Process control is essential in chemical engineering and has diverse applications in automation, manufacturing, scheduling, etc. In this cross-disciplinary work, we shift the domain focus from the control of machines to the control of multiple intelligent agents. Our goal is to improve the optimization problem-solving process, such as optimal regulation of emerging technologies, in a multi-agent system. Achieving that improvement would have potential value both within and outside the chemical engineering community. This work also illustrates the possibility of applying process systems engineering techniques, especially process control, beyond chemical plants. It is very common to observe crowds of individuals solving similar problems with similar information in a largely independent manner. We argue here that the crowds can become more efficient and robust problem-solvers, by partially following the average opinion. This observation runs counter to the widely accepted claim that the wisdom of crowds deteriorates with social influence. The key difference is that individuals are self-interested and hence will reject feedbacks that do not improve their performance. We propose a multi-agent control-theoretic methodology, soft regulation, to model the collective dynamics and compute the degree of social influence, i.e., the level to which one accepts the population feedback, that optimizes the problem-solving performance. Soft regulation is a modeling language for multi-agent sociotechnical systems. The state-space formulation captures the individual learning process (i.e., open loop dynamics) as well as the influence of the population feedback in a straightforward manner. It can model a diverse set of existing multi-agent dynamics. Through numerical analysis and linear algebra, we attempt to understand the role of feedback in multi-agent collective dynamics, thus achieving multi-agent control in sociotechnical systems. Our analysis through mathematical proofs, simulations, and a human subject experiment suggests that intelligent individuals, solving the same problem (or similar problems), could do much better by adaptively adjusting their decisions towards the population average. We even discover that the crowd of human subjects could self-organize into a near-optimal setting. This discovery suggests a new coordination mechanism for enhancing individual decision-making. Potential applications include mobile health, urban planning, and policymaking.

Desde o início do século XXI, vivenciamos uma explosão na produção de informações de diversos tipos, tais como fotos, áudios, vídeos, entre outros. Dentre essas informações, existem aquelas em que a informação pode ser dividida em partes menores, mas que devem ser relacionadas seguindo uma ordem total. Um exemplo deste tipo de informação é um arquivo de vídeo que foi dividido em dez segmentos identificados com números de 1 a 10. Para reproduzir o vídeo original a partir dos segmentos é necessário que seus identificadores estejam ordenados. A estrutura denominada tabela de hash distribuída (DHT) tem sido amplamente utilizada para armazenar, atualizar e recuperar esse tipo de informação de forma eficiente em diversos cenários, como monitoramento de sensores e vídeo sob demanda. Entretanto, a DHT apresenta problemas de escalabilidade quando um membro da estrutura não consegue atender as requisições recebidas, trazendo como consequência a inacessibilidade da informação. Este trabalho apresenta uma arquitetura em camadas denominada MATe, que trata o problema da escalabilidade em dois níveis: estendendo a DHT com a introdução de agentes baseados na utilidade e organizando a quantidade de requisições solicitadas. A primeira camada trata a escalabilidade ao permitir a criação de novos agentes com o objetivo de distribuir as requisições evitando que um deles tenha a escalabilidade comprometida. A segunda camada é composta por grupos de dispositivos organizados de tal forma que somente alguns deles serão escolhidos para fazer requisições. A arquitetura foi implementada para dois cenários onde os problemas de escalabilidade acontecem: (i) monitoramento de sensores; e (ii) vídeo sob demanda. Para ambos cenários, os resultados experimentais mostraram que MATe melhora a escalabilidade quando comparada com as implementações originais da DHT. / Since the beginning of the 21st century, we have experienced an explosive growth in the generation of information, such as photos, audios, videos, among others. Within this information, there are some in which the information can be divided and related following a total order. For example, a video file can be divided into ten segments identified with numbers from 1 to 10. To play the original video from these segments, their identifiers must be fully ordered. A structure called Distributed Hash Table (DHT) has been widely used to efficiently store, update, and retrieve this kind of information in several application domains, such as video on demand and sensor monitoring. However, DHT encounters scalability issues when one of its members fails to answer the requests, resulting in information loss. This work presents MATe, a layered architecture that addresses the problem of scalability on two levels: extending the DHT with the introduction of utility-based agents and organizing the volume of requests. The first layer manages the scalability by allowing the creation of new agents to distribute the requests when one of them has compromised its scalability. The second layer is composed of groups of devices, organized in such a way that only a few of them will be chosen to perform requests. The architecture was implemented in two application scenarios where scalability problems arise: (i) sensor monitoring; and (ii) video on demand. For both scenarios, the experimental results show that MATe improves scalability when compared to original DHT implementations.

Este trabalho apresenta uma arquitetura para fornecer, de forma automática, o reforço de aprendizagem após a aplicação de uma avaliação objetiva dentro de um ambiente de Educação à Distância. Para o desenvolvimento da arquitetura foram usadas tecnologias das áreas de Ontologia e Agentes. A área de Ontologia contribuiu com a construção de materiais de aprendizagem modelados ontologicamente e a construção de ontologias de domínio das disciplinas de Estruturas de Dados I e Análise e Desenvolvimento de Algoritmos I, ministradas nos cursos de Bacharelado em Ciência da Computação e Bacharelado em Sistemas de Informação oferecidos pelo Departamento de Computação da Faculdade de Ciências da UNESP – campus de Bauru - SP. A área de Agentes traz, ao ambiente, agentes de busca capazes de fazer indicações de reforço do conteúdo. A arquitetura foi implementada usando o ambiente de Educação à Distância VirtualCurso, desenvolvido originalmente pelo Laboratório de Tecnologia de Informação Aplicada (LTIA) da UNESP – campus de Bauru - SP. A implementação recebeu o nome de VCurso+. Realizou-se alguns testes com o sistema para verificar se o reforço obteve o efeito esperado na aprendizagem do aluno. / This work presents an architecture to support automatically the apprentice’s reinforcing on environment of Distance Education. In order to develop this architecture technologies from the areas of Ontology & Agents have been used. The Ontology domain has contributed with the building of apprentice’s materials and the construction of domain ontologies for the disciplines of Data Structures I and Analysis & Development of Algorithms I, taught on bachelor's degree of Computer Science and bachelor's degree in Information Systems courses offered at the Computer Science Department of UNESP campus of Bauru SP. The Agents domain has inspered the creation of search agents capable of making indications of content reinforcing. The proposed architecture was implemented in the VirtualCurso enviroment, developped originally at Laboratory of Technology of Information Applied (LTIA) from UNESP campus of Bauru SP and it was called VCurso +. Some tests have been made with the system to verify whether the reinforcing learning proposed by architecture has had the expected effect on the apprentice’s learning.

Esta tese propõe um modelo de adaptação organizacional em Sistemas Multiagentes (SMA) composto por um modelo de organização e um processo de reorganização. O modelo organizacional, chamado de MOISE+, considera três dimensões da organização de um SMA: a estrutura (grupos, papéis e ligações), o funcionamento (planos globais, metas e missões) e as obrigações e permissões dos agentes. A principal característica do modelo MOISE+ é facilitar a mudança organizacional em uma das dimensões sem comprometer as outras. É possível, por exemplo, alterar a estrutura da organização sem alterar o seu funcionamento e vice-versa. O processo de reorganização permite que os próprios agentes realizem a mudança através de quatro etapas: monitoração da organização corrente, projetos de planos de mudança, seleção dos planos e implementação do plano selecionado. A principal característica deste processo é a existência explícita da organização responsável pela reorganização (descrita na notação do modelo MOISE+) permitindo a abertura do processo. Estando a representação deste processo disponível aos agentes, estes podem participar da reorganização, bastando que compreendam a notação do modelo. Este processo foi implementado e avaliado em dois estudos de caso, empresas virtuais e futebol de robôs, nos quais as etapas de seleção e implementação foram enfatizadas, já que existem poucos trabalhos que abordam estas duas etapas do processo de reorganização. / This thesis proposes a reorganization model for Multiagent Systems (MAS). This reorganization model is composed by both an organizational model and a reorganization process. The organization model, called MOISE+, is based on three dimensions: the organizational structure (composed by roles, groups, links, etc.), the organizational functioning (composed by global plans, missions, goals, etc.), and the agents\' obligations and permissions. The MOISE+ model main feature is the independence among these dimensions. It is therefore possible to change the organizational structure without any change in the functioning. The reorganization process enables the agents themselves to perform the reorganization along four phases: monitoring (when to reorganize), design (ways of building a new organization), selection (how to choose a new organization), and implementation (how to change the current running organization). The reorganization process is explicitly described in MOISE+ notation and the agents have access to it. This property enables any agent to participate in the reorganization when he understands MOISE+ notation. The reorganization process is thus open since many agents can joint it. This process was implemented and evaluated in two domains: virtual enterprises and robot soccer. The experiments have focused mainly on the selection and implementation phases since the research on those phases is not very common in the literature.

Sistemas multiagentes muitas vezes apresentam características que os aproximam de sociedades de agentes e, como as humanas, possuem normas e organizações com o objetivo de coordenar as interações entre os seus membros. Coalizão é um tipo de estrutura organizacional temporária, montada com objetivos específicos. A teoria dos jogos estuda formalmente o fenômeno coalizional, se detendo em demonstrações de propriedades e características dessa estrutura. A área de sistemas multiagentes, por outro lado, tem mostrado significativo interesse nas estruturas coalizionais como forma de organizar a cooperação entre os agentes, dedicando-se ao desenvolvimento de algoritmos para formação de coalizões. Esse trabalho apresenta um algoritmo de formação de coalizões para compartilhamento de recursos, denominado heurística ganha-ganha. Definimos um modelo que utiliza a abstração de recursos para representar tanto a posse de habilidades e objetos, quanto para representar os objetivos dos agentes. Um jogo de votação k-ponderado é utilizado para implementar o processo decisório de quais coalizões são válidas e o algoritmo testa iterativamente cada vizinhança de um agente na busca de associações vantajosas. Demonstramos que o algoritmo incrementa monotonicamente o bem-estar da sociedade e converge para uma estrutura coalizional. Também mostramos empiricamente que a heurística é eficiente para compartilhamento de recursos em situações de abundância de recursos, montando coalizões em poucas iterações e com uma quantidade grande de agentes. / Multiagent systems frequently show characteristics that come closer to agent societies and, like the humans ones, have norms and organizations in order to coordinate the interactions of its members. Coalition is a type of temporary organizational structure, assembled with specific goals. Game theory formally studies the coalitional phenomenon focusing in the demonstrations of properties and characteristics of this structure. The area of multiagent systems, on the other hand, has devoted significant interest in coalition structures as a way to organize cooperation between its members, and has been dedicated to the development of algorithms for coalition formation. This dissertation presents an algorithm to coalition formation named win-win heuristic. We define a model which uses the abstraction of resources to represent either, the possession of abilities and objects, or to represent the agents target. A k-weight voting game is used to implement the decision process of what coalitions are worth and an iteratively algorithm tests each agent neighborhood in the pursue of better associations. We demonstrate that the algorithm monotonically increases the society welfare and converges to a coalitional structure. We also show empirically that the heuristic is efficient for resource sharing in situations of availability of resources, building coalitions with few iterations and a large amount of agents.

Nesta tese, propõe-se integrar, de modo sistemático, modelos organizacionais para prover interoperabilidade organizacional em sistemas multiagentes abertos. A interoperabilidade organizacional consiste em permitir que, em uma mesma organização de agentes, coexistam agentes desenvolvidos a partir de modelos organizacionais distintos. Para realizar tal intuito, aplica-se uma abordagem baseada em técnicas da Engenharia Dirigida por Modelos. A abordagem proposta divide-se em três etapas principais: (1) representar a estrutura conceitual de modelos organizacionais através de metamodelos; (2) identificar correspondências entre as estruturas conceituais dos modelos organizacionais utilizando como heurística a idéia de dimensões e padrões de modelagem organizacional ; e (3) combinar os metamodelos dos modelos organizacionais seguindo um algoritmo de junção de modelos baseado em grafos e articulações de grafos. Utilizando a abordagem proposta, quatro modelos organizacionais existentes são integrados: os modelos AGR, STEAM, MOISE+ e ISLANDER. Para ilustrar o uso da integração de metamodelos no problema de interoperabilidade organizacional, especifica-se e implementa-se uma infraestrutura de interoperabilidade organizacional chamada MAORI (Model-driven Architecture for Organizational Interoperability). A infraestrutura MAORI é estruturada em duas camadas: (1) a camada M2M, na qual a integraçãoo de modelos realizada é utilizada para traduzir especificações e estados organizacionais, e (2) a camada ORI, na qual infraestruturas organizacionais são interconectadas. Para testar e validar toda a proposta, a infraestrutura MAORI foi implementada para as infraestruturas organizacionais S-MOISE+ e MADKIT, e uma aplicaçào denominada WritePaper foi desenvolvida. A partir das implementações, alguns experimentos foram realizados. Os resultados obtidos demonstram que a integração de modelos organizacionais proposta, realizada na infraestrutura de interoperabilidade MAORI, efetivamente interconecta infraestruturas organizacionais distintas e permite que, em uma mesma organização de agentes, coexistam agentes desenvolvidos a partir de modelos organizacionais distintos. / This thesis proposes to integrate systematically organizational models for providing organizational interoperability in open multiagent systems. Organizational interoperability consists in enabling the participation, in the same agent organization, of agents developed with different organizational models. For this purpose, it is applied an approach based on techniques from Model-driven Engineering. The approach is divided in three main steps: (1) to represent the conceptual structure of organizational models by means of metamodels; (2) to identify correspondences among the conceptual structures of the organizational models using as heuristics the idea of organizational modeling dimensions and patterns; (3) to combine the metamodels of the organizational models by following a merge algorithm based on graphs and graph articulations. Using the proposed approach, four existing organizational models are integrated: AGR, STEAM, MOISE+ and ISLANDER. To illustrate the use of the integration, an organizational interoperability infrastructure called MAORI (Model-driven Architecture for Organizational Interoperability) is specified and implemented. The MAORI infrastructure is structured in two layers: (1) the M2M layer, in which the integration of models is used to translate the organizational specifications and states, and (2) the ORI layer, in which the organizational infrastructures are interconnected. To test and validate the approach, MAORI infrastructure was implemented for the organizational infrastructures S-MOISE+ and MADKIT, and an application called Write Paper was developed. Using the implementations, some experiments were done. The results show that the organizational model integration, implemented in the MAORI infrastructure, effectively interconnects distinct organizational infrastructures and enables the participation, in the same agent organization, of agents developed with different organizational models.

In this work we present a multiagent Fleet Coordination Problem (FCP). In this formulation, agents seek to minimize the fuel consumed to complete all deliveries while maintaining acceptable on-time delivery performance. Individual vehicles must both (i) bid on the rights to deliver a load of goods from origin to destination in a distributed, cooperative auction and (ii) choose the rate of travel between customer locations. We create two populations of adaptive agents, each to address one of these necessary functions. By training each agent population in separate source domains, we use transfer learning to boost initial performance in the target FCP. This boost removes the need for 300 generations of agent training in the target FCP, though the source problem computation time was less than the computation time for 5 generations in the FCP. / Graduation date: 2012

The decentralized Markov decision process (Dec-POMDP) is a powerful formal model for studying multiagent problems where cooperative, coordinated action is optimal, but each agent acts based on local data alone. Unfortunately, it is known that Dec-POMDPs are fundamentally intractable: they are NEXP-complete in the worst case, and have been empirically observed to be beyond feasible optimal solution.To get around these obstacles, researchers have focused on special classes of the general Dec-POMDP problem, restricting the degree to which agent actions can interact with one another. In some cases, it has been proven that these sorts of structured forms of interaction can in fact reduce worst-case complexity. Where formal proofs have been lacking, empirical observations suggest that this may also be true for other cases, although less is known precisely.This thesis unifies a range of this existing work, extending analysis to establish novel complexity results for some popular restricted-interaction models. We also establish some new results concerning cases for which reduced complexity has been proven, showing correspondences between basic structural features and the potential for dimensionality reduction when employing mathematical programming techniques.As our new complexity results establish that worst-case intractability is more widespread than previously known, we look to new ways of analyzing the potential average-case difficulty of Dec-POMDP instances. As this would be extremely difficult using the tools of traditional complexity theory, we take a more empirical approach. In so doing, we identify new analytical measures that apply to all Dec-POMDPs, whatever their structure. These measures allow us to identify problems that are potentially easier to solve on average, and validate this claim empirically. As we show, the performance of well-known optimal dynamic programming methods correlates with our new measure of difficulty. Finally, we explore the approximate case, showing that our measure works well as a predictor of difficulty there, too, and provides a means of setting algorithm parameters to achieve far more efficient performance.

Efficient organization requires rigorous and systematic information management, which encompasses information processing and decision making. Within the efforts in management science and informatics invested towards advancing the knowledge on, and providing assistance to decision making, this thesis focuses on the conceptualizations and techniques intended to facilitate the identification, evaluation, and selection of decisions during the earliest stages of information systems engineering, whereby the systems of interest are deployed to partly or fully automate various organizational processes, including information processing ones. The overall motivating problem that drove to, and that unites the various contributions presented in this thesis is how to better inform decision making and guide it towards decisions that will increase the quality (as evaluated both by the engineer and the stakeholders) of the information system being engineered. Topics in two key related areas are therefore addressed. First, boundedly rational individuals cannot take engineering decisions by accounting for all information that may be, or actually is available to them. As their information processing abilities are limited and their perception biased, it is necessary to filter the available information to a manageable level, and to bring it to a format that facilitates the rigorous reasoning invested in decision making. Second, it is necessary to provide guidance on how to use the given information in decision making. The first part of this thesis therefore focuses on conceptualizations that facilitate the identification of relevant information and its organization for subsequent analysis, all in the aim of achieving high quality of the system being engineered. In particular, Part I discusses, shows deficiencies, and accordingly revises the conceptual foundations of requirements engineering, a field of information systems engineering that focuses on the identification and analysis of requirements communicated by the stakeholders to the engineer of the system. The novelty of the suggested revision lies primarily in (i) the separation between functional and nonfunctional (i.e., quality) requirements grounded in a foundational ontology, (ii) the introduction of stakeholders' communicated attitudes as important sources of information for the evaluation of alternative requirements engineering decisions, (iii) the reformulation of the so-called ``requirements problem' -- which precisely defines when the requirements engineering effort is successfully completed -- to account for attitudes and nonfunctional requirements, and (iv) the recognition of the importance of defeasible reasoning in the search for a solution to the requirements problem. Acknowledging the importance of defeasible reasoning leads -- in Part II -- to the study of how defeasible reasoning can be incorporated into established decision making processes involved in the identification and analysis of requirements. Novelty in Part II lies mainly in (i) the use of argumentation and justification processes in the modeling and analysis of requirements, (ii) the combined use of design rationale approaches with argumentation and justification, (iii) the recognition that the clarity of arguments is variable (due to ambiguity, vagueness, synonymy, and overgenerality of information going into premises and conclusions in arguments), (iv) the definition of a number of techniques for the detection of unclear information and its clarification, and (v) the use of ``clarity' as a criterion for the discrimination among arguments. Part III shows how the conceptualizations and techniques introduced in Parts I and II are applied within and are relevant to the engineering of information systems, including those that rely on heterogenous and distributed components, as in service-oriented and agent-oriented computing.

La complexité croissante des systèmes industriels en termes d'expansion, d'hétérogénéité et de décentralisation entraine de plus en plus de contraintes dans leur fonctionnement. Ces contraintes ont souvent une composante temporelle. Des techniques, telles que les systèmes multi-agents, sont bien adaptées au caractère hétérogène et décentralisé de ces systèmes, elles sont de plus en plus souvent utilisées dans les applications telles que : contrôle, supervision, simulation, pilotage des travaux existent sur la représentation du temps dans la modélisation et le raisonnement sur les systèmes industriels. Dans le contexte multi-agents, peu de travaux prennent en compte ce facteur temporel explicitement et il est quasiment absent de tous les aspects sociaux de l'agent. Dans cette thèse, nous proposons et définissons une société d'agents temporels qui permet la prise en compte explicite du temps a la fois dans le raisonnement individuel et le raisonnement social au sein de l'agent. L'accent est mis d'une part sur l'interaction et d'autre part sur les différents types de dépendances qui existent entre agents et leur évolution au cours du temps. Ce modelé générique, a été applique à la supervision par reconnaissance de scenarios temporels, domaine dans lequel l'exécution des différentes taches est extrêmement contrainte par le temps. Le système stars est une implémentation du modèle. Il permet une reconnaissance distribuée des scenarios par un ensemble d'agents utilisant un raisonnement temporel.