The main aim of transport stakeholders has always been to transport freight efficiently, as this efficiency contributes to the growth and success of their business. A country like Namibia is no different as the efficiency of transport lies in the effective utilisation of carrier capacity in any direction. Due to the various types of freight, transport operators rarely have the capacity to cover all freight movement requests. This research put the empty runs experienced by most of the Namibian transporters at 33%. Empty runs could however be reduced through collaboration and sharing of capacity among transport stakeholders. Multi-agent systems (MAS) are various individual computer agents that are configured independently to interact with other agents to achieve one goal. These systems have been explored as an approach to achieve collaboration among transporter stakeholders. Taking into consideration the characteristics and requirements of MAS, this research was able to conduct a feasibility of its implementation within Namibia. Concluding with an evaluation of available Multi-agent based systems that could achieve collaboration and reduce empty runs in the Namibian transport environment. / Information Science / M.Tech. (Information Technology)

Thesis (MScEng)--Stellenbosch University, 2013. / ENGLISH ABSTRACT: This thesis documents the research conducted into the control of the feeder subsystem of a Reconfigurable Assembly System (RAS). The research was motivated by a new set of modern manufacturing requirements associated with an aggressive and dynamic global market. The motivation can be more specifically attributed to the need for selective automation, through the installation of reconfigurable systems, in the South African manufacturing industry. The objective of the research was to implement and evaluate Multi-Agent Systems (MASs) and IEC 61499 function block systems as potential control strategies for reconfigurable systems. The control strategies were implemented for the control of the feeder subsystem of an experimental RAS at Stellenbosch University. The subsystem's hardware consisted of a singulation unit with a machine vision camera, part magazines and a six DOF pick-'n-place robot. The structure of the control strategies is based on the ADACOR holonic reference architecture. The mapping of the subsystem holons to the structures of the control strategies is explained. The development and implementation of the control strategies, along with the accompanying lower level software, is described in detail. A system reconfigurability assessment was performed and the results are discussed. The assessment was performed at two levels – the Higher Level Control (HLC) (where the control strategies were implemented) and the low level control and hardware. The assessment was done through four reconfiguration experiments. The evaluation of the HLC was done through both quantitative and qualitative performance measures. The implications of the reconfiguration, involved in each of the respective experiments, on the low level software and hardware are discussed. The experimental results show that agent-based control adds more reconfigurability to the feeder subsystem than IEC 61499 function block control, and that agents have more advantages regarding customizability, convertibility and scalability than IEC 61499 function blocks. Also, the ability of agent-based control to implement reconfiguration changes during subsystem operation makes it more suitable to the case study application. / AFRIKAANSE OPSOMMING: Hierdie tesis dokumenteer die navorsing gedoen in die beheer van die voerder sub-stelsel vir 'n herkonfigureerbare monteringstelsel. Die navorsing was gemotiveer deur „n nuwe stel vereistes vir moderne vervaardiging wat met 'n aggresiewe en dinamiese globale mark geassosieer word. Die motivering kan meer spesifiek toegeskryf word aan die behoefte tot selektiewe outomatisasie, deur middel van die implimentering van herkonfigureerbare stelsels, in the Suid-Afrikaanse vervaardigingsnywerheid. Die doel van die navorsing is om multi-agent stelsels en IEC 61499 funksie-blok stelsels, as potensiële beheerstrategiëe vir herkonfigureerbare stelsels, te implementer en evalueer. Die beheerstrategiëe was geïmplementeer vir die voerder sub-stelsel van „n eksperimentele herkonfigureerbare monteringstelsel by Universiteit Stellenbosch. Die hardeware behels „n skeier-eenheid (singulation unit) met „n masjienvisie kamera, onderdeelmagasyne en „n ses-vryheidsgraad gearktikuleerde optel-en-plaas robot. Die struktuur van die beheerstrategiëe is gebaseer op die ADACOR holoniese verwysingsargitektuur. Die afbeelding van die sub-stelsel holons na die struktuur van die beheerstrategiëe word verduidelik. Die ontwikkeling en implementering van die beheerstrategiëe, asook die gepaardgaande laer-vlak programmatuur, word in detail beskryf. Die stelsel se herkonfigureerbaarheid was geassesseer en die resultate daarvan word bespreek. Die assessering was op twee vlakke gedoen – die hoër-vlak beheer (waar die beheerstrategiëe geimplementeer was) en die lae-vlak beheer en hardeware. Die assessering was gedoen deur middel van vier herkonfigurasie eksperimente. Die hoër-vlak beheer was geëvalueer deur beide kwalitatiewe en kwantitatiewe metings. Die implikasies van die herkonfigurasie, betrokke by die onderskeie eksperimente, op die lae-vlak beheer en hardeware word beskryf. Die eksperimentele resultate wys dat agent-baseerde beheer meer herkonfigureerbaarheid tot die voerder sub-stelsel toevoeg as IEC 61499 funksie-blok beheer. Dit is geïdentifiseer dat agente meer voordele inhou ten opsigte van aanpasbaarheid, skakelbaarheid en skaalbaarheid as IEC funksie-blokke. Agent-baseerde beheer laat ook toe dat herkonfigurasieveranderinge tydens sub-stelsel werking geïmplimenteer kan word – dus is dit meer geskik vir aanwending in die gevallestudie.

Algorithmically designed reward functions can influence groups of learning agents toward measurable desired sequential joint behaviours. Influencing learning agents toward desirable behaviours is non-trivial due to the difficulties of assigning credit for global success to the deserving agents and of inducing coordination. Quantifying joint behaviours lets us identify global success by ranking some behaviours as more desirable than others. We propose a real-valued metric for turn-taking, demonstrating how to measure one sequential joint behaviour. We describe how to identify the presence of turn-taking in simulation results and we calculate the quantity of turn-taking that could be observed between independent random agents. We demonstrate our turn-taking metric by reinterpreting previous work on turn-taking in emergent communication and by analysing a recorded human conversation. Given a metric, we can explore the space of reward functions and identify those reward functions that result in global success in groups of learning agents. We describe 'medium access games' as a model for human and machine communication and we present simulation results for an extensive range of reward functions for pairs of Q-learning agents. We use the Nash equilibria of medium access games to develop predictors for determining which reward functions result in turn-taking. Having demonstrated the predictive power of Nash equilibria for turn-taking in medium access games, we focus on synthesis of reward functions for stochastic games that result in arbitrary desirable Nash equilibria. Our method constructs a reward function such that a particular joint behaviour is the unique Nash equilibrium of a stochastic game, provided that such a reward function exists. This method builds on techniques for designing rewards for Markov decision processes and for normal form games. We explain our reward design methods in detail and formally prove that they are correct.

Les environnements domotiques sont des environnements ubiquitaires dans lesquels des équipements domestiques, disséminés dans une habitation, fournissent des services utilisables à distance au travers d'un réseau. Des systèmes domotiques sont proposés pour permettre aux utilisateurs de contrôler les équipements en fonction de leurs besoins.Idéalement, ces systèmes orchestrent l'exécution des services fournis par les équipements pour réaliser des services complexes. Mieux encore, ces systèmes doivent s'adapter à la variété des environnements en termes d'équipements et des besoins des utilisateurs. Ils doivent également pouvoir s'adapter dynamiquement, si possible de manière autonome, au changement de leur contexte d'exécution (apparition ou disparition d'un équipement, évolution des besoins).Dans cette thèse, nous apportons une réponse à cette problématique avec SAASHA, un système domotique multi-agents à base de composants. La combinaison de ses deux paradigmes permet de gérer l'adaptation à trois niveaux : présentation (interfaces utilisateur),organisation (architecture du système) et comportement (architecture interne des agents).Les agents perçoivent le contexte et ses changements. Les utilisateurs se voient proposer une vue dynamique du contexte leur permettant de définir des scénarios personnalisés sous forme de règles. Les agents se répartissent les rôles pour réaliser les scénarios. Ils modifient dynamiquement leur architecture interne grâce à la génération, au déploiement et à l'assemblage de composants pour se doter de nouveaux comportements de contrôle des équipements et des scénarios. Les agents collaborent ainsi pour exécuter les scénarios. En cas de changement, ces trois niveaux d'adaptation sont mis en oeuvre de manière dynamique et autonome pour maintenir la continuité de service. Un prototype de SAASHA, basé sur les standards industriels UPnP et OSGi, a été développé pour évaluer la faisabilité de notre proposition. / Home automation environments are ubiquitous environments where domestic devices, scattered throughout a home, provide services that can be used remotely over a network. Home automation systems are proposed to enable the users of controlling the devices according to their needs. Ideally, these systems orchestrate the execution of the services provided by the devices to achieve complex services. Even more, these systems must adapt to the variety of environments in terms of devices and users needs. They must also be able to adapt dynamically, if possible in an autonomous manner, to the changes of their execution context (appearance or disappearance of a device, changing needs).In this thesis, we provide an answer to this problematic with SAASHA, a multi-agent home automation system based on components. The combination of these two paradigms enables managing the adaptation on three levels: presentation (user interface), organization (system architecture) and behavior (internal architecture of agents). The agents perceive their context and its changes. The Users are offered a dynamic view of the context allowing them to define custom scenarios as rules. The agents divide the roles among them to realize the scenarios. They modify dynamically their internal architecture throughout the generation, deployment and assembly of components to adopt new device control behaviors and scenarios. The agents collaborate to execute the scenarios. In case of a change, these three levels of adaptation are updated dynamically and autonomously to maintain the service continuity. A SAASHA prototype, based on UPnP and OSGi industry standards, has been developed to assess the feasibility of our proposal.

Cette recherche se place dans le domaine de l’ingénierie industrielle, plus exactement aux premières étapes du processus d´innovation (Fuzzy Front-End). Aujourd´hui dans un monde globalisé, où la technologie est à la portée de tous, les entreprises cherchent de nouvelles techniques permettant de se distinguer de la concurrence, et offrent de nouveaux et meilleurs produits ou services ; une manière de conserver leurs avantages comparatifs ou d’en acquérir de nouveaux, pour des clients toujours plus exigeants, et dont les besoins changent constamment. Partant de ce constat, l’innovation se définit comme un processus long et continu, au sein duquel s’utilisent différents modèles. Du fait de la complexité du processus, il n’existe pas de modèle unique réunissant tous les besoins des clients. C’est pourquoi le processus d’innovation, implique différents acteurs de l’entreprise et de son environnement. Ces acteurs, dont l’importance varie, sont souvent d’origines diverses et possèdent des compétences variées. Ils ont des besoins différents, et sont soumis à des contraintes de travail différentes (ressources technologiques et financières, réglementations, etc...). Il paraît ainsi essentiel pour l’entreprise de créer un modèle unique, capable d’identifier les besoins de toutes les parties prenantes, avant d’entamer le processus de développement de nouveaux produits. C’est pour l’entreprise une manière de renforcer le processus d’innovation, et de réduire le taux d´échec lors du lancement de nouveaux produits. Ainsi l’objectif de cette étude est de proposer une nouvelle méthode d’analyse des impacts de l’innovation (A.I), afin de permettre aux entreprises d’identifier les besoins dynamiques des acteurs. Celles-ci pourront ainsi mieux définir leurs nouveaux produits et s’adapter plus facilement aux évolutions du marché. Cet outil fait partie d´une méthodologie de gestion des projets innovants, et doit être utilisé avant la définition des caractéristiques fonctionnelles du nouveau produit. Ceci nous ramène aux premières phases du développement, avant la conception du produit, qui comprennent toutes les étapes depuis la définition des besoins et opportunités pour l´innovation, jusqu’à la prise de décision pour le développement du nouveau produit, appelée Fuzzy Front-End. L’apport théorique de notre travail : Il n´existe pas de définition des impacts de l’innovation (I_n), nous avons donc fait un travail bibliographique sur le concept d’impact. Celui-ci a permis d’explorer les différentes définitions d´impact au sein de différentes disciplines (ingénierie, écologie, psychologie, etc.). Nous avons proposé de cette manière, notre propre définition de l’impact adaptée au contexte de l´innovation. Cette recherche bibliographique, nous a aussi permis d’identifier les différents types d´impacts et de justifier l´utilisation de différentes méthodologies. La contribution méthodologique : l’approche proposée pour l´analyse des impacts de l´innovation (A.I) se décompose en deux étapes. La première vise à modéliser l’« écosystème d´innovation », sur la base des relations, de l´importance et des besoins individuels de chacun des acteurs face au nouveau produit. Ceci grâce à une Analyse des Besoins Dynamiques (A.B.D), qui permettra d´obtenir les besoins agrégés permettant de représenter le système complet, sans considérer les possibles perturbations du milieu extérieur ou les impacts de la même innovation. Pour bien identifier et décrire les besoins des acteurs, nous avons basé la méthodologie sur l´observation de leurs activités. Ainsi, nous avons pu identifier les différents besoins des acteurs de l´innovation sur la base de l’analyse RAR (Ressources, Activités, Résultat), ensuite nous avons classifié les besoins grâce au modèle de KANO et ses améliorations réalisées par Tontini. Une méthode originale d’agrégation des besoins de tous les acteurs est proposée.[...] / This research belongs to the industrial engineering field, focusing on the first stages of the innovation process (Fuzzy Front-End). In a globalized world, where the technology is near to everyone, enterprises are looking for new technologies to be different from their competitors, offering new and better products / services as a strategy to keep or acquire new competitive advantages. Moreover, innovation is a large and continuous process, where different management models are used. But, because of the complexity of the process, no general model is suitable for situations and projects. Thus, the innovation process involves different stakeholders from the company and its environment. These individuals come from different origins (field of knowledge), have different needs, abilities, importance and work under different restrictions (technological or financial resources, work rules, etc.). Consequently, it is essential for the enterprise to identify all the needs of the parts involved in the products. And, a model is required in order to identify the stakeholder’s needs before starting design tasks. This model aims at more robustness in the new product development process and at the reduction of the failure rate when launching new products. Thus, the objective of this PhD thesis is to propose a new method to analyze the impact of the innovation, allowing companies to identify the dynamic actors' needs, to improve the definition of their new products and to adapt to the possible changes of the market. This tool is part of an innovative projects management methodology, used before the elaboration of the functional specifications of a new product. This locates the research in the first stages of development called Fuzzy Front-End: before the genesis of the product. Fuzzy front ends integrate among others the opportunity / needs identification for innovation to the development of the new product decision taking. The proposed approach for the innovation impact analysis (A.I) consists of two stages: the first one intends to model an innovation ecosystem, integrating the relations, the importance of each stakeholder and their individual necessities in regard to the new product. Through a Dynamic Needs Analysis (A.B.D) it will be possible to get the global needs, which represent the whole system, without considering the disturbances of the environment or the impact of the same innovation. Thus, the second stage of the model, through the generation of innovation impact scenarios (S.I.I), evaluates the impact of the innovation the new product may face at its market launching. Due to a new product emergence changes emerge: actors’ performance, their interrelations, the importance and their own needs about the product [...]

[pt] A análise de requisitos não funcionais (RNF) é um desafio e vem sendo explorado na literatura científica. Tal iniciativa deve-se ao fato da existência do problema de se verificar o uso das operacionalizações desse tipo de requisito no software construído. Nessa tese apresenta-se um método, com técnicas e ferramentas de apoio, que analisam se um software está em conformidade com padrões de RNFs estabelecidos em catálogo como alternativa para o problema de análise de RNF. A estratégia adotada nessa tese utiliza agentes autônomos para análise de conformidade de software em relação a operacionalizações de RNF. Para isso, utiliza uma base de conhecimentos de padrões persistidos em um catálogo. Os resultados parciais são indicativos de que a proposta de solução é aplicável. A avaliação da validade dá-se por demonstração de que um método parcialmente automatizado é eficaz na identificação de conformidades. Um diferencial do trabalho apresentado é a ligação dos RNFs a sua efetiva implementação. Para demonstração da tese aplicou-se e customizou-se uma técnica de padrões de RNFs, baseados em orientação a metas, em estudos de caso de exemplos do cotidiano prático de software. Apresentamos também a construção de um framework de agentes, que operam sob notações XML para identificar conformidades de software em relação a um catálogo de RNF. / [en] The analysis of non-functional requirements (NFR) is a challenge and has been explored in the literature. This initiative is due to the fact of the existence of the problem of analysis the use of the NFRs operationalization in software. In this thesis we present a method, with supporting tools and techniques, that checks, if a software complies with standards of non-functional requirements as described in a catalog, as an alternative to the NFR analysis problem. The strategy adopted in this thesis uses autonomous agents to check software compliance regarding the operationalization of an NFR, by using a knowledge base of patterns persisted in a catalog. Initial results show that the proposed solution is applicable. The evaluation of the validity is given by the demonstration that a partially automated method is effective in identifying compliance. This work differs form others by linking NFRs to their effective implementation. A method based on patterns NFRs was used in common software, as to show the application of the proposed strategy. An agent based framework, working with XML descriptions, for checking software compliance with respect to a NFR catalog was built.

[pt] Os benefícios trazidos pela aplicação de Sistemas Multi-Agentes (SMA) são diversos. Através da computação paralela, agentes podem trabalhar em conjunto para explorar melhor a estrutura descentralizada de uma determinada tarefa e acelerar sua conclusão. Além disso, agentes também podem trocar experiências se comunicando, fornecer alto grau de escalabilidade, através da inclusão de novos agentes quando necessário, e ainda fazer com que agentes assumam as atividades de outros agentes em casos de falha. Vários modelos de agentes desenvolvidos até o momento usam o aprendizado por reforço como algoritmo base no processo de aprendizado. Quando o agente está inserido em ambientes pequenos ou discretos, os resultados obtidos com o uso de métodos como Q-learning são satisfatórios. No entanto, quando o ambiente é grande ou contínuo, o uso de métodos de aprendizado por reforço torna-se inviável, devido à grande dimensão do espaço de estados. Nos SMA, este problema é consideravelmente maior, já que a memória necessária passa a crescer exponencialmente com a quantidade de agentes envolvidos na aplicação. Esta tese teve como finalidade o desenvolvimento de um novo modelo de aprendizado autônomo para Sistemas Multi-Agentes (SMA) visando superar estas limitações. O trabalho foi realizado em três etapas principais: levantamento bibliográfico, seleção e implementação do modelo proposto, e desenvolvimento de estudo de casos. O levantamento bibliográfico contemplou o estudo de agentes inteligentes e Sistemas Multi-Agentes, buscando identificar as propriedades e limitações dos algoritmos já desenvolvidos, as aplicações existentes, e as características desejadas em um SMA. A seleção e utilização de um modelo neuro-fuzzy hierárquico da família RL-NFH foi motivada especialmente pela importância de se estender a autonomia e aprendizado de agentes através do quesito inteligência, e pela sua capacidade de superar limitações presentes em algoritmos de aprendizado por reforço tradicionais. Inicialmente, ao modelo anterior foram adicionados os conceitos de satisfatoriedade e não-dominação, com a finalidade de acelerar o processo de aprendizado do algoritmo. Em seguida, o novo modelo multi-agente foi criado, viabilizando o desenvolvimento de aplicações de natureza tanto cooperativa como competitiva, com múltiplos agentes. Os estudos de caso contemplaram situações distintas de cooperação e competição entre agentes autônomos. Foram implementadas três aplicações distintas: uma aplicação benckmark do jogo da presa-predador (Pursuit-Game); um leilão energia elétrica, em que os fornecedores de energia fazem ofertas para atender à previsão de demanda em um período de tempo determinado; e uma aplicação na área de gerenciamento de projetos, onde agentes inteligentes são criados com o objetivo de fornecer estimativas de duração de atividades e automatizar alguns processos realizados pelo Gerente de Projetos. Em todos os Estudos de Caso, os resultados foram comparados com técnicas convencionais e/ou com o desempenho de outros Sistemas Multi-Agente. Os resultados alcançados pelo novo modelo se mostraram promissores. Os testes evidenciaram que o modelo teve a capacidade de coordenar as ações entre agentes totalmente autônomos em diferentes situações e ambientes. Além disso, o novo modelo mostrou-se genérico e flexível, podendo ser usado no futuro em outras aplicações envolvendo múltiplos agentes. / [en] There are several benefits provided by Multi-Agent Systems (MAS). Through parallel computing, agents can work together to better explore the decentralized structure of a given task and speed up its completion. In addition, agents can also exchange knowledge through communication, provide scalability by adding new agents when appropriate, and replace troubled agents in cases of failures. A great number of existing agent models is based on reinforcement learning algorithms for learning. When the agent works in small or discrete environments, the results obtained with methods such as Qlearning are satisfactory. However, when the environment is large or continuous reinforcement learning methods become unfeasible due to the large state space. In MAS, this problem is considerably greater, since the required memory begins to grow exponentially with the number of agents involved in the application. The main objective of this thesis is to develop a new model of autonomous learning for multi-agents in order to overcome these limitations. The study consisted of three main stages: literature review, new model development and implementation, and case studies. Literature review included the study of intelligent agents and Multi-Agent Systems, seeking to identify the properties and limitations of the algorithms already developed, existing applications, and desired features in the new MAS. The choice of a neuro-fuzzy hierarchical model of the family RL-NFH as a basis was especially motivated by the importance of extending the autonomy and learning of the agents through intelligence. And also, because of its capacity to overcome some of the limitations present in traditional reinforcement learning algorithms. Initially, the concepts of satisficing and non-domination were incorporated into the previous model to accelerate the learning algorithm. Then, the new multi-agent model was elaborated and implemented, enabling the development of cooperative and competitive applications, with multiple agents. Case studies have covered different situations of cooperation and competition between autonomous agents. Three applications were considered: the Pursuit-Game benckmark game, an electricity auction, where energy suppliers make offers to meet forecast demand in a given period of time, and an application in project management area, where intelligent agents are created to provide activity duration estimates and to automate some processes done usually by the Project Manager. In all case studies, results were compared with conventional techniques and/or the performance of other MAS. The results achieved by the new model are encouraging. The tests showed that the new system has the capacity to coordinate actions between fully autonomous agents in different situations and environments. Moreover, the new model is strongly generic and flexible. Due to these properties, it can be used in future in several other applications involving multiple agents.

[pt] A tecnologia de agentes tem ganho destaque, tanto na academia quanto na indústria, no que se refere ao desenvolvimento de sistemas distribuídos. Linguagens de modelagem e processos de desenvolvimento foram criados para documentar e formalizar o desenvolvimento de sistemas baseados em agentes. Entretanto, as linguagens de modelagem existentes ou se baseiam em noções de uma tecnologia de desenvolvimento anterior (orientação a objetos) ou não possuem artefatos (modelos) que denotam as características de agência existentes no sistema. Esta tese propõe uma linguagem de modelagem para sistemas baseados em agentes, a LM-SMA, que gera artefatos (modelos) que mostram a modelagem de aspectos de agência, tais como adaptação, aprendizado, interação e autonomia. A LM-SMA ainda possui artefatos que permitem a modelagem da parte do sistema que não é composta por agentes, por meio de ontologias. / [en] The agent technology is gaining acceptance, both in academy and industry, with regards to distributed systems development. Modeling languages and development processes were created to formalize the development of agent based systems. Nevertheless, existing modeling languages are either based on previous development methods (object oriented) or they do not have artifacts (models) that show the agency characteristics that exist in a system. This thesis proposes a modeling language, for agent based systems, that generates artifacts that model agency aspects, such as adaptation, learning, interaction and autonomy. The language has artifacts that allow the modeling of the non- agent part of an agent based system, using ontology.

[pt] Governança trata heterogeneidade, autonomia e diversidade de interesses entre diferentes agentes em um sistema multi-agente (SMA), estabelecendo um conjunto de normas. A maioria dos mecanismos de aplicação de leis verifica a violação de normas apenas sob o aspecto da interação entre os agentes. Entretanto, em SMAs com a presença de agentes heterogêneos e independentemente projetados, existirão mensagens confidenciais, que só serão percebidas por seus remetentes e destinatários, e execução de ações que só serão observáveis pelos agentes que as executam ou por agentes que sofrem as conseqüências da execução dessas ações. Este trabalho apresenta uma abordagem que implementa um mecanismo de governança em SMA baseado em testemunhos. Agentes podem testemunhar fatos que estão relacionados à violação de normas as quais eles têm conhecimento. Este mecanismo é composto por três sub-sistemas: Reputação, julgamento e sanção. O foco deste trabalho está apenas no sub-sistema de julgamento, responsável por receber os testemunhos e prover decisões apontando se o agente realmente violou uma norma. Será mostrada a arquitetura deste sub-sistema e um processo de julgamento genérico. Finalmente, será ilustrada a utilização deste mecanismo através de um estudo de caso. / [en] Governance copes with the heterogeneity, autonomy and diversity of interests among different agents in a multi-agent system (MAS) by establishing a set of norms. Most of the governance enforcement mechanisms usually check norm violations from the point of view of interaction protocols. However, in MAS, with the presence of heterogeneous and independently designed agents, there will be private messages, that will only be perceived by their senders and receivers, and execution of actions that will only be observed by the agents that execute them or by the agents that suffers from their consequences. This work presents a governance mechanism for MAS based on testimonies. Agents can witness facts that are related to norm violations witch they are aware of. The proposed mechanism is composed by three sub-systems: Reputation, judgment and sanction. This work focuses only in the judgment sub- system, witch is responsible for receiving testimonies and providing a decision, pointing out if an agent has really violated a norm. The judgment sub-system architecture and a generic judgment process will be presented. Finally, the use of this mechanism will be exemplified by a case study.

[pt] As linguagens de modelagem de agentes visam representar o sistema e seus agentes através de diagramas, os quais permitem explicitar seus objetivos, planos e ações. Mesmo as linguagens fornecendo todos esses mecanismos de representação, alguns sistemas ainda não podem ser expressos de maneira ideal, pois tais linguagens visam representar sistemas inteiros e não uma linha de produção com várias instâncias. O método proposto visa num primeiro momento determinar o mapeamento dos pontos de flexibilização em agentes de software. Planos e ações podem ser flexibilizados por apresentarem características de variabilidade. A variabilidade pode ser interpretada de duas formas: a variabilidade de planos que possibilita que um agente tenha vários planos distintos, acarretando aplicações distintas em função do seu plano; e a variabilidade de ações que possibilita que ações sejam executadas de maneiras distintas o que também acarreta aplicações distintas. Com isso podem ser observadas ações e planos abstratos que serão herdados por outras ações e planos concretos e que definirão uma instância para novas aplicações. Para isto o método proposto utiliza diagramas, tags e documentação para guiar a instanciação dos planos e ações. Esses agentes gerarão instâncias de aplicações devido seus planos e ações os quais determinam uma linha de produto de software e conseqüentemente a idéia de framework. Esta abordagem traz para o mundo dos agentes as vantagens observadas com o uso destas técnicas em orientação a objeto. Além disso, as linguagens atuais poderão utilizar concomitantemente tal método. Os benefícios da solução serão apresentados através de estudos de caso. / [en] The current agents´ modeling languages aim at representing the system and theirs agents in a clear way by diagrams, which permit shows theirs goals, plans and actions. Even with all provided by the language, some systems can´t be represented a correct form yet, because the current agents´ modeling languages represent only a whole system and not a product line. In this way the method proposed tries to determine a mapping of the flexibility points into software agents. The flexibility points into agents are defined to theirs plans and actions. Such points can be flexible if show a variability characteristic. The variability is presented by two points of view: the variability of plans and variability of action, where the variability of plans can enable many distinct plan for an agent, in other words, accept distinct applications for each one of theirs plans; and the variability of actions that enable the actions are executed in distinct form that result in distinct applications. With that, we can see an abstract actions and abstract plans will be inherited by the concrete actions and concrete plans, and that in the future will define new applications. For this the method uses the diagrams, tags and documentation. The documentation is used like a guide in a plan instance and action instance. After, these agents can generate the instance of new applications by yours owner plans and actions that´s determine the software product line and consequently it is possible to use the framework idea. With this approach it is possible to introduce into agent world all the advantage of frameworks and product lines, techniques that are traditionally used in object orientation. Moreover, the approach can be used concomitantly with current agents´ modeling languages. The benefits of the approach will be shown in more details through a case study.