Embora haja muitas pesquisas na área de teamwork, o desenvolvimento de times de agentes para ambientes complexos continua sendo um desafio, especialmente se esses ambientes possuem requisitos de tempo-real. Muitas ferramentas foram desenvolvidas, mas não existe uma bala de prata e aquelas mais gerais possuem sérios problemas com o requisito de tempo-real. Este trabalho introduz uma nova ferramenta baseada em proxy, usando a teoria de Joint Intentions, que ajuda agentes a serem membros de uma equipe dentro de um ambiente parcialmente observável, dinâmico, estocástico e com requisitos de tempo-real. Os experimentos de validação dessa ferramenta são executados no jogo de computador Unreal Tournament 2004, que fornece um ambiente com tais características. / Although there are many researches in teamwork, the development of agent teams for complex environments still presents many challenges, especially if these environments have real-time requirements. Many tools have been developed, but there is not a silver bullet, and the most general tools have serious problems with real-time requirements. This paper introduces a new proxy-based tool, based on Joint Intentions, to help agents to be a teammate in partilly observable, dynamic and stochastic environments with real-time requirements. The validation experiments for this proxy were performed in the computer game Unreal Tournament 2004, that provides an environment with such features.

Embora haja muitas pesquisas na área de teamwork, o desenvolvimento de times de agentes para ambientes complexos continua sendo um desafio, especialmente se esses ambientes possuem requisitos de tempo-real. Muitas ferramentas foram desenvolvidas, mas não existe uma bala de prata e aquelas mais gerais possuem sérios problemas com o requisito de tempo-real. Este trabalho introduz uma nova ferramenta baseada em proxy, usando a teoria de Joint Intentions, que ajuda agentes a serem membros de uma equipe dentro de um ambiente parcialmente observável, dinâmico, estocástico e com requisitos de tempo-real. Os experimentos de validação dessa ferramenta são executados no jogo de computador Unreal Tournament 2004, que fornece um ambiente com tais características. / Although there are many researches in teamwork, the development of agent teams for complex environments still presents many challenges, especially if these environments have real-time requirements. Many tools have been developed, but there is not a silver bullet, and the most general tools have serious problems with real-time requirements. This paper introduces a new proxy-based tool, based on Joint Intentions, to help agents to be a teammate in partilly observable, dynamic and stochastic environments with real-time requirements. The validation experiments for this proxy were performed in the computer game Unreal Tournament 2004, that provides an environment with such features.

Embora haja muitas pesquisas na área de teamwork, o desenvolvimento de times de agentes para ambientes complexos continua sendo um desafio, especialmente se esses ambientes possuem requisitos de tempo-real. Muitas ferramentas foram desenvolvidas, mas não existe uma bala de prata e aquelas mais gerais possuem sérios problemas com o requisito de tempo-real. Este trabalho introduz uma nova ferramenta baseada em proxy, usando a teoria de Joint Intentions, que ajuda agentes a serem membros de uma equipe dentro de um ambiente parcialmente observável, dinâmico, estocástico e com requisitos de tempo-real. Os experimentos de validação dessa ferramenta são executados no jogo de computador Unreal Tournament 2004, que fornece um ambiente com tais características. / Although there are many researches in teamwork, the development of agent teams for complex environments still presents many challenges, especially if these environments have real-time requirements. Many tools have been developed, but there is not a silver bullet, and the most general tools have serious problems with real-time requirements. This paper introduces a new proxy-based tool, based on Joint Intentions, to help agents to be a teammate in partilly observable, dynamic and stochastic environments with real-time requirements. The validation experiments for this proxy were performed in the computer game Unreal Tournament 2004, that provides an environment with such features.

Submitted by Mariana Dornelles Vargas (marianadv) on 2015-05-27T13:31:09Z No. of bitstreams: 1 adaptacao_conteudos.pdf: 1461923 bytes, checksum: 1272ba554f9ec06de579b9f337be5ff5 (MD5) / Made available in DSpace on 2015-05-27T13:31:09Z (GMT). No. of bitstreams: 1 adaptacao_conteudos.pdf: 1461923 bytes, checksum: 1272ba554f9ec06de579b9f337be5ff5 (MD5) Previous issue date: 2011 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Os objetos de aprendizagem (OAs) têm tomado um papel importante no contexto educacional atual, onde os meios computacionais se tornam cada vez mais presentes em sala de aula impulsionados por instituições governamentais. Dentre as características necessárias para utilização efetiva dos OAs a bibliografia enumera dois pontos principais, o reuso e o compartilhamento. Com o avanço de novos meios computacionais o conceito de reuso tem adquirido novas proporções vislumbrando a utilização do conteúdo educacional disponibilizado pelos OAs em diversos ambientes computacionais. Dentre estes novos ambientes computacionais, os dispositivos móveis e a televisão digital (TVD) estendem o conceito de e-learning criando duas novas vertentes, são elas o m-learning e t-learning respectivamente. Apesar disso, os OAs criados para os ambientes de e-learning dificilmente podem ser reutilizados nos novos ambientes devido às dificuldades de apresentação de conteúdo nos dispositivos móveis e de TVD. Para dar apoio à visualização de OAs nestes novos ambientes o presente trabalho apresenta conceitos relacionados à OAs, a TVD e aos agentes de software para basear a análise do sistema multiagentes desenvolvido. Este sistema tem como principal funcionalidade a adaptação de conteúdo para os ambientes de t-learning e m-learning. A intenção do presente trabalho é apresentar uma arquitetura, utilizando diagramas da Agent UML, que comporte a adaptação de OAs para que possam ser visualizados em múltiplos ambientes e aplicar esta arquitetura a casos de uso voltados à televisão digital. Através do sistema desenvolvido foi possível adaptar múltiplos formatos de conteúdos de OAs para que sejam visualizados nos dispositivos do Sistema Brasileiro de Televisão Digital (SBTVD). / Learning Objects (LO) are assuming an important role in current educational context, where computational resources are a common presence in the classroom due to government incentives. Among the needed characteristics for right using of LOs by teachers and students, two main aspects are highlighted by the literature, the reusing and sharing. With the advances in the new computing environments the concept of reusing has been taking new proportions viewing the utilization of its educational content in several computing environments. Among these new computing environments, the mobile devices and the digital television (DTV) extends the concept of e-learning creating two new strands, they are the m-learning and t-learning respectively. Nevertheless, the LOs created for e-learning environment hardly can be used in the new environments due to the difficulties of presenting on mobile and DTV devices. To help the presenting of LOs in these new environments this work presents concepts related to LOs, DTV and agents based software aiming the analysis of a multiagent system developed. This system has as your main feature the adaptation of content for t-learning and m-learning environments. The intention of this work is present an architecture, using Agent UML (AUML) diagrams, that supports the LOs adaptation for viewing in multiple environments and apply this architecture to digital television use cases. Through the system developed should be possible to adapt multiple LOs content formats to be viewed on the devices of the Brazilian System of Digital Television (SBTVD).

Cette thèse propose un modèle organisationnel et holonique de l'environnement pour la simulation des déplacements de piétons dans des bâtiments. Une foule de piétons peut être considérée comme un système composé d'un grand nombre d'entités en interaction, dont la dynamique globale ne peut se réduire à la somme des comportements de ses composants. La simulation multiniveau fondée sur les modèles multiagents holoniques constitue une approche permettant d'analyser la dynamique de tels systèmes. Elle autorise leur analyse en considérant plusieurs niveaux d'observation (microscopique, mésoscopique et macroscopique) et prend en compte les ressources de calcul disponibles. Dans ces systèmes, l'environnement est considéré comme l'une des parties essentielles. La dynamique des piétons composant la foule est alors clairement distinguée de celle de l'environnement dans lequel ils se déplacent. Un modèle organisationnel décrivant la structure et la dynamique de l'environnement est proposé. L'environnement est structurellement décomposé en zones, sous-zones, etc. Les organisations et les rôles de cet environnement sont projetés dans une société d'agents ayant en charge de simuler la dynamique de l'environnement et les différentes missions qui lui sont classiquement assignées dans les systèmes multiagents. Ce modèle précise également les règles de passage entre deux niveaux d'observation. Ainsi, chaque agent appartenant au modèle de l'environnement tente d'utiliser une approximation des comportements de ses sous-zones afin de limiter la consommation de ressources durant la simulation. La qualité de l'approximation entre ces deux niveaux d'observation est évaluée avec des indicateurs énergétiques. Ils permettent de déterminer si l'agent approxime correctement les comportements des agents associés aux sous-zones. En sus du modèle organisationnel et holonique proposé, nous présentons un modèle concret de la simulation de voyageurs dans un terminal d'aéroport. Ce modèle concret est implanté sur les plateformes JaSIM et Janus. / This work presents a holonic organizational model of the environment for the simulation of pedestrians in buildings. A crowd of pedestrians is considered as a system composed of a large number of interacting entities. The global dynamics of this system cannot be reduced to the sum of the behaviors of its components, Multilevel simulation based on holonic multiagent models is one approach to analyze the dynamics of such systems. It allows their analysis by considering several levels of observation (microscopic, mesoscopic and macroscopic) and the available computing resources. In these systems, the environment is considered as an essential part. The behavior of the crowd is clearly distinguished from the behavior of the environment in which the pedestrians move. An organizational model is proposed to describe the structure and the dynamics of the indoor environment. This environment is structurally divided into areas, sub-areas, etc. Organizations and roles are mapped into a society of agents in charge of simulating the dynamics of the environment and their various missions in multiagent systems. This model also specifies the rules for changing the level of observation dynamically. Thus, each agent belonging to the model of the environment tries to use an approximation of behaviors of its sub-zones, and at the same time to minimize the resource consumption. The quality of the approximation between these two levels is evaluated with energy-based indicators. They help to determine if the agent approximates the behaviors of its sub-agents correctly. In addition to the organizational and holonic model proposed in this work, we present a concrete model of the simulation of passengers in an airport terminal. This concrete model is implemented on the platforms JaSIM and Janus.

Multiagent systems (MAS) are considered a high level abstration for design and engineering of complex systems. Such systems are characterized by organization structures and coordination process more articulated and dynamic. Usually, agents cooperate and coexist in an environment. In addition, there is a general consense in the research community that an environment is an essential part of open and dynamic MAS. However, given the dynamic characteristics present in complex systems, only the use of the agent-based paradigm in the software development does not guarantee the flexibility and scalability of the project ahead of the inevitable changes on requirements. For this reason, the component-based development have been identified as promising in the building of applications with greater ability to adapt to the changes of its requirements. This work aims at developing a component-based framework for engineering open MAS enviroments. The framework is based on the concept of dynamic software composition and supported by a specification that demand mapping from agents concepts to components in order to ensure the flexibility and reusability provided in the component approach. In addition, i) agents and resources are used to compose the software, ii) components are used to compose agents, and iii) objects and aspects are used to develop the functional and non-functional components requirements. The favorable results of this proposal were checked in experiments developed in four case studies. / Sistemas Multiagentes (SMA) são considerados um alto nível de abstração para projeto e engenharia de sistemas complexos, tendo sido caracterizados por estruturas de organização e processos de coordenação cada vez mais articulados e dinâmicos. Geralmente, agentes cooperam e coexistem dentro de um ambiente. Há um consenso geral na comunidade de agentes que os ambientes são parte essencial dos SMA dinâmicos e abertos. Contudo, diante das características dinâmicas presentes em tais SMA, apenas a utilização do paradigma de agentes no desenvolvimento de software não garante a flexibilidade e escalabilidade do projeto frente a inevitáveis mudanças de requisitos do mesmo. Por sua vez, o desenvolvimento baseado em componentes tem sido apontado como promissor na construção de aplicações com maior capacidade de adaptação a mudanças nos seus requisitos. Este trabalho tem como objetivo o desenvolvimento de um arcabouço para engenharia de ambientes de SMA abertos baseado no conceito de composição dinâmica de software. O arcabouço é baseado em uma especificação que procura mapear os conceitos de agentes em componentes para garantir a flexibilidade e reutilização provida na abordagem de componentes. Agentes e recursos são utilizados para compor o software, componentes são utilizados para compor agentes, e objetos e aspectos são utilizados para implementar as características funcionais e não-funcionais dos componentes. Os resultados favoráveis da presente proposta foram verificados nos experimentos realizados em quatro estudos de casos.