Dans le cadre des évolutions du bâtiment, il est de plus en plus nécessaire d'interconnecter des objets communicants. Cette démarche est cependant freinée du fait de l'absence d'un protocole de communication standard. La maison intelligente n'est pas un nouveau concept, et l'automatisation de la maison est devenue un sujet de recherche à la mode ces dernières années. La problématique de cette thèse se focalise sur l'interopérabilité de systèmes communicants. Plus précisément, nous nous intéressons à la manière de créer une coopération entre différents dispositifs d'un environnement, afin de masquer la complexité à l'usager. Dans un premier temps, l'interopérabilité entre ces objets est réalisée par la mise en place d'un écosystème virtuel où les dispositifs peuvent communiquer leurs états ou l'état de l'environnement. C'est l'abstraction des systèmes. Dans ce mémoire, nous abordons comment et pourquoi les systèmes sont nécessairement abstraits lorsque l'on s'intéresse à la prise en compte du contexte de l'installation. L'étude de la prise en compte du contexte à permis de procurer les données scientifiques à l'entreprise Overkiz afin qu'elle puisse réaliser l'abstraction des dispositifs domotiques. Ensuite, la coopération entre les objets communicants est mise en œuvre par un outil mathématique permettant de modéliser le comportement d'un environnement composé d'actionneurs, de capteurs et d'interfaces utilisateurs. Le comportement est formalisé au moyen de la théorie des automates étendus et plus particulièrement par la théorie des systèmes de transitions symboliques à entrées/sorties (IOSTS). Nous synthétisons alors le pilotage par un "contrôleur" du comportement global d'une installation, à partir de règles de contraintes ou de règles d'action. / In the context of the building developments, it is increasing necessary to interconnect communicating objects. This approach is slow down because of the lack of a standard communication protocol. The smart home is not a new concept, and home automation has become a fashion research topic in recent years. The problem of this thesis focuses on the interoperability of communicating systems. Specifically, we focus on how to establish cooperation between different devices of an environment in order to hide the complexity to a user. Initially, interoperability between these objects is achieved by the establishment of a virtual ecosystem where devices can communicate their states or the state of the environment. It is the abstraction of systems. In this paper, we discuss how and why systems are necessarily abstract when we are interested in taking into consideration the context of the installation. The study of consideration of context allowed to provide scientific data to the company Overkiz in order to realize abstraction of home automation devices. Second, cooperation among communicating objects is implemented by a mathematical tool to model the behavior of an environment cibsustubg of actuators, sensors and user interfaces. The behavior is formalized using the automata theory and more specificaly the automaton theory of Input/Output Symbolic Transition Systems (IOSTS). We synthesize then how to synthetise a "controller" of a global behavior of an installation, from constraint rules or rules of action.

Muitos dos processos que afetam a distribuição, estrutura e dinâmica de uma população arbórea vegetal podem ocorrer muito cedo na vida de uma planta. Desta forma, o estudo dos processos de germinação e estabelecimento das plantas e fundamental para entender o crescimento populacional e a dispersão de suas sementes. Neste trabalho, apresentamos modelos de vegetação que usam autômatos celulares para representar o crescimento de espécies arbóreas nas áreas alagáveis da floresta amazônica considerando a dispersão de suas sementes, o recrutamento dos indivíduos e a influência da inundação. O primeiro dos modelos representa a dinâmica de vegetaão sem considerar diretamente a influência do pulso de inundação. O segundo modelo representa a influência da inundação via a mortalidade dos indivíduos da população. O terceiro modelo acrescenta a variação no tempo da altura da inundação,adicionando informações sobre o relevo da regi ão. É feita uma análise da influência de cada parâmetro principal na dinâmica da vegetação. Além disso apresentamos uma análise dos resultados que avaliam o comportamento do modelo, bem como a consistência e a estabilidade dos resultados obtidos pelos modelos computacionais. / Most processes that aect the distribution, structure and dynamics of tree populations occur early in the plant life cycle. Hence, it is very important to study seed germination and seedling establishment processes to understand the growth and the dispersion of plant populations. In this work, we present three vegetation models to represent the tree growth process in flooded areas of the Amazon Forest that use cellular automata. We consider seed dispersion, seedling establishment, in uence of light, recruitment of individuals, as well as the ood gradient in the models.The rst model represents the influence of flooding in the vegetation dynamic only through data and adequate hypotheses about the ooded regions. The second model represents the influence of ooding in the vegetation dynamic taking into account its in uence on mortalities. The third model represents the ood varying along time and includes topographic information.We analyze the in uence of each parameter on the vegetation dynamic. Furthermore, we present a statistical analysis of the results, evaluating the model behavior and the consistency and stability of the results obtained by the computational models.

As variáveis de processo determinam as propriedades dos produtos resultantes dos processos de fundição ou de soldagem, que são função da sua macro e microestrutura bruta de solidificação. Um dos parâmetros importantes para se determinar as propriedades de um produto é a posição da transição colunarequiaxial (CET) e, por este motivo, o entendimento dos fenômenos físicos que causam esta transição é essencial. Com o intuito de se prever a formação da CET, surgiram os métodos empíricos e os modelos matemáticos, que são divididos em dois grandes grupos: modelos determinísticos e modelos estocásticos. Estes dois grupos foram bem estudados, porém nunca foram comparados entre si, particularmente em relação à previsão da posição da CET. O presente trabalho tem como um primeiro objetivo preencher esta lacuna através da comparação entre estes modelos. No entanto, o objetivo principal é apresentar, implementar e validar um novo modelo matemático, denominado de híbrido estocástico-determinístico (CADE -\"Cellular Automaton Deterministic\"), que combine características importantes e vantajosas de cada um dos dois grupos de modelos. Inicialmente, um modelo representante do grupo dos modelos estocásticos foi implementado e validado frente a resultados disponíveis na literatura. Durante esta validação, foi necessária a elaboração de um critério baseado na razão de aspecto dos grãos para a identificação da CET nas macroestruturas calculadas pelo modelo. Estes resultados foram então comparados com os resultados de modelos determinísticos para, após cuidadosa discussão, possibilitar a proposta e implementação do modelo híbrido. Os modelos determinísticos que utilizam o critério mecânico para prever o bloqueio de grãos colunares e a ocorrência da CET mostram regiões colunares em geral maiores que as previstas pelo modelo estocástico. Por outro lado, os modelos determinísticos que utilizam um critério de bloqueio a partir da interação do campo de concentração de soluto ao redor dos grãos prevêem uma CET em posições semelhantes às calculadas pelos modelos estocásticos. O modelo implementado no presente trabalho é capaz de prever a macroestrutura bruta de solidificação e ainda utilizar as equações tradicionalmente empregadas nos modelos determinísticos, sem a necessidade de qualquer método extra para prever a posição da frente de crescimento colunar ou o seu bloqueio por grãos equiaxiais. / The processing variables determine many properties of the products obtained by casting and welding processes and these properties, on the other hand, are strongly affected by the as-cast micro and macrostructure. Particularly the position of the columnar-to-equiaxed transition (CET), which determines the amount of columnar and equiaxed grains in the macrostructure, has an important effect on the properties of as-cast parts. Therefore, understanding the important physical phenomena that cause and affect the formation of the CET plays a crucial role in predicting the ascast macrostructure. To predict the CET formation, empirical methods and mathematical models have been developed. These models are frequently divided into two main groups: deterministic and stochastic. Both groups have been thoroughly studied, but a comparison between them was never attempted, especially regarding the prediction of the CET position. One of the main objectives of the present work is to fulfill this gap by carefully comparing these models. Nevertheless, the most important objective is to propose, implement, and validate a hybrid stochastic-deterministic model, referred to as CADE (Cellular Automaton Deterministic), that combines some important and well-known features of each model. Initially, a model from the stochastic group was implemented and validated using results available in the literature and then used to analyze the effects of some processing variables on the CET prediction. To carry out this analyzes, a criteria based on the aspect ratio of the grains was proposed and developed to identify the CET region from macrostructure images calculated by the model. The results were compared with those obtained by deterministic models and finally led to the development of the new proposed model. This new model has some characteristics from each group of mathematical models and, for this reason, was denoted as hybrid. A deterministic model based on a mechanical blocking criterion to block columnar grains and define the CET position showed, for the most part, larger columnar regions than those predicted by the stochastic model. A deterministic model with a solutal blocking criterion to predict the CET showed results similar to those calculated with the stochastic model. The model proposed in the present work (CADE) was able to predict the as-cast macrostructure using the well-established deterministic equations, without the need for a new method to track columnar grains or predict their blocking by equiaxed grains.

Made available in DSpace on 2016-03-15T19:37:40Z (GMT). No. of bitstreams: 1 Alex Martins.pdf: 1863982 bytes, checksum: 63a3f4efd397697e6bc129fa070520d5 (MD5) Previous issue date: 2011-10-19 / Synchronization can be understood as a temporal organization of events, able of emerging in complex systems, as neural networks. Here, random graph and cellular automaton (CA) are used to represent neural networks, in order to investigate the occurrence of synchronism in such networks. The network coupling topology is of Newman-Watts type, formed by regular lattice with additional random connections. Two parts with this structure are connected by random links. Results obtained from numerical simulations with this model indicate variety of oscillatory behavior: there are cases in which both parts oscillate with equal, multiple and submultiple periods; and cases without oscillation. Investigations were performed concerning the relation among oscillatory behavior and maximum activity, the time to reach such an activity, the minimum average path length, size of the network, the percentage of random connections added and the rules of the CA state transition. Synchronous behavior was found in more than 75% of 28000 simulations accomplished. The system dynamics is influenced more by variations on the number of time steps in which a cell remains firing than by alterations on the lattice size or on the percentage of the randomly added links. / Pode-se entender sincronismo como uma organização temporal de eventos, possível de emergir em sistemas complexos, como redes neurais. Aqui, usam-se grafo aleatório e autômato celular (AC) para representar redes neurais, a fim de investigar a ocorrência de sincronismo em tais redes. A topologia de acoplamento da rede é do tipo Newman-Watts, formada por uma grade regular com ligações aleatórias acrescentadas. Duas partes com essa estrutura são conectadas por ligações aleatórias. Resultados obtidos por simulações numéricas com esse modelo indicam diversidade de comportamento oscilatório: há casos em que as duas partes oscilam em períodos iguais, múltiplos e submúltiplos; e casos sem oscilação. Investigaram-se as relações entre comportamento oscilatório e a atividade máxima, o tempo para se alcançar essa atividade, o comprimento do caminho mínimo médio, o tamanho da rede, a porcentagem de ligações aleatórias adicionadas, e as regras de transição de estado do AC. Comportamento síncrono foi encontrado em mais de 75% das 28.000 simulações realizadas. A dinâmica do sistema é mais influenciada por variações no número de passos de tempo em que a célula permanece disparando do que por alterações no tamanho do reticulado ou no percentual das ligações aleatórias adicionais.

Made available in DSpace on 2016-03-15T19:38:12Z (GMT). No. of bitstreams: 1 Douglas Nunes de Oliveira.pdf: 2639611 bytes, checksum: aa451c77b140c40c8a504474462ded1e (MD5) Previous issue date: 2008-09-17 / Probabilistic cellular automata can be used to model the spreading of contagious diseases in a population composed by susceptible, infected and recovered individuals. At each time step, a susceptible individual can either remain susceptible or contract the disease from infected individuals, where the probability related to the contagion depends on the number of infected individuals in contact with this susceptible individual. At each time step, an infected individual may (probabilistically) either remain infected or recuperate or die by the disease or die by other causes. A recovered individual, at each iteration, can either remain as he/she is or die. When an infected or recovered individual dies, a susceptible one appears in his/her place; thus, the population remains constant. Here, genetic algorithms are employed to identify the probability values concerning the processes of infection, cure and death, from epidemiological data from Arizona (USA) for measles. The goal is to obtain a model based on probabilistic rules of state transitions able of reproducing this time series and to verify the quality of the model prediction. This work reveals that the predictions are strongly influenced by the lattice dimension of the cellular automaton and by limitations imposed to the probability values. / Autômatos celulares probabilistas podem ser usados para modelar a propagação de doenças contagiosas numa população composta por indivíduos suscetíveis, infectados e recuperados da infecção. A cada passo de tempo, um indivíduo suscetível pode ou permanecer suscetível ou contrair a doença de infectados, sendo a probabilidade associada ao contágio dependente do número de infectados em contato com esse suscetível. A cada passo de tempo, um indivíduo infectado pode (probabilisticamente) permanecer infectado, ou se recuperar, ou morrer pela doença ou morrer de outras causas. Um indivíduo recuperado pode, a cada iteração, ou permanecer como está ou morrer. Quando um indivíduo infectado ou recuperado morre, nasce, em seu lugar, um suscetível, de modo que a população permanece constante. Aqui, algoritmos genéticos são empregados para identificar os valores das probabilidades associadas aos processos de infecção, recuperação e morte, a partir de dados epidemiológicos do Arizona (EUA) para catapora. O objetivo é obter um modelo baseado em regras probabilistas de transição de estados capaz de reproduzir essa série temporal e verificar a qualidade da previsão do modelo. Este trabalho revela que as previsões são fortemente influenciadas pelo tamanho do reticulado do autômato celular e por restrições impostas aos valores das probabilidades.

Le séchage des dispersions colloïdales a été et demeure très étudié en raison de son extrême complexité et de son utilisation dans de nombreuses applications. Dans cette thèse, nous avons développé une simulation, d'abord unidirectionnelle, basée sur le principe des automates cellulaires, qui traite la problématique du séchage horizontal et vertical. Ce travail permet de prédire, par calcul numérique, la distribution des particules et la position des fronts de séchage dans les dépôts de dispersions sous forme de films minces. Nos résultats ont montré que la pression existant dans le fluide est la somme des pressions de Laplace et hydrostatique. Par rapport aux modèles existants, cela modifie la convection des particules dans la partie fluide de la dispersion. La diffusion collective des particules chargées a été étudiée également. Contrairement aux prédictions théoriques antérieures, nous avons pu montrer que la diffusion collective des particules chargées pouvait jouer un rôle majeur, y compris dans le cadre de l’approximation de lubrification. Finalement, la simulation 1D a été étendue en 2D, ce qui a permis de comprendre la raison pour laquelle deux fronts dans deux directions perpendiculaire (cas d’une géométrie rectangulaire) avancent à des vitesses différentes. Une comparaison entre les données expérimentales et le calcul numérique sur le profil du film et la vitesse des fronts de séchage pour une dispersion de silice montre un bon accord. / Drying of colloidal dispersions, given their uses in several fields in everyday life, has been the subject of many studies for a long time. In this thesis, we first developed an unidirectional simulation, based on the principle of the cellular automaton, which deals with the problem of horizontal and vertical drying. This work makes it possible to predict, by numerical calculation, the distribution of the particles and the position of drying fronts in deposits in form of thin films.The profile of the film in the liquid part was studied. Our results have shown that the pressure in the fluid is the sum of the Laplace and hydrostatic pressures. This result affects the dynamics of particles in the fluid part of the dispersion, in particular convection. The collective diffusion of charged particles has also been studied. Contrary to what was predicted in previous theoretical models, we were able to show that the collective diffusion of the charged particles could be important even within the lubrication approximation. Finally, the 1D simulation was extended to 2D in order to understand the reason why two fronts in perpendicular directions (case of a rectangular geometry) advance at different speeds. A comparison between the experimental data for the drying of a silica dispersion and the numerical calculation shows good agreement.

Tato disertační práce je zaměřena na vysokorychlostní simulace dopravy a jejich přesnou kalibraci pomocí různých typů dopravních dat. Práce se po úvodním popisu motivace pro samotný výzkum nejdříve věnuje současnému stavu poznání, a dále rozdělení simulátorů dopravy, zejména podle typu dat, se kterými se v nich pracuje. Úpravou existujícího řešení je navržen vlastní mikrosimulační model, který je založen na principu celulárního automatu. S tímto novým modelem je pak experimentováno, zpočátku z pohledu rychlosti simulace a další rozšiřitelnosti. Je navržena a popsána technika, kterou je možné navržený model významně akcelerovat a následně provádět simulace rychleji než v reálném čase i pro rozsáhlá území. Práce dále přistupuje k samotné kalibraci modelu, ke které byl využit evoluční přístup. Je představena metoda pro efektivní způsob optimalizace parametrů mikrosimulačního modelu, která nevyžaduje citlivostní analýzu a je schopná nalézt jinak obtížně nastavitelné parametry modelu. Kvalita získaných optimalizovaných modelů byla analyzována jak pomocí makroskopických, tak i mikroskopických dopravních dat, a to i s ohledem na jejich reálné vlastnosti, tj. chybějící vzorky. Dále jsou zhodnoceny výkonnostní a jiné kvalitativní parametry vlastního přístupu v porovnání s existujícími řešeními, přičemž bylo dosaženo významného zlepšení. Nakonec jsou představeny nástroje, které v rámci řešení vznikly. Na závěr je uvedeno další zaměření výzkumu, a to zejména s ohledem na využití výsledků práce v praxi.

The main goal of the work is to introduce new ideas how traditional approaches for designing an operation system and associated software can be improved to be a part of automatic software evolution. It is generally supposed that algorithms found by the genetic programming processes cannot be used for exact calculations but only for approximate solutions. Several examples of software evolution are introduced, to show that quite precise solutions can be achieved. To reach this goal, characteristics of tree-like structures with approaches based on cellular automata features are combined in a new promising technique of algorithm representation, joining benefits of both concepts. An application has been developed based on these new genetic programming concepts and it is supposed it can be a part of a future automatic software evolution process.

This thesis main focus is microscopic traffic sumulation. Part of this work is the design and implementation of microsimulation model based on cellular automaton. Implemented model supports calibration with genetic algorithm. The results of calibration and simulations are included.

The aim of this paper is to introduce the readers to the field of cellular automata, their design and their usage for structural design. Genetic algorithms are usually involved in designing complicated cellular automata, and because of that they are also briefly described here. For the purposes of this work sorting networks are considered as suitable structures to be designed using cellular automata, however, they are not a part of the automata but they are generated separately by modified rules of a local transition function.