Economic changes such as rising açaí prices and the availability of off-farm employment are transforming the landscape of the Amazonian várzea, subject to decision-making at the farming household level. Land use change results from complex human-environment interactions which can be addressed by an agent-based model. An agent-based model is a simulation model composed of autonomous interacting entities known as agents, built from the bottom-up. Coupled with cellular automata, which forms the agents’ environment, agent-based models are becoming an important tool of land use science, complementing traditional methods of induction and deduction. The decision-making methods employed by agent-based models in recent years have included optimization, imitation, heuristics, classifier systems and genetic algorithms, among others, but multiple methods have rarely been comparatively analyzed. A modular agent-based model is designed to allow the researcher to substitute alternative decision-making methods. For a smallholder farming community in Marajó Island near Ponta de Pedras, Pará, Brazil, 21 households are simulated over a 40-year period. In three major scenarios of increasing complexity, these households first face an environment where goods sell at a constant price throughout the simulated period and there are no outside employment opportunities. This is followed by a scenario of variable prices based on empirical data. The third scenario combines variable prices with limited employment opportunities, creating multi-sited households as members emigrate. In each scenario, populations of optimizing agents and heuristic agents are analyzed in parallel. While optimizing agents allocate land cells to maximize revenue using linear programming, fast and frugal heuristic agents use decision trees to quickly pare down feasible solutions and probabilistically select between alternatives weighted by expected revenue. Using distributed computing, the model is run through several parameter sweeps and results are recorded to a cenral database. Land use trajectories and sensitivity analyses highlight the relative biases of each decision-making method and illustrate cases where alternative methods lead to significantly divergent outcomes. A hybrid approach is recommended, employing alternative decision-making methods in parallel to illustrate inefficiencies exogenous and endogenous to the decision-maker, or allowing agents to select among multiple methods to mitigate bias and best represent their real-world analogues.

The amount of information on the World Wide Web (WWW) is rapidly growing in pace and topic diversity. This has made it increasingly difficult, and often frustrating, for information seekers to retrieve the content they are looking for as information retrieval systems (e.g., search engines) are unable to decipher the relevance of the retrieved information as it pertains to the information they are searching for. This issue can be decomposed into two aspects: 1) variability of information relevance as it pertains to an information seeker. In other words, different information seekers may enter the same search text, or keywords, but expect completely different results. It is therefore, imperative that information retrieval systems possess an ability to incorporate a model of the information seeker in order to estimate the relevance and context of use of information before presenting results. Of course, in this context, by a model we mean the capture of trends in the information seeker's search behaviour. This is what many researchers refer to as the personalized search. 2) Information diversity. Information available on the World Wide Web today spans multitudes of inherently overlapping topics, and it is difficult for any information retrieval system to decide effectively on the relevance of the information retrieved in response to an information seeker's query. For example, the information seeker who wishes to use WWW to learn about a cure for a certain illness would receive a more relevant answer if the search engine was optimized into such domains of topics. This is what is being referred to in the WWW nomenclature as a 'specialized search'. This thesis maintains that the information seeker's search is not intended to be completely random and therefore tends to portray itself as consistent patterns of behaviour. Nonetheless, this behaviour, despite being consistent, can be quite complex to capture. To accomplish this goal the thesis proposes a Multi-Agent Personalized Information Retrieval with Specialization Ontology (MAPIRSO). MAPIRSO offers a complete learning framework that is able to model the end user's search behaviour and interests and to organize information into categorized domains so as to ensure maximum relevance of its responses as they pertain to the end user queries. Specialization and personalization are accomplished using a group of collaborative agents. Each agent employs a Reinforcement Learning (RL) strategy to capture end user's behaviour and interests. Reinforcement learning allows the agents to evolve their knowledge of the end user behaviour and interests as they function to serve him or her. Furthermore, REL allows each agent to adapt to changes in an end user's behaviour and interests. Specialization is the process by which new information domains are created based on existing information topics, allowing new kinds of content to be built exclusively for information seekers. One of the key characteristics of specialization domains is the seeker centric - which allows intelligent agents to create new information based on the information seekers' feedback and their behaviours. Specialized domains are created by intelligent agents that collect information from a specific domain topic. The task of these specialized agents is to map the user's query to a repository of specific domains in order to present users with relevant information. As a result, mapping users' queries to only relevant information is one of the fundamental challenges in Artificial Intelligent (AI) and machine learning research. Our approach employs intelligent cooperative agents that specialize in building personalized ontology information domains that pertain to each information seeker's specific needs. Specializing and categorizing information into unique domains is one of the challenge areas that have been addressed and various proposed solutions were evaluated and adopted to address growing information. However, categorizing information into unique domains does not satisfy each individualized information seeker. Information seekers might search for similar topics, but each would have different interests. For example, medical information of a specific medical domain has different importance to both the doctor and patients. The thesis presents a novel solution that will resolve the growing and diverse information by building seeker centric specialized information domains that are personalized through the information seekers' feedback and behaviours. To address this challenge, the research examines the fundamental components that constitute the specialized agent: an intelligent machine learning system, user input queries, an intelligent agent, and information resources constructed through specialized domains. Experimental work is reported to demonstrate the efficiency of the proposed solution in addressing the overlapping information growth. The experimental work utilizes extensive user-centric specialized domain topics. This work employs personalized and collaborative multi learning agents and ontology techniques thereby enriching the queries and domains of the user. Therefore, experiments and results have shown that building specialized ontology domains, pertinent to the information seekers' needs, are more precise and efficient compared to other information retrieval applications and existing search engines.

近年來，在電腦動畫的應用中，虛擬人群模擬的需求越來越多；但人群運動的模擬對於動畫設計師而言，仍是一件十分繁瑣耗時的工作。過去有許多研究曾以虛擬力場模擬簡單的生物群聚行為，但所模擬出的動畫品質與虛擬力場的參數及虛擬環境息息相關，因此經常需要以人工的方式耗時地調整出適當的虛擬力場參數。因此，我們提議以此問題定義成一個基因演算法的問題，針對不同的移動行為，定義適切的適應函數，再由系統根據不同環境自動演化出適當的虛擬力權重組合，以供產生不同人群移動行為之動畫時參考。在本篇論文中，我們已完成基因演算法的設計及人群動畫模擬系統，並設計了不同的典型環境進行電腦模擬實驗，以驗證此方法的可行性。 / The demands for virtual crowd simulation have been increasing in recent years but creating realistic crowd motions remains a complex and time-consuming task for a computer animator. In the literature、much work has been proposed to use virtual forces to simulate the motion of a group of virtual creatures such as birds and fishes. However、the quality of the simulations largely depends on the weights of the component virtual forces as well as the scene where the agents are situated. Usually it requires the animator to tune these parameters for a specific scene in order to obtain the desired result. In this thesis、we propose to use genetic algorithm to generate an optimal set of weighting parameters for composing virtual forces according to the given environment and desired movement behavior. We have implemented the proposed genetic algorithm as well as the crowd simulation system. Extensive experiments have also been conducted to study the effects of typical scenes and behaviors on the parameter sets and verify the feasibility of the approach.

The impact of spatial characteristics on the dynamics of urban development is a topic of great interest in urban studies. The interaction between the residents and the spatial characteristics is of particular interest in the context of urban models where some of the most famous urban models have been based on the process of individual settlements in some specific parts of cities.This research investigates the dynamism of urban development modeled by Cellular Automata and Multi-Agent System. The urban development, in this study embraces urban renewal and residential mobility. It corresponds to the residential mobility of households, attracted by residential and centrality comfort; these comforts are crystallized in some areas and residences of Strasbourg. The diversity and quality of these comforts become criteria for residential choice in a way that the households seek for proximity to these comforts.The Cellular Automata in this study, models the spatial characteristics of urban spatial units and they are identified by some inherent attributes that are equal to the comfort in residences and urban areas. The Multi- Agent System represent a system in which the population of the city interact between them and between them and the city; the agents delegate the socio-professional classes of households. During the spatiotemporal change, the aspiration of households forms the socio-spatio-temporal development of the city.

The amount of information on the World Wide Web (WWW) is rapidly growing in pace and topic diversity. This has made it increasingly difficult, and often frustrating, for information seekers to retrieve the content they are looking for as information retrieval systems (e.g., search engines) are unable to decipher the relevance of the retrieved information as it pertains to the information they are searching for. This issue can be decomposed into two aspects: 1) variability of information relevance as it pertains to an information seeker. In other words, different information seekers may enter the same search text, or keywords, but expect completely different results. It is therefore, imperative that information retrieval systems possess an ability to incorporate a model of the information seeker in order to estimate the relevance and context of use of information before presenting results. Of course, in this context, by a model we mean the capture of trends in the information seeker's search behaviour. This is what many researchers refer to as the personalized search. 2) Information diversity. Information available on the World Wide Web today spans multitudes of inherently overlapping topics, and it is difficult for any information retrieval system to decide effectively on the relevance of the information retrieved in response to an information seeker's query. For example, the information seeker who wishes to use WWW to learn about a cure for a certain illness would receive a more relevant answer if the search engine was optimized into such domains of topics. This is what is being referred to in the WWW nomenclature as a 'specialized search'. This thesis maintains that the information seeker's search is not intended to be completely random and therefore tends to portray itself as consistent patterns of behaviour. Nonetheless, this behaviour, despite being consistent, can be quite complex to capture. To accomplish this goal the thesis proposes a Multi-Agent Personalized Information Retrieval with Specialization Ontology (MAPIRSO). MAPIRSO offers a complete learning framework that is able to model the end user's search behaviour and interests and to organize information into categorized domains so as to ensure maximum relevance of its responses as they pertain to the end user queries. Specialization and personalization are accomplished using a group of collaborative agents. Each agent employs a Reinforcement Learning (RL) strategy to capture end user's behaviour and interests. Reinforcement learning allows the agents to evolve their knowledge of the end user behaviour and interests as they function to serve him or her. Furthermore, REL allows each agent to adapt to changes in an end user's behaviour and interests. Specialization is the process by which new information domains are created based on existing information topics, allowing new kinds of content to be built exclusively for information seekers. One of the key characteristics of specialization domains is the seeker centric - which allows intelligent agents to create new information based on the information seekers' feedback and their behaviours. Specialized domains are created by intelligent agents that collect information from a specific domain topic. The task of these specialized agents is to map the user's query to a repository of specific domains in order to present users with relevant information. As a result, mapping users' queries to only relevant information is one of the fundamental challenges in Artificial Intelligent (AI) and machine learning research. Our approach employs intelligent cooperative agents that specialize in building personalized ontology information domains that pertain to each information seeker's specific needs. Specializing and categorizing information into unique domains is one of the challenge areas that have been addressed and various proposed solutions were evaluated and adopted to address growing information. However, categorizing information into unique domains does not satisfy each individualized information seeker. Information seekers might search for similar topics, but each would have different interests. For example, medical information of a specific medical domain has different importance to both the doctor and patients. The thesis presents a novel solution that will resolve the growing and diverse information by building seeker centric specialized information domains that are personalized through the information seekers' feedback and behaviours. To address this challenge, the research examines the fundamental components that constitute the specialized agent: an intelligent machine learning system, user input queries, an intelligent agent, and information resources constructed through specialized domains. Experimental work is reported to demonstrate the efficiency of the proposed solution in addressing the overlapping information growth. The experimental work utilizes extensive user-centric specialized domain topics. This work employs personalized and collaborative multi learning agents and ontology techniques thereby enriching the queries and domains of the user. Therefore, experiments and results have shown that building specialized ontology domains, pertinent to the information seekers' needs, are more precise and efficient compared to other information retrieval applications and existing search engines.

En esta tesis se propone el uso de agentes inteligentes en entornos de aprendizaje en línea con el fin de mejorar la asistencia y motivación del estudiante a través de contenidos personalizados que tienen en cuenta el estilo de aprendizaje del estudiante y su nivel de conocimiento. Los agentes propuestos se desempeñan como asistentes personales que ayudan al estudiante a llevar a cabo las actividades de aprendizaje midiendo su progreso y motivación. El entorno de agentes se construye a través de una arquitectura multiagente llamada MASPLANG diseñada para dar soporte adaptativo (presentación y navegación adaptativa) a un sistema hipermedia educativo desarrollado en la Universitat de Girona para impartir educación virtual a través del web.Un aspecto importante de esta propuesta es la habilidad de construir un modelo de estudiante híbrido que comienza con un modelo estereotípico del estudiante basado en estilos de aprendizaje y se modifica gradualmente a medida que el estudiante interactúa con el sistema (gustos subjetivos). Dentro del contexto de esta tesis, el aprendizaje se define como el proceso interno que, bajo factores de cambio resulta en la adquisición de la representación interna de un conocimiento o de una actitud. Este proceso interno no se puede medir directamente sino a través de demostraciones observables externas que constituyen el comportamiento relacionado con el objeto de conocimiento. Finalmente, este cambio es el resultado de la experiencia o entrenamiento y tiene una durabilidad que depende de factores como la motivación y el compromiso.El MASPLANG está compuesto por dos niveles de agentes: los intermediarios llamados IA (agentes de información) que están en el nivel inferior y los de Interfaz llamados PDA (agentes asistentes) que están en el nivel superior. Los agentes asistentes atienden a los estudiantes cuando trabajan con el material didáctico de un curso o una lección de aprendizaje. Esta asistencia consiste en la recolección y análisis de las acciones de los estudiantes para ofrecer contenidos personalizados y en la motivación del estudiante durante el aprendizaje mediante el ofrecimiento de contenidos de retroalimentación, ejercicios adaptados al nivel de conocimiento y mensajes, a través de interfaces de usuario animadas y atractivas. Los agentes de información se encargan del mantenimiento de los modelos pedagógico y del dominio y son los que están en completa interacción con las bases de datos del sistema (compendio de actividades del estudiante y modelo del dominio).El escenario de funcionamiento del MASPLANG está definido por el tipo de usuarios y el tipo de contenidos que ofrece. Como su entorno es un sistema hipermedia educativo, los usuarios se clasifican en profesores quienes definen y preparan los contenidos para el aprendizaje adaptativo, y los estudiantes quienes llevan a cabo las actividades de aprendizaje de forma personalizada. El perfil de aprendizaje inicial del estudiante se captura a través de la evaluación del cuestionario ILS (herramienta de diagnóstico del modelo FSLSM de estilos de aprendizaje adoptado para este estudio) que se asigna al estudiante en su primera interacción con el sistema. Este cuestionario consiste en un conjunto de preguntas de naturaleza sicológica cuyo objetivo es determinar los deseos, hábitos y reacciones del estudiante que orientarán la personalización de los contenidos y del entorno de aprendizaje. El modelo del estudiante se construye entonces teniendo en cuenta este perfil de aprendizaje y el nivel de conocimiento obtenido mediante el análisis de las acciones del estudiante en el entorno.

In recent years many people have built Personal Assistant Agents, Information Agents and the like, and have simply added them to the operating system as auxiliary applications, without regard to architecture. This thesis argues that an agent architecture, one designed as a sophisticated representation of an individual user, should be embedded deep in the device system software, with at least equal status to the GUI – the graphical user interface. A sophisticated model of the user is then built, drawing upon contemporary Analytical Psychology – the Psychology of Subselves. The Shadowboard Agent architecture is then built upon that user model, drawing both structural and computational implications from the underlying psychology. An XML DTD file named Shadowboard.dtd is declared as a practical manifestation of the semantics of Shadowboard. An implementation of the Shadowboard system is mapped out, via a planned conversion of two existing integrated systems: SlimWinX, an event-driven GUI system; and XSpaces, an object-oriented tuplespace system with Blackboard-like features. The decision making mechanism passes logic terms and contraints between the various sub-agent components (some of which take on the role of Constraint Solvers), giving this agent system some characteristics of a Generalised Constraint Solver. A Shadowboard agent (built using the system) consists of a central controlling autonomous agent named the Aware Ego Agent, and any number of sub-agents, which collectively form an integrated but singular whole agent modelled on the user called the Digital Self. One such whole-agent is defined in a file named DigitalSelf.xml – which conforms to the schema in Shadowboard.dtd - which offers a comprehensive and generic representation of a user’s stance in a 24x7 network, in particular - the Internet. Numerous types of Shadowboard sub-agents are declared.

Cette thèse présente une contribution sur les problèmes de contrôle de réseaux d'irrigations en tenant compte des contraintes de communication grâce à une approche multicouches d’intelligence distribuée. Les analyses détaillées de chaque couche avec les résultats analytiques et les simulations seront décrites dans les différents chapitres. Ils mettent l'accent sur l'intérêt de l'approche multicouches, plus précisément sur son efficacité et sa fiabilité pour la supervision, l'optimisation multi-objectifs et le contrôle coopératif distribué sur des systèmes complexes de transport d'eau.La première couche analysé est le réseau hydraulique composé de canaux d’écoulements à surface libre, de sous-réseaux maillés de tuyaux sous pression et des structures hydrauliques (pompes vannes, ..). En intégrant les équations de Saint-Venant pour décrire l’écoulement physique des fluides en surface libre et la méthode Lattice Boltzmann pour la simulation du fluide, nous obtenons un modèle non linéaire discret pour les canaux à surface libre. Les structures hydrauliques sont généralement traitées comme des limites internes des biefs (tronçons) et modélisées par des relations entre les variables de flux et de pression.Permettant l'échange d'informations entre les éléments du système de contrôle, le réseau de communication sera considéré comme la deuxième couche. La résolution des problèmes d’hétérogénéités des systèmes et des communications (par exemple les retards de diffusion dans le réseau, la perte de paquets, la consommation d'énergie) sera étudié en introduisant une architecture de réseau hybride avec un routage dynamique basé sur les exigences de Qualité de Service (QoS) des applications de contrôle. Pour le routage dynamique dans le réseau, une composition pondérée de certaines métriques standards est proposée afin que le protocole de routage utilisant cette métrique composite converge sans boucle avec une « route » optimum. Grâce à différents scénarios de simulation, plusieurs critères de performance du réseau ont été évalués. La comparaison des résultats de simulation permet de valider l'intérêt de cette approche de composition pour le routage dynamique.Une troisième couche propose un système de contrôle réactif optimal développé pour la régulation du réseau d'irrigation dans un modèle étendue à grande échelle : Distributed Cooperative Model Predictive Control (DCMPC). Cette partie aborde la mise en œuvre de différentes stratégies de contrôle (centralisées, décentralisées et distribuées) et intègre la communication coopérative entre les contrôleurs MPC locaux afin d’améliorer les performances global es du système. La gestion de la divergence dans l'échange d'informations entre les contrôleurs est considérée comme un problème de consensus et résolue en utilisant un protocole de consensus asynchrone. Cette approche du contrôle distribué basée sur le paradigme des systèmes multi-agents, fournit une solution garantissant que tous les contrôleurs aient une vue cohérente de certaines valeurs des données nécessaires pour le calcul de décision. Un cas d’application sur un canal d'irrigation est étudié dans les simulations. La comparaison des résultats de simulations valide les avantages de l'approche du contrôle distribué coopératif par rapport aux autres stratégies de contrôle. / This thesis presents control problems of irrigation network with communication constraints and a multi-layer approach to solve these problems in a distributed manner. Detailed discussions of each layer with analytical and simulation results are described throughout several chapters. They emphasize the potential interest of the multi-layer approach, more precisely its efficiency and reliability for supervision, multi-objective optimization and distributed cooperative control of complex water transport systems. Conventionally, the first layer to be considered is the hydraulic network composed of free-surface channels, hydraulic structures and mesh subnetwork of pressurized pipes. By coupling the Saint-Venant equations for describing the physics of free-surface fluid and the Lattice Boltzmann method for the fluid simulation, a discrete-time nonlinear model is obtained for channel reaches. The hydraulic structures are usually treated as internal boundaries of reaches and modeled by algebraic relationships between the flow and pressure variables. To enable the exchange of information among the control system’s components, a communication network is considered in the second layer. Solving challenging problems of heterogeneous devices and communication issues (e.g., network delay, packet loss, energy consumption) is investigated in this thesis by introducing a hybrid network architecture and a dynamic routing design based on Quality of Service (QoS) requirements of control applications. For network routing, a weighted composition of some standard metrics is proposed so that the routing protocol using the composite metric achieves convergence, loop-freeness and path-optimality properties. Through extensive simulation scenarios, different network performance criteria are evaluated. The comparison of simulation results can validate the interest of this composition approach for dynamic routing. Finally, the third layer introduces an optimal reactive control system developed for the regulatory control of large-scale irrigation network under a Distributed Cooperative Model Predictive Control (DCMPC) framework. This part discusses the implementation of different control strategies (e.g., centralized, decentralized, and distributed strategies) and how the cooperative communication among local MPC controllers can be included to improve the performance of the overall system. Managing divergent (or outdated) information exchange among controllers is considered in this thesis as a consensus problem and solved by an asynchronous consensus protocol. This approach based on the multi-agent system paradigm to distributed control requires each controller to agree with its neighbors on some data values needed during action computation. For simulations, a particular benchmark of an irrigation channel is considered. The comparison of simulation results validate the benefits of the distributed cooperative control approach over other control strategies.

Le travail présenté dans ce mémoire contribue au domaine de l’entreprise étendue et le développement des systèmes d’informations distribués. et le développement des systèmes d’informations distribués. C’est bien évidemment un sujet d’étude important pour la communauté Logistique (chaîne logistique), mais aussi pour la communauté Génie logiciel. C’est dans cette perspective que se situent les objectifs de proposer une modélisation de la chaîne logistique globale dans un environnement de transport intermodal de porte à porte en vue de résoudre les problèmes : (i) d’interfaces entre les différents acteurs intervenants le long de la chaîne et (ii) de rupture des charges engendrés par les opérations de transfert de marchandises d’un mode de transport à l’autre. Ainsi, l’amélioration de performance de la chaine logistique dépend fortement du niveau de coopération et de coordination, et surtout du partage et de la validité des informations et des connaissances, entre ces différents acteurs de la chaîne (organisateur du transport, transitaire, fournisseur, etc.). L’applicatif se situe dans le corridor de la vallée de Seine. Le travail s’inscrit dans le projet APLOG (Amélioration et Performance de la LOgistique Globale) financé par la région Haute Normandie. / The work presented in this thesis contributes to the field of the extended enterprise and the development of distributed information systems. This is obviously an important subject of study for the logistics community (supply chain), but also for the software engineering community. It is in this context that the objectives are to provide a model of global supply chain in an intermodal environment door to door service to solve the problems: (i) interfaces between the different actors involved along the chain and (ii) trans-loading operations generated by the goods transfer operations from one mode of transport to another. Thus, improved performance of the supply chain depends heavily on the level of cooperation and coordination, and especially sharing and validity of information and knowledge between the different actors in the chain (transport organizer, forwarding, Supplier, etc.). The application is located in the corridor of the Seine valley. The work is part of the project APLOG (Performance Improvement and global logistics) financed by the Haute Normandie region.

La gestion de l'eau peut être un domaine complexe, incertain et conflictuel. Dans différentes régions du monde il se confronte à de nombreux problèmes, tel que la disparité des intérêts liés à la ressource de l’eau, plusieurs décideurs, des réseaux administratives complexes, la distribution d'eau inefficace, divers événements socio-politiques et le changement climatique. Par conséquent, la prise de décision a lieu dans un système fortement interconnecté, dans lequel ni les ramifications décisionnels ni la complexité de ses impacts peuvent être négligés. Dans la région des Pouilles, la rareté de l'eau est le principal problème croissant qui touche les communautés humaines et plus qu’humaines.La pénurie d'eau génère la nécessité d'améliorer les processus décisionnels collaboratifs avec agents multiples. Les chercheurs suggèrent que la «tragédie des communs» pourrait être évitée si une ressource partagée était gérée de manière collective. Cela nécessite le développement d'outils dynamiques d’aide à la décision. Ceux-ci devraient être capables d'intégrer les différents cadres de problèmes détenues par les décideurs, de clarifier les différences entre les cadres, de soutenir la création d'un processus collaboratif pour la structuration d’un problème et de fournir des plates-formes communes et des espaces d'interaction.À cet égard, nous avons construit un espace d'interaction dynamique (DIS), mettant en évidence les points critiques opératoires et permettant aux analystes d'identifier une définition commune du problème. Les nouveaux défis de la collecte et de l'échange de connaissances et de la représentation des concepts structurés peuvent être résolus par une approche combinée. Les systèmes multi-agents joints aux systèmes dynamiques pourraient fournir des alternatives non conventionnelles qui utilisent des composants physiques et sociales, avec une attention particulière sur les comportements individuels et collectifs dans la gestion des ressources avec plusieurs décideurs.Dans notre étude de cas, le modèle a été utilisé comme une plate-forme pour la modélisation des organisations multi-agents, afin de soutenir la prise de décision collective dans la gestion de l'eau. Le modèle est capable de représenter un système de gestion de l'eau distribuée complexe, où les comportements simulés sont basées sur des observations sur le terrain et sur la participation des parties prenantes. De plus, l'approche de système multi-agents permet l'interaction et la formalisation des comportements des usagers de l'eau dans le processus de gestion. Une modélisation type systèmes dynamiques dans un environnement d'interaction entre agents de décision, nous permet d’intégrer explicitement les différents cadres et de simuler les interactions lors de l'adoption d'une nouvelle politique. Le modèle peut montrer comment la compréhension limitée de l'espace d'interaction affecte les actions suivies par chacun des décideurs et, enfin, comment elle pourrait conduire à des mécanismes de résistance systémique. En conclusion, le résultat est l’image la plus riche possible de la situation du problème existant, qui traite de la gestion de l'eau d'irrigation dans les systèmes agricoles. / Water resource management can be a complex, uncertain and conflictual domain. It faces numerous problems in many regions of the world, such as the disparity of interests associated with the water resource, multiple decision makers, complex networks of administration, inoperative water distribution, various socio-political events and climate change. Consequently, environmental decision-making takes place in a highly interconnected system, in which neither the decisional ramifications nor the complexity of its impacts can be neglected. In the Apulia Region, water scarcity is the main rising problem and is affecting human and more-than-human communities.Water scarcity generates the need to enhance collaborative multi-agent decision-making processes. Researchers suggest that the “tragedy of commons” could be avoided when a shared resource is at stake, provided that communities interact and operate in a collective way and avoid, for example, the market rules constraints. This requires the development of dynamic decision-aiding tools. They should be capable to integrate the different problem frames held by the decision makers, to clarify the differences among those frames, to support the creation of a collaborative problem structuring process and to provide shared platforms and interaction spaces.In this regard, we built a dynamic interaction space (DIS), highlighting the operative criticalities and allowing the analysts to identify a shared problem definition. The emerging issues of gathering and exchanging knowledge and representing structured concepts can be solved through a combined approach. Multi-agent systems joined with system dynamics can provide unconventional alternatives that use physical and social components, with a particular focus on individual and collective behaviours in resource management with multiple decision makers.In our case study, the model was used as a platform for modelling multi-agent organizations, in order to support collective decision-making in water management. The model is capable of representing a distributed complex water management system, where simulated behaviours are based on field observations and on the participation of stakeholders. What is more, the multi–agent system approach enables the interaction and allows to formalize theIrene Pluchinotta – “Multi-Agent Modelling For Distributed Intelligent Decision In Water Management”iibehaviours of water users in the management process. A system dynamics modelling in an environment of interacting decision agents, allows us to explicitly consider the different frames and to simulate interactions when adopting a new policy. The model can showcase how the limited understanding of the interaction space affects the actions followed by each decision-makers and, finally, how it could lead to policy resistance mechanisms. In conclusion, the result is the richest possible picture of the existing problem situation that deals with irrigation water management in agricultural systems.