. En parallèle, nous analysons les concepts associés à la problématique des systèmes multi-agents, systèmes qui étudient l'émergence de comportements collectifs résultant de l'activité coopérative de plusieurs entités autonomes "intelligentes" sur lesquelles l'expertise est répartie. Ces études convergent vers un modèle de simulation issu des techniques de simulation à événements discrets, dans lequel nous intégrons une modélisation du système décisionnel à base d'agents. Enfin, l'implémentation de ce modèle sera basée sur des techniques de simulation orientée objets (objets actifs, programmation concurrente), et sera validée sur deux exemples, l'un pédagogique et conceptuel, 1 'autre industriel.

Cette thèse est une proposition pour la conception de comportements crédibles dans les simulations informatiques pour agents situés dans un environnement virtuel. Le comportement d'un agent se définit à partir de l'observation des actions qu'il exécute dans un environnement. Chaque action exécutée résulte d'un choix de l'agent parmi l'ensemble des actions qu'il peut effectuer. Le comportement se construit donc à partir des capacités de l'agent lui permettant de résoudre ses buts, le raisonnement, et d'un choix influencé par ses traits de caractère, l'individualité. Ma contribution consiste à définir des comportements qui sont composés de ces deux parties : le raisonnement et l'individualité. Cette thèse se concentre sur la proposition d'un mécanisme de sélection d'action pour la partie individualité du comportement. Ce mécanisme se base sur les notions de motivations et d'alternatives. Les motivations sont l'expression de traits du caractère de l'agent, elles influencent le choix de l'action à exécuter. Les alternatives sont les résolutions possibles calculées par la partie raisonnement du comportement. Le mécanisme de sélection d'action s'appuie sur les alternatives pour déterminer, à l'aide des motivations, la meilleure action à exécuter à chaque instant. Les motivations influençant le comportement de l'agent sont définies indépendamment du contexte d'application permettant leur réutilisation pour d'autres agents et d'autres simulations. Ma contribution vise également à apporter un enrichissement au projet CoCoA par l'apport d'un mécanisme de sélection d'action concret basé sur les motivations ainsi que la réalisation d'un atelier de conception de comportement.

Les simulations multi-agents reproduisent un phénomène en modélisant intuitivement son fonctionnement au niveau microscopique. Ce fonctionnement est décrit par le comportement d'entités autonomes qui agissent dans un environnement commun. Toutefois, les approches actuelles restreignent les interactions à des effets de bord ou ne fournissent aucune méthodologie réifiant la notion d'interaction. La conception de simulations contenant un grand nombre d'agents interagissant de manière variée s'en trouve complexifiée. Nous soutenons que pour faciliter la conception des simulations, il est préférable que toute entité soit concrétisée par un agent et tout comportement par une interaction. Le moteur de simulation doit de plus être clairement séparé des agents et interactions, de sorte que tout le système multi-agents soit régi par le même algorithme de simulation. Une telle approche procure de nombreux avantages tels que l'automatisation de l'implémentation, la réutilisabilité des interactions ou la conception graduelle du modèle du phénomène. En nous fondant sur ces principes, nous avons développé une approche centrée sur les interactions (IODA) composée d'une pyramide d'outils : un modèle formel, un ensemble d'algorithmes de simulation et une méthodologie. Nous confirmons la faisabilité de cette approche par une plateforme de simulation paramétrable (JEDI) fidèle au modèle formel et un environnement de développement intégré (JEDI-Builder) qui automatise le passage du modèle IODA au code JEDI. Nous montrons ainsi que la concrétisation logicielle des interactions a conduit à une unification du concept d'agent et à une simplification du processus de conception de simulations.

La coordinació i assignació de tasques en entorns distribuïts ha estat un punt important de la recerca en els últims anys i aquests temes són el cor dels sistemes multi-agent. Els agents en aquests sistemes necessiten cooperar i considerar els altres agents en les seves accions i decisions. A més a més, els agents han de coordinar-se ells mateixos per complir tasques complexes que necessiten més d'un agent per ser complerta. Aquestes tasques poden ser tan complexes que els agents poden no saber la ubicació de les tasques o el temps que resta abans de que les tasques quedin obsoletes. Els agents poden necessitar utilitzar la comunicació amb l'objectiu de conèixer la tasca en l'entorn, en cas contrari, poden perdre molt de temps per trobar la tasca dins de l'escenari. De forma similar, el procés de presa de decisions distribuït pot ser encara més complexa si l'entorn és dinàmic, amb incertesa i en temps real. En aquesta dissertació, considerem entorns amb sistemes multi-agent amb restriccions i cooperatius (dinàmics, amb incertesa i en temps real). En aquest sentit es proposen dues aproximacions que permeten la coordinació dels agents. La primera és un mecanisme semi-centralitzat basat en tècniques de subhastes combinatòries i la idea principal es minimitzar el cost de les tasques assignades des de l'agent central cap als equips d'agents. Aquest algoritme té en compte les preferències dels agents sobre les tasques. Aquestes preferències estan incloses en el bid enviat per l'agent. La segona és un aproximació d'scheduling totalment descentralitzat. Això permet als agents assignar les seves tasques tenint en compte les preferències temporals sobre les tasques dels agents. En aquest cas, el rendiment del sistema no només depèn de la maximització o del criteri d'optimització, sinó que també depèn de la capacitat dels agents per adaptar les seves assignacions eficientment. Addicionalment, en un entorn dinàmic, els errors d'execució poden succeir a qualsevol pla degut a la incertesa i error de accions individuals. A més, una part indispensable d'un sistema de planificació és la capacitat de re-planificar. Aquesta dissertació també proveeix una aproximació amb re-planificació amb l'objectiu de permetre als agent re-coordinar els seus plans quan els problemes en l'entorn no permeti la execució del pla. Totes aquestes aproximacions s'han portat a terme per permetre als agents assignar i coordinar de forma eficient totes les tasques complexes en un entorn multi-agent cooperatiu, dinàmic i amb incertesa. Totes aquestes aproximacions han demostrat la seva eficiència en experiments duts a terme en l'entorn de simulació RoboCup Rescue. / Distributed task allocation and coordination have been the focus of recent research in last years and these topics are the heart of multi-agent systems. Agents in these systems need to cooperate and consider the other agents in their actions and decisions. Moreover, agents may have to coordinate themselves to accomplish complex tasks that need more than one agent to be accomplished. These tasks may be so complicated that the agents may not know the location of them or the time they have before the tasks become obsolete. Agents may need to use communication in order to know the tasks in the environment, otherwise, it may take a long time to find the tasks into the scenario. Similarly, the distributed decisionmaking process may be even more complex if the environment is dynamic, uncertain and real-time. In this dissertation, we consider constrained cooperative multi-agent environments (dynamic, uncertain and real-time). In this regard, we propose two approaches that enable the agents to coordinate themselves. The first one is a semi-centralized mechanism based on combinatorial auction techniques and the main idea is minimizing the cost of assigned tasks from the central agent to the agent teams. This algorithm takes into account the tasks' preferences of the agents. These preferences are included into the bid sent by the agent. The second one is a completely decentralized scheduling approach. It permits agents schedule their tasks taking into account temporal tasks' preferences of the agents. In this case, the system's performance depends not only on the maximization or the optimization criterion, but also on the agents' capacity to adapt their schedule efficiently. Furthermore, in a dynamic environment, execution errors may happen to any plan due to uncertainty and failure of individual actions. Therefore, an indispensable part of a planning system is the capability of replanning. This dissertation is also providing a replanning approach in order to allow agents recoordinate his plans when the environmental problems avoid fulfil them. All these approaches have been carried out to enable the agents to efficiently allocate and coordinate all their complex tasks in a cooperative, dynamic and uncertain multi-agent scenario. All these approaches have demonstrated their effectiveness in experiments performed in the RoboCup Rescue simulation environment.

Made available in DSpace on 2014-12-17T15:47:03Z (GMT). No. of bitstreams: 1 GivanaldoRS_TESE.pdf: 2106802 bytes, checksum: 88486cf095bfcefea309b73b76e7de67 (MD5) Previous issue date: 2013-10-25 / This thesis proposes an architecture of a new multiagent system framework for hybridization of metaheuristics inspired on the general Particle Swarm Optimization framework (PSO). The main contribution is to propose an effective approach to solve hard combinatory optimization problems. The choice of PSO as inspiration was given because it is inherently multiagent, allowing explore the features of multiagent systems, such as learning and cooperation techniques. In the proposed architecture, particles are autonomous agents with memory and methods for learning and making decisions, using search strategies to move in the solution space. The concepts of position and velocity originally defined in PSO are redefined for this approach. The proposed architecture was applied to the Traveling Salesman Problem and to the Quadratic Assignment Problem, and computational experiments were performed for testing its effectiveness. The experimental results were promising, with satisfactory performance, whereas the potential of the proposed architecture has not been fully explored. For further researches, the proposed approach will be also applied to multiobjective combinatorial optimization problems, which are closer to real-world problems. In the context of applied research, we intend to work with both students at the undergraduate level and a technical level in the implementation of the proposed architecture in real-world problems / A presente tese prop?e uma arquitetura multiagente para hibridiza??o de metaheur?sticas, inspirada na t?cnica de Otimiza??o por Nuvem de Part?culas, e tem como principal contribui??o a proposta de uma abordagem efetiva para resolu??o de problemas de otimiza??o combinat?ria. A escolha da Otimiza??o por Nuvem de Part?culas como inspira??o deu-se pelo fato desta t?cnica ser inerentemente multiagente, permitindo explorar os recursos dos sistemas multiagente, tais como as t?cnicas de aprendizado e coopera??o. Na arquitetura proposta, as part?culas s?o agentes aut?nomos com mem?ria e m?todos de decis?o e de aprendizagem, utilizando estrat?gias de busca para se moverem no espa?o de solu??es. Os conceitos de posi??o e velocidade, originalmente definidos na Otimiza??o por Nuvem de Part?culas, s?o redefinidos para esta abordagem. A arquitetura proposta foi aplicada ao Problema do Caixeiro Viajante e ao Problema Quadr?tico de Aloca??o, realizando experimentos computacionais que comprovaram sua efetividade. Os resultados dos experimentos foram bastante promissores, apresentando desempenho satisfat?rio, considerando que o potencial da arquitetura proposta ainda n?o foi totalmente explorado. Em pesquisas futuras, a abordagem proposta ser? aplicada a problemas de otimiza??o combinat?ria multiobjetivo, os quais s?o mais pr?ximos aos problemas do mundo real. No ?mbito da pesquisa aplicada, pretende-se trabalhar tanto com alunos em n?vel de gradua??o como em n?vel t?cnico a aplica??o da arquitetura proposta em problemas pr?ticos do mundo real

As pessoas têm acesso a uma vasta gama de informações devido a grande oferta e aos recursos da Internet, porém despendem muito tempo na busca do que realmente é interessante ou útil para elas. A dificuldade de encontrar a informação correta é aumentada quando a informação disputa a atenção de uma pessoa com uma série de outras informações não tão relevantes. Procurando minimizar esta dificuldade e auxiliar no acesso a informação interessante, são aplicados desde sistemas de recuperação de informação até sistemas de filtragem de informação. Os sistemas de recuperação são amplamente difundidos na Internet através dos motores de busca (por exemplo, google.com, av.com, citeseer.ist.psu.edu), porém um problema neste tipo de aplicação constitui-se na necessidade do usuário apresentar os termos (palavras-chave) que são relevantes para a consulta. filtragem de informação, tendo como representante os Sistemas de Recomendação, surge como uma nova abordagem que procura liberar o usuário da exigência de criar consultas com palavras-chave, ou seja, a filtragem baseada em conteúdo procura casar o perfil do usuário e o conteúdo dos itens a serem recomendados, e então, oferecer alguns destes itens aos usuários. Por fim, surgiram sistemas que não se baseavam na análise do conteúdo dos itens, mas sim na reputação de um item junto aos outros usuários, ou seja, o usuário recebe a recomendação de um item que pode ser do seu interesse frente à colaboração de outros usuários que avaliaram o item. A questão maior nesta abordagem está no quanto a opinião de um usuário que avaliou um item é relevante para servir como colaboração na elaboração da recomendação para outro usuário. Esta tese constitui-se em uma proposta para modelar e incluir a relevância da opinião do usuário no processo de recomendação colaborativa, ou seja, apresenta uma abordagem de Sistemas de Recomendação para recomendar itens baseando-se em informação adicional − definida como relevância da opinião do usuário − além das típicas informações utilizadas na grande maioria dos Sistemas de Recomendação. Esta inclusão da relevância da opinião constitui-se em uma alternativa para que o usuário alvo da recomendação consiga identificar qual a importância de um determinado item recomendado frente à relevância de opinião dos recomendadores. A idéia apresentada é a de que pessoas com maior relevância de opinião poderiam melhor avaliar e recomendar itens. / Nowadays, people have access to a huge amount of information due to the Internet's resources. However they spend too much time searching for interesting, adequate or useful information. The difficulty to find worthwhile information increases when interesting things dispute the user's attention. Information retrieval and information filtering systems are applicable in order to minimize search difficulties, aiming to aid the user in the search for worthwhile information. Information retrieval systems are widely spread in the Internet through search engines (e.g., google.com, av.com, citeseer.ist.psu.edu). However there is a problem in this kind of application, which consists in compelling the user to know the terms (keywords) that are relevant for the search. Recommender Systems are an information filtering solution. They present a different approach that frees the user from creating queries with keywords. It means that the system tries to match the user's profile (historical interests) with the content of items to be recommended, and then offers these items to the user (recommendee). In parallel, an alternative approach to item recommendation was proposed, this one based on the offering of items based on other users’ opinion, i.e. the user receives an item recommendation based on the evaluation of other users (collaborative filtering or social filtering). However, a different question is raised here − how much the opinion of a user who evaluated an item is relevant to be employed in the recommendation process applying a collaborative method? This thesis presents a new approach to model and include in the collaborative recommendation process additional information named Recommender's Rank, which represents the relevance of the user's opinion and complements the typical information used in the large majority of Recommender Systems. This approach is an alternative to aid the user to identify the importance of a recommended item based on other users' opinions, as people with higher relevance of opinion are more likely to better evaluate and recommend items.

Cette thèse s’intéresse à la question de l’évaluation des Systèmes Multi-Agents (SMA). Les caractéristiques propres que possèdent ces derniers, notamment en termes d’autonomie, de distribution, de dynamique et de socialité, ont grandement contribué à l’élargissement de leurs champs d’application, mais en contrepartie, elles ont rendu leur analyse plus ardue. Ainsi, les méthodes d’évaluation dans les systèmes informatiques classiques s’avèrent insuffisantes à analyser les SMA, étant donné qu’elles ne tiennent pas compte de leurs spécificités. L’objectif de cette thèse consiste donc à proposer une approche générique pour l’évaluation des SMA en se basant sur la mesure de leurs caractéristiques fonctionnelles. A cet effet, le besoin de disposer d’informations sur l’exécution du système à évaluer est manifeste. C’est dans ce cadre qu’une nouvelle approche d’observation des SMA est proposée. Les résultats de ces observations sont exploités pour construire une abstraction du système sous forme d’un modèle, lequel est étudié pour définir les mesures de performances. L’analyse se focalise sur deux caractéristiques essentielles, à la base de la dynamique et de la socialité des SMA : la communication et l’organisation. Les expérimentations de la solution proposée portent sur deux applications multi-agents. La première est une application de diagnostic des pannes dans un environnement industriel et la seconde est une application de pilotage et de gestion de la production dans les chaînes logistiques. / This thesis focuses on the issue of MultiAgent Systems (MAS) evaluation. The MAS own characteristics, namely autonomy, distribution, dynamicity and sociality, have greatly contributed to the expansion of their application scope; but in return they made their analysis more difficult. Thus, evaluation methods in classic computer systems are insufficient to analyse MAS, since they do not take into account their specificities. The objective of this thesis is to provide a generic approach for the evaluation of MAS by measuring their functional characteristics. To this end, the need for information about the execution of the system to be evaluated is evident. In this context, a new approach to observe MAS is proposed. The results of these observations are exploited to build an abstraction model of the system which is studied in order to define performance metrics. The analysis focuses on two key characteristics, at the basis on the dynamics and sociality in MAS: communication and organization. The experiments of the proposed solution are performed on two multiagent applications. The first is an application of fault diagnosis in an industrial environment and the second is an application of control and production planning in supply chains.

L'observation des systèmes complexes artificiels à grande échelle est difficile en raison de leur dimension et de leur ouverture. Cette thèse propose le modèle SMA-R (Système Multi-Agent Récursif) basé sur la récursivité pour l'observation multi-niveau des systèmes complexes artificiels. À partir d'un Système Multi-Agent donné, ce modèle est capable de construire une représentation abstraite multi-niveau. Le modèle d'agent récursif proposé possède un module contenant les connaissances, les capacités et le contexte de récursivité; un module pour observer les changements; deux mécanismes pour construire ou détruire les niveaux abstraits; un module d'interaction récursive prenant en charge des comportements collectifs et individuels. Pour appliquer ce modèle à des SMA physiquement décentralisés, nous avons proposé une architecture générique décentralisée d'agents récursifs en adoptant les concepts du modèle OSI qui offre en avant les capacités que nous cherchons: multi niveau, encapsulation, échange de messages virtuels et physiques... Un framework décentralisé générique a été développé permettant aux agents réels de construire les niveaux d'observation. Ce framework est appliqué à l'observation d'un réseau de capteurs sans fil. / Observation of large scale artificials systems is difficult because of their dimension and their openness. This thesis proposes a model SMA-R (Recursive Multi-Agent Systems) based on recursion for multi-level observation of artificial complex systems. From a given SMA, this model is able to build multi-level of abstractions. The model's agent has a module containning knowledge, skills and context of recursion, an eye to observe changes; two mechanisms to build or destroy an abstract level, a module of recursive interaction to support collective and individual behaviors. For applying this model to SMA physically decentralized, we proposed a generic decentralized architecture for recursive agent by adopting the concepts of the OSI model which offerts forward capabilities that we look for: multi-level encapsulation, exchange of virtual and physical messages ... A generic decentralized framework was developed allowing applicatives agents to build multi-levels observation. This framework is applied to the observation of a wireless sensor network.

[pt] Normas são aplicadas em sistemas multiagentes como mecanismos capazes de restringir o comportamento dos agentes de software com o objetivo de alcançar uma ordem social desejável. Entretanto, essas normas podem entrar em conflito, como por exemplo, uma norma que proíbe um agente de realizar uma determinada ação e outra norma que obriga o mesmo agente a realizar a mesma ação no mesmo intervalo de tempo. A decisão do agente sobre quais normas serão cumpridas pode ser definida com base nas recompensas e punições normativas e nos objetivos do agente. No entanto, em determinadas situações a avaliação desses atributos pode não ser o suficiente para permitir que o agente efetue uma tomada de decisão satisfatória. Nesse contexto, foi elaborada uma abordagem que considera traços de personalidade em agentes de software para aprimorar o processo de resolução de conflitos normativos e a escolha dos planos para tomada decisões, além de realizar a comparação da abordagem proposta com diferentes abordagens encontradas na literatura. / [en] Norms are applied in multiagent systems as mechanisms capable of restricting the behavior of software agents in order to achieve a desirable social order. However, norms eventually can be conflicting - for example, when there is a norm that prohibits an agent to perform a particular action and another norm that obligates the same agent to perform the same action in the same period of time. The agent s decision about which norms to fulfill can be defined based on rewards, punishments and agent goals. Sometimes, this balance will not be enough to allow the agent to make the best decision. In this context, this proposal introduces an approach that considers the agent s personality traits in order to improve the plan decision-making process and resolving normative conflicts. Our approach s applicability and validation is demonstrated by an experiment that reinforces the importance of considering the norms both in the agent and society s points of view.

As pessoas têm acesso a uma vasta gama de informações devido a grande oferta e aos recursos da Internet, porém despendem muito tempo na busca do que realmente é interessante ou útil para elas. A dificuldade de encontrar a informação correta é aumentada quando a informação disputa a atenção de uma pessoa com uma série de outras informações não tão relevantes. Procurando minimizar esta dificuldade e auxiliar no acesso a informação interessante, são aplicados desde sistemas de recuperação de informação até sistemas de filtragem de informação. Os sistemas de recuperação são amplamente difundidos na Internet através dos motores de busca (por exemplo, google.com, av.com, citeseer.ist.psu.edu), porém um problema neste tipo de aplicação constitui-se na necessidade do usuário apresentar os termos (palavras-chave) que são relevantes para a consulta. filtragem de informação, tendo como representante os Sistemas de Recomendação, surge como uma nova abordagem que procura liberar o usuário da exigência de criar consultas com palavras-chave, ou seja, a filtragem baseada em conteúdo procura casar o perfil do usuário e o conteúdo dos itens a serem recomendados, e então, oferecer alguns destes itens aos usuários. Por fim, surgiram sistemas que não se baseavam na análise do conteúdo dos itens, mas sim na reputação de um item junto aos outros usuários, ou seja, o usuário recebe a recomendação de um item que pode ser do seu interesse frente à colaboração de outros usuários que avaliaram o item. A questão maior nesta abordagem está no quanto a opinião de um usuário que avaliou um item é relevante para servir como colaboração na elaboração da recomendação para outro usuário. Esta tese constitui-se em uma proposta para modelar e incluir a relevância da opinião do usuário no processo de recomendação colaborativa, ou seja, apresenta uma abordagem de Sistemas de Recomendação para recomendar itens baseando-se em informação adicional − definida como relevância da opinião do usuário − além das típicas informações utilizadas na grande maioria dos Sistemas de Recomendação. Esta inclusão da relevância da opinião constitui-se em uma alternativa para que o usuário alvo da recomendação consiga identificar qual a importância de um determinado item recomendado frente à relevância de opinião dos recomendadores. A idéia apresentada é a de que pessoas com maior relevância de opinião poderiam melhor avaliar e recomendar itens. / Nowadays, people have access to a huge amount of information due to the Internet's resources. However they spend too much time searching for interesting, adequate or useful information. The difficulty to find worthwhile information increases when interesting things dispute the user's attention. Information retrieval and information filtering systems are applicable in order to minimize search difficulties, aiming to aid the user in the search for worthwhile information. Information retrieval systems are widely spread in the Internet through search engines (e.g., google.com, av.com, citeseer.ist.psu.edu). However there is a problem in this kind of application, which consists in compelling the user to know the terms (keywords) that are relevant for the search. Recommender Systems are an information filtering solution. They present a different approach that frees the user from creating queries with keywords. It means that the system tries to match the user's profile (historical interests) with the content of items to be recommended, and then offers these items to the user (recommendee). In parallel, an alternative approach to item recommendation was proposed, this one based on the offering of items based on other users’ opinion, i.e. the user receives an item recommendation based on the evaluation of other users (collaborative filtering or social filtering). However, a different question is raised here − how much the opinion of a user who evaluated an item is relevant to be employed in the recommendation process applying a collaborative method? This thesis presents a new approach to model and include in the collaborative recommendation process additional information named Recommender's Rank, which represents the relevance of the user's opinion and complements the typical information used in the large majority of Recommender Systems. This approach is an alternative to aid the user to identify the importance of a recommended item based on other users' opinions, as people with higher relevance of opinion are more likely to better evaluate and recommend items.