Motion planning is defined as the problem of computing a feasible trajectory for an agent to follow. It is a well-studied problem with applications in fields such as robotics, control theory and artificial intelligence. In the last decade there has been an increased interest in algorithms for motion planning under uncertainty where the agent does not know the state of the environment due to, e.g. motion and sensing uncertainties. One approach is to generate an initial feasible trajectory using for example an algorithm such as RRT* and then improve that initial trajectory using local optimization. This thesis proposes a new modification of the RRT* algorithm that can be used to generate initial paths from which initial trajectories for the local optimization step can be generated. Unlike standard RRT*, the modified RRT* generates multiple paths at the same time, all belonging to different families of solutions (homotopy classes). Algorithms for motion planning under uncertainty that rely on local optimization of trajectories can use trajectories generated from these paths as initial solutions. The modified RRT* is implemented and its performance with respect to computation time and number of paths found is evaluated on simple scenarios. The evaluations show that the modified RRT* successfully computes solutions in multiple homotopy classes. Two methods for motion planning under uncertainty, Trajectory-optimized LQG (T-LQG), and a belief space variant of iterative LQG (iLQG) are implemented and combined with the modified RRT*. The performance with respect to cost function improvement, computation time and success rate when following the optimized trajectories for the two methods are evaluated in a simulation study. The results from the simulation studies show that it is advantageous to generate multiple initial trajectories. Some initial trajectories, due to for example passing through narrow passages or through areas with high uncertainties, can only be slightly improved by trajectory optimization or results in trajectories that are hard to follow or with a high collision risk. If multiple initial trajectories are generated the probability is higher that at least one of them will result in an optimized trajectory that is easy to follow, with lower uncertainty and lower collision risk than the initial trajectory. The results also show that iLQG is much more computationally expensive than T-LQG, but that it is better at computing control policies to follow the optimized trajectories.

The topic of this thesis is the theoretical and numerical research of optimal control problems for uncertain nonlinear systems, described by semilinear parabolic differential equations with additive noise, where the state is not completely available. Based on a paper by Kazufumi Ito and Karl Kunisch, which was published in 2006 with the title "Receding Horizon Control with Incomplete Observations", we analyze a Model Predictive Control (MPC) approach where the resulting linear problems on small intervals are solved with a Linear Quadratic Gaussian (LQG) design. Further we define a performance index for the MPC/LQG approach, find estimates for it and present bounds for the solutions of the underlying Riccati equations. Another large part of the thesis is devoted to extensive numerical studies for an 1+1- and 3+1-dimensional problem to show the robustness of the MPC/LQG strategy. The last part is a generalization of the MPC/LQG approach to infinite-dimensional problems.

This master’s thesis examines how a small MIMO lighting system can be identified and controlled. Two approaches are examined and compared; the first approach is a dynamic model using state space representation, where the system identification technique is Recursive Least Square, RLS, and the controller is an LQG controller; the second approach is a static model derived from the physical properties of light and a feedback feed-forward controller consisting of a PI controller coupled with a Control Allocation, CA, technique. For the studied system, the CA-PI approach significantly outperforms the LQG-RLS approach, which leads to the conclusion that the system’s static properties are predominant compared to the dynamic properties.

A utilização de sistemas semi-ativos no controle de movimento e vibração vem crescendo muito nos últimos anos, e está se firmando como a opção mais econômica em muitas aplicações. Os sistemas de controle semi-ativos baseados em fluidos magneto-reológicos (MR) já são empregados no controle de vibrações desde máquinas de lavar a grandes pontes suspensas, e até mesmo no controle de vibração de edifícios sujeitos a abalos sísmicos. No controle de movimento, eles têm sido aplicados em aparelhos de fisioterapia, próteses de joelho humano e embreagens. No setor automobilístico, os sistemas semi-ativos são empregados em sistemas de controle da suspensão e sistemas anti-rolagem, e também no controle de sistemas de torque e tração. Visto a crescente importância dos sistemas de controle semi-ativo, este trabalho aborda este assunto dividindo-o em duas partes: a primeira, apresentada no volume 1, trata do projeto de atuadores magneto-reológicos e do seu modelamento dinâmico voltado para projetos de controle; e a segunda parte, apresentada no volume 2, trata de projeto de controladores para sistemas semi-ativos, mais especificamente voltados para o controle de suspensão veicular semi-ativa, buscando a melhoria do desempenho em conforto, segurança e dirigibilidade do veículo. Na primeira parte, a maior contribuição deste trabalho está em preencher uma lacuna na literatura no que diz respeito a uma técnica sistemática de projeto de atuadores magneto-reológicos e da forma mais adequada de sua representação dinâmica para projetos de controle, que é feita a partir de dados experimentais. Na segunda parte deste trabalho buscou-se desenvolver um sistema de controle economicamente viável para veículos utilitários (caminhonetes principalmente). Estes são os veículos que mais carecem de sistemas de suspensão capazes de atender a uma grande variedade de situações (variação de pista, variação de carga), e cujo mercado exige uma relação custo/benefício melhor que aquela dos mercados de carros de luxo, onde estes sistemas são empregados atualmente. Como resultados mais importantes deste trabalho têmse: uma proposição para o uso da lei de controle LQG/LTR em sistemas não estritamente próprios; um modelo de um veículo utilitário completo com sete graus de liberdade e capaz de representar as situações mais importantes do comportamento dinâmico deste tipo de veículo; o desenvolvimento de uma estratégia de controle robusto baseado na técnica LQG/LTR, adequada para trabalhar com atuadores semi-ativos, e que utiliza sensores de custo compatível com a aplicação. O sistema de controle proposto foi capaz de melhorar o desempenho em conforto e segurança, evitando situações de perda de aderência dos pneus e comportamentos dinâmicos indesejáveis destes veículos, que foi verificado experimentalmente e através de simulações em computador. / The application of semi-active systems to motion and vibration control has significantly grown during the last years, and it is becoming an economically viable option for several applications. Semi-active systems based on magneto-rheological (MR) fluids were applied to vibration control ranging from washing machines to suspended bridges, and also in vibration control of buildings under seismic tremors. Regarding the movement control, they have been used in gym apparati, human knee prosthesis and clutches. More specifically in the automotive industry, semi-active systems have been applied in suspension and anti-roll systems, and also in torque and traction control systems. This work treats this subject dividing it into two parts: the first one, presented in volume 1, deals with the design of semi-active actuators based on magneto-rheological fluids and dynamic modeling; and the second one, presented in volume 2, deals with the design of semi-active control for vehicular suspension systems. The control main objective is the performance improvement in comfort, safety and handling features of a vehicle. In the first part, the main contribution of this work is to fulfill a literature gap regarding a systematic procedure for design of magneto-rheological actuators and to find a better representation for the dynamic behavior regarding the control system development, which is done based on experimental data analysis. The second part of this work leaded with the development of a vehicular suspension system for utility vehicles (for SUVs, and mainly for light trucks), which need suspension systems able to face a great number of situations (road variation, load variation, etc.), and which present stability problems due to its high gravity center height. Important results of this work are: the development of a robust control strategy based on LQG/LTR techniques for non-strictly proper systems, the development of a dynamic model for light trucks and SUVs with seven degrees of freedom; the development of a robust control strategy based on the LQG/LTR synthesis, suit to work with semi-active actuators, and using sensors with costs compatible to the application. The proposed control system was able to improve the performance of comfort and safety, avoiding the loss of adherence between tires and the ground and other undesirable dynamic behaviors of these vehicles, that was verified experimentally and through computer simulations.

Com o aperfeiçoamento de materiais compósitos de alto desempenho, surge a possibilidade do desenvolvimento de estruturas inteligentes, onde atuadores e sensores piezelétricos estão integrados na estrutura com sistemas de controle adequados para a atenuação de vibrações. Projetos multidisciplinares se tornam cada vez mais complexos e sofisticados, envolvendo diversas fontes de incertezas que devem ser analisadas e quantificadas. O escopo principal desse trabalho visa o estudo da propagação de incertezas em estruturas de materiais compósitos laminados com atuadores e sensores piezelétricos, onde entradas e parâmetros do projeto podem ser fontes aleatórias e/ou nebulosas. Para atingir esse objetivo é adotada a metodologia fuzzy, com a aplicação de otimização de cortes-α. Essa técnica é utilizada na presença de informações vagas ou imprecisas acerca da aleatoriedade presente. Nesse estudo projetam-se, através do método dos elementos finitos, estruturas em forma de placa e casca de material compósito laminado com atuadores e sensores piezelétricos acoplados, controlados pelos reguladores Linear Quadratic Regulator (LQR) e Linear Quadratic Gaussian (LQG). Inicialmente são realizados estudos de otimização para encontrar a melhor localização dos componentes piezelétricos pelos Gramianos de controlabilidade e observabilidade, assim como os fatores de ponderação das leis de controle. O desenvolvimento é realizado no espaço modal reduzido visando um melhor desempenho computacional. As métricas escolhidas para avaliação do controle de vibração e análise das saídas incertas do sistema são baseadas nas energias cinética, potencial e elétrica. Também apresentam-se estudos de envelopes relacionados ao deslocamentos e às frequências naturais da estrutura devido às incertezas. Os resultados mostraram que as otimizações por corte-α para tratar números fuzzy nesse tipo de problema são robustas e eficientes, encontrando-se valores extremos das saídas desejadas. Além de ser um método não intrusivo, também pode ser utilizado em problemas com um número elevado de parâmetros incertos como entrada. / The possibility of developments of smart structures arises with high performance composite materials improvements, where piezoelectric actuators and sensors are embedded into the structures, following a suitable control laws for vibration attenuation. Multidisciplinary projects are becoming highly complex and sophisticated, involving several sources of uncertainty that should be analyzed and quantified. The main objective for this work is to study the uncertainty propagation in composite laminate structures with embedded piezoelectric actuators and sensors, considering random and/or fuzzy sources for the inputs and design parameters. To accomplish this objective, it is adopted the fuzzy α-cut optimizations methodology. This technique is used when the available information related to the actual randomness is vague or imprecise. In this study, laminated composite shells and plates structures are designed and analyzed by the finite element method, where embedded piezoelectric actuators and sensors controlled by Linear Quadratic Regulator (LQR) and Linear Quadratic Gaussian (LQG) are present. Initially, optimization analyses are executed to find the best arrangement for the piezoelectric material using controllability and observability Gramians metrics, as well as the best controller parameters. This study is developed in the reduced modal space looking for computational costs savings. The chosen rating metrics for the vibration control and uncertainty analysis are based on kinetic, potential and electrical energies. Structural displacements and natural frequency envelopes due uncertainty are also studied and presented. The results have shown that the fuzzy α-cut optimizations methodology is robust and efficient to find extreme values for the sought outputs. In addition to being a non-intrusive method, it is also able to deal with a large number of uncertain input parameters.

Submitted by Cleide Dantas (cleidedantas@ufpa.br) on 2014-07-31T13:11:15Z No. of bitstreams: 2 license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) Dissertacao_ControladoresRobustosTipo.pdf: 1406069 bytes, checksum: 94652acee018b9469e9601f509fb6582 (MD5) / Approved for entry into archive by Ana Rosa Silva (arosa@ufpa.br) on 2014-09-04T15:03:43Z (GMT) No. of bitstreams: 2 license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) Dissertacao_ControladoresRobustosTipo.pdf: 1406069 bytes, checksum: 94652acee018b9469e9601f509fb6582 (MD5) / Made available in DSpace on 2014-09-04T15:03:43Z (GMT). No. of bitstreams: 2 license_rdf: 23898 bytes, checksum: e363e809996cf46ada20da1accfcd9c7 (MD5) Dissertacao_ControladoresRobustosTipo.pdf: 1406069 bytes, checksum: 94652acee018b9469e9601f509fb6582 (MD5) Previous issue date: 2014 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / O objetivo principal desta dissertação é apresentar uma solução eficiente, prática e de simples implementação para um problema recorrente em projetos de controladores robustos multivariáveis do tipo LQG/LTR: a elevada ordem que estes controladores podem obter dependendo das complicações apresentadas pelo sistema dificultando para que este possa ser controlado de maneira satisfatória. Para que esta meta seja alcançada, é apresentada uma técnica de redução do modelo de sistemas com metodologia bastante descomplicada, dispensando qualquer necessidade de complexas programações para a sua utilização. Esta metodologia porém, é somente aplicável a uma classe bastante específica de sistema. Em suma, o sistema deve possuir variáveis de estado desacopladas do restante do sistema, ou seja, variáveis que não sofram influências de outras e que também não provoquem grande efeito nas saídas do sistema. Foi escolhido um sistema multivariável de sexta ordem, com duas entradas e duas saídas para que a técnica de redução de ordem de modelo seja testada. Este sistema possui as características especiais mencionadas anteriormente bem como exige o projeto de compensador dinâmico e a adição de integradores às suas saídas para que seja controlado adequadamente. Este trabalho pretende apresentar o procedimento de todo o projeto mencionado, desde a obtenção de um modelo de ordem reduzida até a implementação do controlador LQG/LTR. Em seguida, o controlador obtido é testado através de diversas simulações e os resultados encontrados são discutidos para a avaliação da eficácia e da praticidade do método proposto para obtenção de controladores de ordem reduzida. / This thesis’ main goal is to introduce an efficient, practical and easy to implement solution to a recurring problem in projects of LQG/LTR multivariable robust controllers: the high order these controllers may obtain depending on the complications presented by the system hampering its control in a satisfactory way. For this goal to be achieved, a system model reduction technique with very simple methodology is introduced, dispensing any needs of complex programming for its use. This methodology however, is only applicable to a very specific class of system. Summarizing, the system must have state variables decoupled from the rest of the system, that is, variables that don’t not influenced by others and that also don’t cause major effects on the system’s outputs. It was chosen a sixth order multivariable system having two inputs and two outputs for the model order reduction be tested. This system has the special characteristics mentioned before and also demands a dynamic compensator project as well as the integrators addition to its outputs so it can be controlled adequately. This text intends to show the procedure for the whole project, since the reduced order model achievement to the LQG/LTR controller implementation. Then, the obtained controller is tested through several simulations and the attained results are discussed for effectiveness and practicality evaluation of the proposed method for reduced order controller project.

Ce travail est consacré au contrôle en boucle fermée des perturbations se développant linéairement dans des écoulements laminaires et incompressibles de types oscillateurs et amplificateurs de bruit. La loi de contrôle, calculée selon la théorie du contrôle LQG, est basée sur un modèle d'ordre réduit de l'écoulement obtenu par projection de Petrov-Galerkin. La stabilisation d'un écoulement de cavité de type oscillateur est traitée dans une première partie. Il est montré que la totalité de la partie instable de l'écoulement (les modes globaux instables) ainsi que la relation entrée-sortie (action de l'actionneur sur le capteur) de la partie stable doivent être captées par le modèle réduit afin de stabiliser le système. Les modes globaux, modes POD et modes BPOD sont successivement évalués comme bases de projection pour modéliser la partie stable. Les modes globaux ne parviennent pas à reproduire le comportement entrée-sortie de la partie stable et par conséquent ne peuvent stabiliser l'écoulement que lorsque l'instabilité du système est initialement faible (nombre de Reynolds proche de la criticité). En revanche, les modes POD et plus particulièrement BPOD sont capable d'extraire la dynamique entrée-sortie stable et permettent de stabiliser efficacement l'écoulement. La seconde partie de ce travail est consacrée à la réduction de l'amplification des perturbations sur une marche descendante. L'influence de la localisation du capteur et de la fonctionnelle de coût sur la performance du compensateur est étudiée. Il est montré que la troncature du modèle réduit peut rendre le système bouclé instable. Finalement, la possibilité de contrôler une simulation non-linéaire avec un modèle linéaire est évaluée.

Ce travail concerne la commande optimale robuste des systèmes hydrauliques à surface libre<br />(canaux d'irrigation). Nous nous sommes intéressés à deux approches de synthèse de<br />commande optimale. La première approche consiste à synthétiser une loi de commande<br />optimale LQG /H2-LTR (de dimension finie) avec pondérations fréquentielles robuste vis-àvis<br />des erreurs engendrées par la réduction à un modèle de dimension finie des équations de<br />Saint Venant. Le modèle réduit est obtenu par collocation orthogonale à partir du modèle<br />linéarisé tangent de Saint Venant. Un observateur est également proposé qui permet de<br />reconstruire l'état du système à partir des seuls états mesurés à l'amont et à l'aval de chaque<br />bief. Le régulateur optimal robuste et l'observateur ont été testés sur différents modèles de<br />référence ainsi que sur un micro-canal expérimental réel. Ils sont comparés aux résultats<br />obtenus par d'autres méthodes de régulation connues. La seconde approche de synthèse<br />consiste à faire l'approximation en dimension finie d'une loi de commande (de dimension<br />infinie) obtenue à partir des équations de Saint Venant linéarisées mais non réduites. Nous<br />présentons dans ce rapport des résultats liés à l'analyse et à la synthèse du régulateur optimal<br />LQ en dimension infinie appliquée aux équations de Saint Venant. Nous décrivons ensuite le<br />moyen d'obtenir une approximation en dimension finie du régulateur LQ sur la base de<br />l'équation de Riccati d'opérateurs associée au problème.

No description available.

Sheet and film forming processes refer to a set of processes that have a 2-dimensional sheet or film as their output. They are typically characterized by a scanning sensor which moves between the edges of the sheet in a periodic manner. This work is concerned with control and monitoring of such processes. There are three main contributions in this study. The first and foremost contribution is the reformulation of the Linear Quadratic Gaussian (LQG) objective function to give one the ability to trade-off between control over either dimension of the sheet. The second contribution is a method to derive the LQG trade-off surface between output variability in either dimension and input variance. The third contribution is a set of data driven techniques for performance assessment of these processes. Simulation results using the model used by Bergh and MacGregor (1987) are provided to support the proposed methods. / Process Control