Made available in DSpace on 2016-08-17T14:52:45Z (GMT). No. of bitstreams: 1 Yonara Magalhaes.pdf: 3488130 bytes, checksum: e183eccb37268ce608c84048d98d467a (MD5) Previous issue date: 2003-01-31 / Cooperative process of teach-learning at a distance of the society of intelligent agents of environment MATHNET. It is presented architecture of the society of agents of the MATHNET that to provide the process with teach-learning. It is in exchange for distinguished the Tutor Agent and the model of communication, based messages and that it uses the FIPA-ACL Agent Communication Language, between the Tutor Agent and excessively the participant agents of this society. The main characteristics and the paper (responsibilities) of the Tutor Agent describe. Its conceptual model is constructed. Its main Cases Use, interactions with several agents of this architecture and the protocols of communication of the Tutor Agent with the other agents are shaped. The Tutor Agent creates itself using a tool of construction of agents. One implements the Tutor Agent and its Cases Use to inside demonstrate to its relevance and importance of the system and the model of communication. / Processo de ensino-aprendizagem cooperativo à distância da sociedade de agentes inteligentes do ambiente MATHNET. Apresenta-se a arquitetura da sociedade de agentes do MATHNET que provê o processo de ensino-aprendizagem. Destaca-se o Agente Tutor e o modelo de comunicação, baseado em troca de mensagens e que utiliza a Linguagem de Comunicação FIPA-ACL, entre o Agente Tutor e os demais agentes participantes dessa sociedade. Descrevem-se as principais características e o papel (responsabilidades) do Agente Tutor. Constrói-se o seu modelo conceitual. Modelam-se os seus principais Casos de Uso, as interações com os demais agentes dessa arquitetura e os protocolos de comunicação do Agente Tutor com os outros agentes. Cria-se o Agente Tutor utilizando para isto uma ferramenta de construção de agentes. Implementa-se o Agente Tutor e seus Casos de Uso para demonstrar sua relevância e importância dentro do sistema e o modelo de comunicação aqui proposto.

Durant les dernières années, la taille des systèmes ainsi que leur complexité ont pas mal évolué, entrainant le besoin d'approches distribuées pour la commande et l'aide à la décision. Cette thèse porte sur la résolution d'un problème incluant une commande distribuée et une aide à la décision, l'allocation dynamique de ressource dans un réseau.Pour résoudre ce problème, nous avons étudié un algorithme basé sur un consensus qui ne nécessite pas de calcul centralisé, et qui soit capable de traiter des applications modélisées par des systèmes dynamiques ou par des fonctions sans mémoires. La principale contribution de ce travail de thèse est d'avoir prouvé, en utilisant des outils issus de la théorie des graphes etl'analyse de la passivité, que le contrôleur atteint la solution optimale de façon asymptotique, sans obligation d'avoir une information complète.Afin d'illustrer la pertinence de notre résultat principal, plusieurs applications en ingénierie ont été étudiées, incluant la commande distribuée pour l'économie d'énergie dans des bâtiments intelligents, la gestion des clients dans un environnement de "smart grids", et le développement d'une méthode exacte d'optimisation distribuée pour un problème d'allocation de ressources soumis à des contraintes sur les bornes inférieures.Enfin, nous étudions les techniques d'allocation de ressources basées sur les modèlesde dynamique de populations. Pour les rendre distribuées, nous introduisons le concept dedynamique de populations "pas bien mélangées". Nous montrons que ces dynamiques peuventêtre utilisées pour des structures d'informations contraintes. Même si les dynamiquesde populations "pas bien mélangées" utilisent des informations partielles, ellesconservent des propriétés similaires aux dynamiques classiques qui utilisent desinformations complètes. Plus spécifiquement, la conservation de masse et la convergencevers l'équilibre de Nash sont prouvées. / Since the complexity and scale of systems have been growing in the last years, distributed approaches for control and decision making are becoming more prevalent. This dissertation focuses on an important problem involving distributed control and decision making, the dynamic resource allocation in a network. To address this problem, we explore a consensus--based algorithm that does not require any centralized computation, and that is capable to deal with applications modeled either by dynamical systems or by memoryless functions. The main contribution of our research is to prove, by means of graph theoretical tools and passivity analysis, that the proposed controller asymptotically reaches an optimal solution without the need of full information. In order to illustrate the relevance of our main result, we address several engineering applications including: distributed control for energy saving in smart buildings, management of the customers of an aggregating entity in a smart grid environment, and development of an exact distributed optimization method that deals with resource allocation problems subject to lower--bound constraints. Finally, we explore resource allocation techniques based on classic population dynamics models. In order to make them distributed, we introduce the concept of non--well--mixed population dynamics. We show that these dynamics are capable to deal with constrained information structures that are characterized by non--complete graphs. Although the proposed non--well--mixed population dynamics use partial information, they preserve similar properties of their classic counterpart, which uses full information. Specifically, we prove mass conservation and convergence to Nash equilibrium. / Dado que la complejidad y la escala de los sistemas sehan ido incrementando en los últimos años, las técnicas centralizadas de control y toma de decisiones están siendo reemplazadas por métodos distribuidos. Esta tesis se centra en un importante problema que involucra control y toma de decisiones distribuidas: la asignación dinámica de recursos en redes. Para abordar este problema, exploramos un algoritmo basado en consenso que no requiere computación centralizada, y que puede ser usado en aplicaciones modeladas ya sea por sistemas dinámicos o funciones sin memoria. La principal contribución de esta tesis es probar, por medio de teoría de grafos y pasividad, que el algoritmo propuesto alcanza asintóticamente una solución óptima sin la necesidad de usar información completa. Para ilustrar la relevancia del resultado principal de esta disertación, abordamos varias aplicaciones en ingeniería,incluyendo: el control distribuido en edificios inteligentes orientado a la eficiencia energética, la gestión de los clientes de un agregador en una red inteligente en la que se aplican estrategias de respuesta de la demanda, y el desarrollo de un método de optimización exacto que permite incluir restricciones de límite inferior. Finalmente, se exploran otras técnicas de asignación derecursos inspiradas en modelos de dinámicas poblacionales. Se introduce el concepto de poblaciones no—bien—mezcladas, y se muestra que las dinámicas asociadas a este tipo de poblaciones cuentan con una estructura de información local, caracterizada por grafos que no son completos. A pesar de que las dinámicas propuestas usan información parcial, ellas preservan características similares a las dinámicas poblacionales clásicas que usan información completa.

Zum Verständnis menschlichen sozialen, administrativen und wirtschaftlichen Verhaltens, das als Spiel bzw. strategische Interaktion aufgefasst werden kann, reichen die rein analytischen Methoden nicht aus. Es ist nötig, Daten menschlichen strategischen Verhaltens zu sammeln. Basierend auf Daten lässt sich solches Verhalten modellieren, simulieren bzw. vorhersagen. Der theoretische Teil der Zielsetzung wird über praxisorientierte Konzeptualisierung strategischer Interaktion realer Agenten - Menschen und Maschinen - und gegenseitige Integration der Konzepte aus Spieltheorie und Multiagentensysteme erreicht, die über die bisherigen Ansätze hinausgehen. Der praktische Teil besteht darin, ein allgemein verwendbares System zu entwerfen, das strategische Interaktionen zwischen realen Agenten mit maximalen wissenschaftlichen Nutzen durchführen kann. Die tatsächliche Implementation ist eines der Ergebnisse der Arbeit. Ähnliche vorhandene Systeme sind GDL-Server (für Maschinen) [Genesereth u.a., 2005] und z-Tree (für Menschen) [Fischbacher, 2007]. Die Arbeit ist in drei Bereiche unterteilt - (1) Entwicklung von Sprachen für die Beschreibung eines Spiels, (2) ein auf diesen Sprachen basierendes Softwaresystem und (3) eine Offline-Analyse der u.a. mit dem System bereits gesammelten Daten als Beitrag zur Möglichkeiten der Verhaltensbeschreibung. Die Innovation dieser Arbeit besteht nicht nur darin ,einzelne Bereiche mit einander zu kombinieren, sondern auch Fortschritte auf jedem Bereich für sich allein zu erreichen. Im Bereich der Spielbeschreibungssprachen, werden zwei Sprachen - PNSI und SIDL - vorgeschlagen, die beide Spiele bei imperfekter Information in diskreter Zeit definieren können. Dies ist ein Fortschritt gegenüber der bisherigen Sprachen wie Gala und GDL. Speziell die auf Petrinetzen basierende Sprache PNSI kann gleichermaßen für Gameserver und für spieltheoretische Algorithmen von z.B. GAMBIT verwendet werden. Das entwickelte System FRAMASI basiert auf JADE [Bellifemine u.a., 2001] und ist den bisherigen Client-Server-Lösungen durch Vorteile der Multiagentensysteme voraus. Mit dem entstandenen System wurde bereits ein Experiment entsprechend den Standards der experimentellen Spieltheorie durchgeführt und somit die Praxistauglichkeit nachgewiesen. Das Experiment hatte als Ziel, Daten zur menschlichen Unvorhersagbarkeit und zur Vorhersagefähigkeit anderer zu liefen. Dafür wurden Varianten von \"Knobeln\" verwendet. Die Daten dieses Experiments sowie eines Experiments einer externen Arbeitsgruppe mit ähnlicher Motivation wurden mit Hilfe von Datamining analysiert. Dabei wurden die in der Literatur berichteten Gesetzmäßigkeiten des Verhaltens nachgewiesen und weitere Gesetzmäßigkeiten entdeckt.:Einführung Grundlagen Verwandte Arbeiten Sprachen für Spielbeschreibung Implementation der Spielinfrastruktur Beschreibung Strategischen Verhaltens Resümee Ergebnisse Ausblick / To understand human social, administrative and economic behavior, which can be considered as a game or strategic interaction, the purely analytical methods do not suffice. It is necessary to gather data of human strategic behavior. Based on data, one can model, simulate and predict such behavior. The theoretical part of the objective is achieved using a practice oriented conceptualization of the real agents\' - humans and machines - strategic interaction and mutual integration of the concepts from game theory and multi-agent systems, which go beyond the related work. The practical part is the design of an universally usable system that can perform the strategic interactions between real agents with maximum scientific benefit. The current implementation is one of the results of the work. Similar existing systems are GDL-server (for machines) [Genesereth et al., 2005] and z-Tree (for humans) [Fischbacher, 2007]. The work is divided in three fields - (1) development of languages for the description of a game, (2) a software system based on these languages and (3) an offline analysis of the data already gathered among other things using the system as a contribution to behavior definition facilities. The innovation of this work does not consist only in combining of the several fields to each other, but also in achieving of improvements in every field on its own. In the field of game definition languages, two languages are proposed - PNSI and SIDL, which both can define games of imperfect information in discrete time. It is an improvement comparing with hitherto languages as Gala and GDL. Especially, the Petri net based language PNSI can likewise be used for game servers and game theoretic algorithms like GAMBIT. The developed system FRAMASI is based on JADE [Bellifemine et al., 2001] and is ahead of the hitherto client-server solutions through the advantages of the multi-agent systems. Using the originated system, an experiment has been conducted according to the standards from the experimental game theory, and thus demonstrated the practicability. The experiment had the objective to provide data on the human unpredictability and the ability to predict others. Therefore, variants of Roshambo were used. The data from this experiment and from an experiment of an external workgroup with a similar motivation were analyzed using data mining. As results, the regularities of the behavior reported in literature have been demonstrated and further regularities have been discovered.:Einführung Grundlagen Verwandte Arbeiten Sprachen für Spielbeschreibung Implementation der Spielinfrastruktur Beschreibung Strategischen Verhaltens Resümee Ergebnisse Ausblick

Over the past few years, mining data located in heterogeneous and geographically distributed sites have been designated as one of the key important issues. Loading distributed data into centralized location for mining interesting rules is not a good approach. This is because it violates common issues such as data privacy and it imposes network overheads. The situation becomes worse when the network has limited bandwidth which is the case in most of the real time systems. This has prompted the need for intelligent data analysis to discover the hidden information in these huge amounts of distributed databases. In this research, we present an incremental approach for building an efficient Multi-Agent based algorithm for mining real world databases in geographically distributed sites. First, we propose the Distributed Multi-Agent Association Rules algorithm (DMAAR) to minimize the all-to-all broadcasting between distributed sites. Analytical calculations show that DMAAR reduces the algorithm complexity and minimizes the message communication cost. The proposed Multi-Agent based algorithm complies with the Foundation for Intelligent Physical Agents (FIPA), which is considered as the global standards in communication between agents, thus, enabling the proposed algorithm agents to cooperate with other standard agents. Second, the BitTable Multi-Agent Association Rules algorithm (BMAAR) is proposed. BMAAR includes an efficient BitTable data structure which helps in compressing the database thus can easily fit into the memory of the local sites. It also includes two BitWise AND/OR operations for quick candidate itemsets generation and support counting. Moreover, the algorithm includes three transaction trimming techniques to reduce the size of the mined data. Third, we propose the Pruning Multi-Agent Association Rules algorithm (PMAAR) which includes three candidate itemsets pruning techniques for reducing the large number of generated candidate itemsets, consequently, reducing the total time for the mining process. The proposed PMAAR algorithm has been compared with existing Association Rules algorithms against different benchmark datasets and has proved to have better performance and execution time. Moreover, PMAAR has been implemented on real world distributed medical databases obtained from more than one hospital in Egypt to discover the hidden Association Rules in patients¿ records to demonstrate the merits and capabilities of the proposed model further. Medical data was anonymously obtained without the patients¿ personal details. The analysis helped to identify the existence or the absence of the disease based on minimum number of effective examinations and tests. Thus, the proposed algorithm can help in providing accurate medical decisions based on cost effective treatments, improving the medical service for the patients, reducing the real time response for the health system and improving the quality of clinical decision making.

In the last decades, significant theoretical advances have been made on the field of distributed mulit-agent control theory. One of the most common systems that can be modelled as multi-agent systems are the so called formation control problems, in which a network of mobile agents is controlled to move towards a desired final formation. These problems additionally pose practical challenges, namely limited access to information about the global state of the system, which justify the use distributed and localized approaches for solving the control problem. The problem is further complicated if partial or no information is known about the dynamic model of the system. A widely used fundamental challenge of this approach in this setting is that the state-action space size scales exponentially with the number of agents, rendering the problem intractable for a large networks. This thesis presents a scalable and localized reinforcement learning approach to a traditional multi-agent formation control problem, with collision avoidance. A scalable reinforcement learning advantage actor critic algorithm is presented, based on previous work in the literature. Sub-optimal bounds are calculated for the accumulated reward and policy gradient localized approximations. The algorithm is tested on a two dimensional setting, with a network of mobile agents following simple integrator dynamics and stochastic localized policies. Neural networks are used to approximate the continuous value functions and policies. The formation control with collisions avoidance formulation and the algorithm presented show good scalability properties, with a polynomial increase in the number of function approximations parameters with number of agents. The reduced number of parameters decreases learning time for bigger networks, although the efficiency of computation is decreased compared to state of the art machine learning implementations. The policies obtained achieve probably safe trajectories although the lack of dynamic model makes it impossible to guarantee safety. / Under de senaste decennierna har betydande framsteg gjorts inom området för distribuerad mulit-agent reglerteori. Ett av de vanligaste systemen som kan modelleras som multiagentsystem är de så kallade formationskontrollproblemen, där ett nätverk av mobila agenter styrs för att röra sig mot en önskad slutlig formation. om systemets globala tillstånd, vilket motiverar användningen av distribuerade och lokaliserade tillvägagångssätt för att lösa det reglertekniska problemet. Problemet kompliceras ytterligare om delvis eller ingen information är känd om systemets dynamiska modell. Ett allmänt använt tillvägagångssätt för modellfri kontroll är reinforcement learning (RL). En grundläggande utmaning med detta tillvägagångssätt i den här miljön är att storleken på state-action utrymmet skalas exponentiellt med antalet agenter, vilket gör problemet svårlöst för ett stort nätverk. Detta examensarbete presenterar en skalbar och lokaliserad reinforcement learning metod på ett traditionellt reglertekniskt problem med flera agenter, med kollisionsundvikande. En reinforcement learning advantage actor critic algoritm presenteras, baserad på tidigare arbete i litteraturen. Suboptimala gränser beräknas för den ackumulerade belönings- och policygradientens lokaliserade approximationer. Algoritmen testas i en tvådimensionell miljö, med ett nätverk av mobila agenter som följer enkel integratordynamik och stokastiska lokaliserade policyer. Neurala nätverk används för att approximera de kontinuerliga värdefunktionerna och policyerna. Den presenterade formationsstyrningen med kollisionsundvikande formulering och algoritmen visar goda skalbarhetsegenskaper, med en polynomisk ökning av antalet funktionsapproximationsparametrar med antalet agenter. Det minskade antalet parametrar minskar inlärningstiden för större nätverk, även om effektiviteten i beräkningen minskar jämfört med avancerade maskininlärningsimplementeringar. De erhållna policyerna uppnår troligen säkra banor även om avsaknaden av dynamisk modell gör det omöjligt att garantera säkerheten. / En las últimas décadas, se han realizado importantes avances teóricos en el campo de la teoría del control multiagente distribuido. Uno de los sistemas más comunes que se pueden modelar como sistemas multiagente son los llamados problemas de control de formación, en los que se controla una red de agentes móviles para alcanzar una formación final deseada. Estos problemas plantean desafíos prácticos como el acceso limitado a la información del estado global del sistema, que justifican el uso de algoritmos distribuidos y locales para resolver el problema de control. El problema se complica aún más si solo se conoce información parcial o nada sobre el modelo dinámico del sistema. Un enfoque ampliamente utilizado para el control sin conocimiento del modelo dinámico es el reinforcement learning (RL). Un desafío fundamental de este método en este entorno es que el tamaño de la acción y el estado aumenta exponencialmente con la cantidad de agentes, lo que hace que el problema sea intratable para una red grande. Esta tesis presenta un algoritmo de RL escalable y local para un problema tradicional de control de formación con múltiples agentes, con prevención de colisiones. Se presenta un algoritmo “advantage actor-”critic, basado en trabajos previos en la literatura. Los límites subóptimos se calculan para las aproximaciones locales de la función Q y gradiente de la política. El algoritmo se prueba en un entorno bidimensional, con una red de agentes móviles que siguen una dinámica de integrador simple y políticas estocásticas localizadas. Redes neuronales se utilizan para aproximar las funciones y políticas de valor continuo. La formulación de del problema de formación con prevención de colisiones y el algoritmo presentado muestran buenas propiedades de escalabilidad, con un aumento polinómico en el número de parámetros con el número de agentes. El número reducido de parámetros disminuye el tiempo de aprendizaje para redes más grandes, aunque la eficiencia de la computación disminuye en comparación con las implementaciones de ML de última generación. Las politicas obtenidas alcanzan trayectorias probablemente seguras, aunque la falta de un modelo dinámico hace imposible garantizar la completa prevención de colisiones. / A les darreres dècades, s'han realitzat importants avenços teòrics en el camp de la teoria del control multiagent distribuït. Un dels sistemes més comuns que es poden modelar com a sistemes multiagent són els anomenats problemes de control de formació, en els què es controla una xarxa d'agents mòbils per assolir una formació final desitjada. Aquests problemes plantegen reptes pràctics com l'accés limitat a la informació de l'estat global del sistema, que justifiquen l'ús d'algorismes distribuïts i locals per resoldre el problema de control. El problema es complica encara més si només es coneix informació parcial sobre el model dinàmic del sistema. Un mètode àmpliament utilitzat per al control sense coneixement del model dinàmic és el reinforcement learning (RL). Un repte fonamental d'aquest mètode en aquest entorn és que la mida de l'acció i l'estat augmenta exponencialment amb la quantitat d'agents, cosa que fa que el problema sigui intractable per a una xarxa gran. Aquesta tesi presenta un algorisme de RL escalable i local per a un problema tradicional de control de formació amb múltiples agents, amb prevenció de col·lisions. Es presenta un algorisme “advantage actor-”critic, basat en treballs previs a la literatura. Els límits subòptims es calculen per a les aproximacions locals de la funció Q i gradient de la política.’ Lalgoritme es prova en un entorn bidimensional, amb una xarxa ’dagents mòbils que segueixen una dinàmica ’dintegrador simple i polítiques estocàstiques localitzades. Xarxes neuronals s'utilitzen per aproximar les funcions i les polítiques de valor continu. La formulació del problema de formació amb prevenció de col·lisions i l'algorisme presentat mostren bones propietats d'escalabilitat, amb un augment polinòmic en el nombre de paràmetres amb el nombre d'agents. El nombre reduït de paràmetres disminueix el temps d'aprenentatge per a les xarxes més grans, encara que l'eficiència de la computació disminueix en comparació amb les implementacions de ML d'última generació. Les polítiques obtingudes aconsegueixen trajectòries probablement segures, tot i que la manca d'un model dinàmic fa impossible garantir la prevenció completa de col·lisions.

This thesis explores the integration of IoT protocols to enhance supply chain efficiency and sustainability by developing a flexible automated system. The research covers the integration of a palletization optimizer with a flexible robotic platform, a project conducted in collaboration with OpiFlex and Linköping University. Flexibility and sustainability in production, particularly in the food and beverage industry, are critical yet challenging to achieve. This research addresses these challenges by proposing a system that aligns the output with customer needs by combining these technologies. The research employs a combination of case study and exploratory methodologies. The development approach synthesizes elements from Set-Based Design, Point-Based Design, and Agile development frameworks. The primary research questions focus on identifying the best system architecture for integrating the palletization optimizer with a lower-level automation platform and outlining the steps needed to transform this integration into a commercially viable product. The system includes the optimizer, capable of processing customer orders and configuring products on mixed output pallets, integrated with a flexible robotic system provided by OpiFlex. The work involved evaluating communication protocols, MQTT, OPC UA, and TCP/IP, and designing robust interactions and interfaces between the subsystems. The results demonstrate the system's architecture and interaction protocols.  The thesis concludes with a discussion of the results in comparison to the application scenario and the standards consulted. The conclusion is that the chosen interface practices should remain largely intact but be re-developed using an OPC UA-based architecture. The main reasons for this are its support for both pub/sub and client-server models, increased security, and greater support for enterprise application integration. However, depending on the specific application, the downsides of OPC UA may outweigh its benefits.