Recent debiasing techniques are incorporated into the Ensemble Kalman-Bucy Filter (EnKBF). Specifically, a novel double randomization is applied. The EnKBF is a Monte Carlo (MC) method that approximates the Kalman-Bucy Filter (KBF), which in turn can be seen as the continuous-time version of the celebrated discrete-time Kalman Filter (KF). The KF is a method that combines sequential observations with an underlying dynamics model to predict the state of the quantity of interest. Our interest in the EnKBF comes from its relevance in high dimensions, where it overcomes the curse of dimensionality and outperforms other standard methods like the Particle Filter. We will consider debiasing techniques (also termed unbiased estimators) in order to improve the error-to-cost rate. Unbiased estimators are variance reduction techniques that produce unbiased and finite variance estimators. Applications of the EnKBF are numerous, from atmospheric sciences, numerical weather prediction, finance, machine learning, among others. Thus, improving the EnKBF is of interest. Numerical tests are done in order to evaluate the cost and the error-to-cost rate of the algorithm, where we consider Ornstein-Uhlenbeck processes. Specifically, a numerical comparison with the Multilevel Ensemble Kalman-Bucy Filter (MLEnKBF) is made using two different unbiased estimators, the coupled sum and the single term estimators. Additionally, we test two variants of the EnKBF, the Vanilla EnKBF, and the Deterministic EnKBF. We find that the error-to-cost rate is virtually the same, although the cost of the unbiased EnKBF is much higher.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-03-03T13:37:53Z No. of bitstreams: 1 eduardodasilvacastro.pdf: 10234661 bytes, checksum: c58d694820eabc593fe5f1b06aa4f93d (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-03-06T20:14:38Z (GMT) No. of bitstreams: 1 eduardodasilvacastro.pdf: 10234661 bytes, checksum: c58d694820eabc593fe5f1b06aa4f93d (MD5) / Made available in DSpace on 2017-03-06T20:14:38Z (GMT). No. of bitstreams: 1 eduardodasilvacastro.pdf: 10234661 bytes, checksum: c58d694820eabc593fe5f1b06aa4f93d (MD5) Previous issue date: 2011-02-02 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Com o desenvolvimento de novas tecnologias nas áreas de materiais, equipamentos eletrônicos e computação, a concepção de projetos estruturais vem sendo alterada. Estruturas cada vez mais leves e esbeltas vêm sendo construídas, o que, em alguns casos, tem levado a problemas de vibrações excessivas. Como forma de solucionar tais problemas pode-se empregar técnicas de controle ativo. O controle ativo estrutural consiste basicamente em impor forças de controle à estrutura visando a redução das amplitudes de vibração. Normalmente utiliza-se atuadores como macacos hidráulicos para a imposição das forças de controle. Uma das ferramentas mais importantes usadas na concepção de um sistema de controle ativo são os algoritmos numéricos usados no cálculo das forças de controle. Em geral estes algoritmos são baseados na resposta monitorada da estrutura e a eficácia do sistema de controle está diretamente ligada à qualidade dos algoritmos empregados. Dentre os algoritmos usados no controle ativo estão aqueles decorrentes do controle ótimo, definido por um regulador quadrático para sistemas de comportamento linear. Nesse caso, para o cálculo das forças de controle é necessária a determinação da matriz de Riccati, obtida através de métodos tais como: o algoritmo de Potter, o método da retro-integração temporal, o algoritmo LQR e o algoritmo baseado no método de Newton- Raphson, proposto nesta dissertação de mestrado. Um dos grandes obstáculos para a aplicação do controle ótimo em estruturas reais é que, em geral, os algoritmos de controle demandam o monitoramento de todos os graus de liberade (GLs) da estrutura. Alternativamente, pode-se utilizar métodos para a estimativa das respostas dinâmicas dos GLs não monitorados tais como os algoritmos denominados observadores apresentados neste trabalho. Finalmente pode-se afirmar que os ruídos inerentes aos sinais dos GLs monitorados podem prejudicar a qualidade do controle ativo. Desta forma faz-se também neste trabalho a avaliação da aplicação do filtro Kalman-Bucy visando a redução das perturbações geradas pelos ruídos em sistemas de controle ativo. Em suma, faz-se nesse trabalho uma avaliação de algoritmos numéricos aplicados ao controle ativo de vibrações mecânicas onde três aspectos inerentes aos algoritmos de controle são abordados: 1) exatidão no cálculo da matriz de Riccati; 2) eficiência do uso de algoritmos com a metodologia dos observadores de estado para estimativa de GLs não monitorados; 3) eficiência do uso do filtro de Kalman-Bucy para a redução de perturbações do sistema de controle geradas por ruídos. Os resultados obtidos mostram que o uso do algoritmo de Newton-Raphson, proposto neste trabalho, apresenta valores mais precisos para a determinação da Matriz de Riccati, levando a maiores reduções de vibrações com maiores magnitudes de forças de controle. Nota-se também que a técnica dos observadores de estado e do filtro de Kalman-Bucy se mostram eficientes nos sistemas de controle analisados. / With the development of new technologies in materials, electronics and computing, the conception of structural projects has been changed. Structures are getting lighter and slender, which in some cases, leads to vibration problems. Those problems can be solved with techniques of active control. The structural active control consists basically on imposing control forces on a structure aiming to reduce the amplitude of vibration. Usually hydraulic actuators are used for the imposition of control forces. One of the most important tool used in an active control system conception are numerical algorithms employed in the calculation of controlling forces. In general these algorithms are based on the response sensors of the structure and the efficiency of the control system is directly related to the quality of the employed algorithms. Among the algorithms used in active control are those arising from optimal control, wich are defined by a quadratic regulator for linear system. In this case, for the calculation of controlling forces is necessary to determine Riccati matrix, which may be obtained by means of Potter’s algorithm, the method of backward integration in time, the LQR algorithm and the algorithm based on Newton-Raphson method, proposed in this dissertation. One of the greatest obstacles for the application of optimal control in real structures is the need for control algorithms, in general, to request a monitoring of all degrees of freedom (DFs) of the structure. Alternatively, one way use methods for estimating the dynamic response of non-sensored DFs. This work presents the analysis of algorithms called state observers used in active control of structures. Finally it can be affirmed that the noise inherent to the DFs signs monitored may harm the quality of the active control. Thus it is also evaluated the implementation of Kalman-Bucy filter in order to reduce the disturbances generated by the noise in control system with state observers. In short, this work is an evaluation of numerical algorithms applied to active control of vibration and the aspects related to control algorithm are: 1) accuracy in the calculation of the Riccati matrix; 2) efficiency in the use of algorithms with the methodology of state observers to estimate unmonitored DFs, 3) influence of noise on the efficiency of active control of structures with state observers. The presented results support the conclusion that the proposed Newton-Raphson algorithm provides more precise values for the Riccati Matrix determination, leading to a better performance of control system. It was also noticed that the techniques of state observers and Kalman-Bucy filter had also good performance for the studied models.

In the Thesis we study the problem of linear filtration of Gaussian signals in finite-dimensional space. We use the Kalman-type equations for the filter to show that the filter depends continuously on the signal. Secondly, we show the same continuity property for the covariance of the error and verify existence and uniqueness of a solution to an integral equation that is satisfied by the filter even under more general assumptions. We present several examples of application of the continuity property that are based on the theory of stochastic differential equations driven by fractional Brownian motion. 1

Filtering for Stochastic Evolution Equations Vít Kubelka Doctoral thesis Abstract Linear filtering problem for infinite-dimensional Gaussian processes is studied, the observation process being finite-dimensional. Integral equations for the filter and for covariance of the error are derived. General results are applied to linear SPDEs driven by Gauss-Volterra process observed at finitely many points of the domain and to delayed SPDEs driven by white noise. Subsequently, the continuous dependence of the filter and observation error on parameters which may be present both in the signal and the obser- vation process is proved. These results are applied to signals governed by stochastic heat equations driven by distributed or pointwise fractional noise. The observation process may be a noisy observation of the signal at given points in the domain, the position of which may depend on the parameter. 1