MICROFICHE COPY AVAILABLE IN ARCHIVES AND ENGINEERING. / Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 1980. / Includes bibliographical references. / by Jeffrey Hastings Lang. / Ph.D.

Thesis (Sc.D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 1981. / MICROFICHE COPY AVAILABLE IN ARCHIVES AND ENGINEERING. / Bibliography: leaves 174-176. / by Victor On-Kwok Li. / Sc.D.

Thesis (M.S.)--Massachusetts Institute of Technology, Dept. of Aeronautics and Astronautics, 1982. / Microfiche copy available in Archives and Barker. / Includes bibliographical references. / by Peter David McQuade. / M.S.

Orientador: Liciana Vaz de Arruda Silveira / Banca: Francisco Torres-Avilés / Banca: Rogério Antonio de Oliveira / Resumo: Dados grupados e um caso particular de dados de sobrevivência com censura intervalar, que ocorre quando as observações são avaliadas nos mesmos intervalos de tempo, e geralmente está associado a dados com grande número de empates, ou seja, quando a proporção de empates e maior que 25% (Chalita et al., 2002), e podem desta forma ser analisados considerando o tempo discreto e ajustando-se modelos a probabilidade do indivíduo falhar em um certo intervalo, dado que sobreviveu ao intervalo anterior (Lawless, 2002). Dentre os possíveis modelos adaptados a este tipo de dados, tem-se o Modelo Logístico e o Modelo de Cox. A comparação entre o ajuste destes dois modelos pode ser feita utilizando o teste escore proposto por Colosimo et al. (2000), Bootstrap não paramétrico ou Critérios de Informação de Akaike (AIC). Uma alternativa a estas téccnicas, do ponto de vista bayesiano, e a utilização do Fator de Bayes. O objetivo deste estudoé comparar o ajuste do Modelo Logístico com o do Modelo de Cox para dados grupados e censurados usando inicialmente critérios clássicos de seleção de modelos: o Critério de Informação de Akaike (AIC), o Critério de Informação de Akaike Corrigido (AICc) e o Critério de Informação Bayesiano (BIC). Posteriormente foi utilizado o Fator de Bayes, assim como o Critério de Informação Deviance (DIC) e adaptações dos critérios clássicos já citados, utilizando as amostras a posteriori geradas por um método MCMC. Como ilustração, foi usado um conjunto de dados referente a uma manifestação clínica da doença de Chagas, conhecida como megacólon chagásico (Almeida, 1996). / Abstract: Grouped data is a particular case of survival data with interval censoring, that occurs when the observations are evaluated at the same time intervals, and is generally associated at data with a large number of draws, or draws ratio more than 25% (Chalita et al., 2002). It can be analyzed considering discrete-time and tting models at the probability of an individual fails in a certain interval, given that survived the previous one (Lawless, 2002). Among the possible models adapted to this type of data we can mention the Logistic Model and the Cox's Model. The comparison between the t of these two models can be made using the score test proposed by Colosimo et al. (2000), nonparametric Bootstrap or the Akaike Information Criterion (AIC). An alternative to these techniques, from the Bayesian point of view, is the Bayes Factor. The purpose of this study is to compare the t of the Logistic Model with the Cox's Model to grouped and censoring data using, initially, classic model selection criteria: Akaike Information Criterion (AIC), Akaike Information Criterion corrected (AICc) and Bayesian Information Criterion (BIC). After that, was used the Bayes Factor, as well as Deviance Information Criterion (DIC) and adaptations of the classic model selection criteria above mentioned, using the posteriori sample generated by a MCMC method. As an example, was used a data set related to a clinical manifestation of Chagas disease known as chagasic megacolon (Almeida, 1996) / Mestre

Thesis (M.S.)--Baylor University, 2005. / Includes bibliographical references (p. 120-121).

Models, modelling and control design play important parts in automatic control. The contributions in this thesis concern topics in all three of these concepts. The poles are of fundamental importance when analyzing the behaviour of a system, and pole placement is an intuitive and natural approach for control design. A novel parameterization for state feedback gains for pole placement in the linear multiple input case is presented and analyzed. It is shown that when the open and closed loop poles are disjunct, every state feedback gain can be parameterized. Other properties are also investigated. Hammerstein models have a static non-linearity on the input. A method for exact compensation of such non-linearities, combined with introduction of integral action, is presented. Instead of inversion of the non-linearity the method utilizes differentiation, which in many cases is simpler. A partial differential equation (PDE) can be regarded as an infinite order model. Many model based control design techniques, like linear quadratic Gaussian control (LQG), require finite order models. Active damping of vibrations in a viscoelastic beam, modelled as a PDE, is considered. The beam is actuated by piezoelectric elements and its movements are measured by strain gauges. LQG design is used, for which different finite order models, approximating the PDE model, are constructed. The so obtained controllers are evaluated on the original PDE model. Minimization of the measured strain yields a satisfactory performance, but minimization of transversal deflection does not. The effect of the model accuracy of the finite order model approximations is also investigated. It turns out that a model with higher accuracy in a specified frequency interval gives controllers with better performance. The wave equation is another PDE. A PDE model, with one spatial dimension, is established. It describes wave propagation in a tube perforated with helical slots. The model describes waves of both extensional and torsional type, as well as the coupling between the two wave types. Experimental data are used for estimation of model parameters, and for assessment of the proposed model in two different cases. The model is found adequate when certain geometrical assumptions are valid.

Thesis (M.Sc.)--Acadia University, 1999. / Includes bibliographical references (leaves 88-89). Also available on the Internet via the World Wide Web.

prepared by: Bernard C. Levy, Alan S. Willsky. / Prepared for Systems Theory and Operations Research Program, Division of Electrical, Computer, and Systems Engineering, National Science Foundation under NSF Grant ECS-8012668. / Description based on: Nov.1, 1980/June 30, 1981.

Made available in DSpace on 2015-12-10T14:23:17Z (GMT). No. of bitstreams: 0 Previous issue date: 2015-02-09. Added 1 bitstream(s) on 2015-12-10T14:29:34Z : No. of bitstreams: 1 000851915.pdf: 551687 bytes, checksum: cf0d5be63962cf4703a354b8a39f9c14 (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Dados grupados e um caso particular de dados de sobrevivência com censura intervalar, que ocorre quando as observações são avaliadas nos mesmos intervalos de tempo, e geralmente está associado a dados com grande número de empates, ou seja, quando a proporção de empates e maior que 25% (Chalita et al., 2002), e podem desta forma ser analisados considerando o tempo discreto e ajustando-se modelos a probabilidade do indivíduo falhar em um certo intervalo, dado que sobreviveu ao intervalo anterior (Lawless, 2002). Dentre os possíveis modelos adaptados a este tipo de dados, tem-se o Modelo Logístico e o Modelo de Cox. A comparação entre o ajuste destes dois modelos pode ser feita utilizando o teste escore proposto por Colosimo et al. (2000), Bootstrap não paramétrico ou Critérios de Informação de Akaike (AIC). Uma alternativa a estas téccnicas, do ponto de vista bayesiano, e a utilização do Fator de Bayes. O objetivo deste estudoé comparar o ajuste do Modelo Logístico com o do Modelo de Cox para dados grupados e censurados usando inicialmente critérios clássicos de seleção de modelos: o Critério de Informação de Akaike (AIC), o Critério de Informação de Akaike Corrigido (AICc) e o Critério de Informação Bayesiano (BIC). Posteriormente foi utilizado o Fator de Bayes, assim como o Critério de Informação Deviance (DIC) e adaptações dos critérios clássicos já citados, utilizando as amostras a posteriori geradas por um método MCMC. Como ilustração, foi usado um conjunto de dados referente a uma manifestação clínica da doença de Chagas, conhecida como megacólon chagásico (Almeida, 1996). / Grouped data is a particular case of survival data with interval censoring, that occurs when the observations are evaluated at the same time intervals, and is generally associated at data with a large number of draws, or draws ratio more than 25% (Chalita et al., 2002). It can be analyzed considering discrete-time and tting models at the probability of an individual fails in a certain interval, given that survived the previous one (Lawless, 2002). Among the possible models adapted to this type of data we can mention the Logistic Model and the Cox's Model. The comparison between the t of these two models can be made using the score test proposed by Colosimo et al. (2000), nonparametric Bootstrap or the Akaike Information Criterion (AIC). An alternative to these techniques, from the Bayesian point of view, is the Bayes Factor. The purpose of this study is to compare the t of the Logistic Model with the Cox's Model to grouped and censoring data using, initially, classic model selection criteria: Akaike Information Criterion (AIC), Akaike Information Criterion corrected (AICc) and Bayesian Information Criterion (BIC). After that, was used the Bayes Factor, as well as Deviance Information Criterion (DIC) and adaptations of the classic model selection criteria above mentioned, using the posteriori sample generated by a MCMC method. As an example, was used a data set related to a clinical manifestation of Chagas disease known as chagasic megacolon (Almeida, 1996)

The main purpose of this dissertation is to define and solve problems on optimal sensing and actuating policies in Cyber-Physical Systems (CPSs). Cyber-physical system is a term that was introduced recently to define the increasing complexity of the interactions between computational hardwares and their physical environments. The problem of designing the ``cyber'' part may not be trivial but can be solved from scratch. However, the ``physical'' part, usually a natural physical process, is inherently given and has to be identified in order to propose an appropriate ``cyber'' part to be adopted. Therefore, one of the first steps in designing a CPS is to identify its ``physical'' part. The ``physical'' part can belong to a large array of system classes. Among the possible candidates, we focus our interest on Distributed Parameter Systems (DPSs) whose dynamics can be modeled by Partial Differential Equations (PDE). DPSs are by nature very challenging to observe as their states are distributed throughout the spatial domain of interest. Therefore, systematic approaches have to be developed to obtain the optimal locations of sensors to optimally estimate the parameters of a given DPS. In this dissertation, we first review the recent methods from the literature as the foundations of our contributions. Then, we define new research problems within the above optimal parameter estimation framework. Two different yet important problems considered are the optimal mobile sensor trajectory planning and the accuracy effects and allocation of heterogeneous sensors. Under the remote sensing setting, we are able to determine the optimal trajectories of remote sensors. The problem of optimal robust estimation is then introduced and solved using an interlaced ``online'' or ``real-time'' scheme. Actuation policies are introduced into the framework to improve the estimation by providing the best stimulation of the DPS for optimal parameter identification, where trajectories of both sensors and actuators are optimized simultaneously. We also introduce a new methodology to solving fractional-order optimal control problems, with which we demonstrate that we can solve optimal sensing policy problems when sensors move in complex media, displaying fractional dynamics. We consider and solve the problem of optimal scale reconciliation using satellite imagery, ground measurements, and Unmanned Aerial Vehicles (UAV)-based personal remote sensing. Finally, to provide the reader with all the necessary background, the appendices contain important concepts and theorems from the literature as well as the Matlab codes used to numerically solve some of the described problems.