No description available.

The project considers the class of deterministic continuous-time optimal control problems (OCPs) with piecewise-affine (PWA) vector fields and polynomial data. The OCP is relaxed as an infinite-dimensional linear program (LP) over space of occupation measures. The LP is then written as a particular instance of the generalized moment problem which is then approached by an asymptotically converging hierarchy of linear matrix inequality (LMI) relaxations. The relaxed dual of the original LP gives a polynomial approximation of the value function along optimal trajectories. Based on this polynomial approximation, a novel suboptimal policy is developed to construct a state feedback in a sample-and-hold manner. The results show that the suboptimal policy succeeds in providing a stabilizing suboptimal state feedback law that drives the system relatively close to the optimal trajectories and respects the given constraints.

O presente trabalho apresenta um método de projeto de controlador com alocação de pólos em sistemas fuzzy utilizando funções de Lyapunov por partes e contínuas no espaço de estado. A idéia principal é utilizar controladores chaveados no espaço de estado para obter uma resposta transitória satisfatória do sistema, obtida pela localização dos pólos. A modelagem fuzzy Takagi-Sugeno é utilizada para representar um sistema não-linear em diversos pontos de linearização através de uma aproximação por vários modelos locais lineares invariantes no tempo. A análise de estabilidade e o projeto de sistemas de controle podem se formulados em termos de desigualdades matriciais lineares (em inglês, linear matrix inequalities (LMIs)), as quais são resolvidas por técnicas de programação convexa. Na análise de estabilidade ou na síntese de um controlador em sistemas fuzzy é necessário resolver um número determinado de LMIs de acordo com o número de modelos locais. Encontrar uma função de Lyapunov comum a todos os modelos locais pode ser inviável, especialmente quando se impõem critérios de desempenho, que aparecem como restrições no contexto de LMIs. A proposta de uma função de Lyapunov por partes objetiva diminuir o conservadorismo na busca de um controlador que leve os pólos de malha fechada à uma região desejada. Resultados de análise e síntese da teoria de sistemas lineares por partes contribuíram para a construção do resultado apresentado. Exemplos com simulação ilustram o método proposto. / This work presents a controller design method for fuzzy dynamic systems based on piecewise Lyapunov functions with constraints on the closed-loop pole location. The main idea is to use switched controllers to locate the poles of the system to obtain a satisfactory transient response. The pole placement strategy allows to specify the performance in terms of the desired time response of the feedback system. The Takagi-Sugeno fuzzy model can approximate the nonlinear system in several linearization points using linear time invariant systems. Thus, a global fuzzy model can be obtained from a fuzzy combination of these linear systems. Stability analysis and design of fuzzy control systems can be efficiently carried out in the context of linear matrix inequalities (LMIs). If the fuzzy system is described by many local models, the resulting set of LMIs may be infeasible. The search for a Lyapunov function in the fuzzy pole placement problem may be easier to be satisfied in a piecewise framework. Some results from piecewise linear systems theory have contributed to the development of the presented technique. Some examples are given to illustrate the proposed method.

Actuellement la nécessité des systèmes d’alimentation d’énergie, capables d’assurer un fonctionnement stable dans des domaines de fonctionnement assez larges avec des bonnes performances dynamiques (rapidité du système, variations limitées de la tension de sortie en réponse aux perturbations de charge ou de tension d’alimentation), devient de plus en plus importante. De ce fait, cette thèse est orientée sur la commande des convertisseurs de puissance DC-DC représentés par des modèles hybrides.En tenant compte de la structure variable de ces systèmes à commutation, un modèle hybride permet de décrire plus précisément le comportement dynamique d’un convertisseur dans son domaine de fonctionnement. Dans cette optique, l’approximation PWA est utilisée afin de modéliser les convertisseurs DC-DC. A partir des modèles hybrides développés, on s’est intéressé à la stabilisation des convertisseurs au moyen des correcteurs à gains commutés élaborés sur la base de fonctions de Lyapunov PWQ, et à l’implantation d’une commande prédictive explicite, en considérant des contraintes sur l’entrée de commande. La méthode de modélisation et les stratégies de commande proposées ont été appliquées sur deux topologies : un convertisseur buck, afin de mieux maîtriser le réglage des correcteurs et un convertisseur flyback avec filtre d’entrée. Cette dernière topologie nous a permis de répondre aux difficultés du point de vue de la commande (comportement à déphasage non-minimal) rencontrées dans la majorité des convertisseurs DC-DC. Les performances des commandes élaborées ont été validées en simulation sur les topologies considérées et expérimentalement sur une maquette du convertisseur buck. / Lately, power supply systems, guaranteeing the global stability for large enough operation ranges with good dynamic performances (small settling time, bounded overshoot of the output voltage in the presence of load or supply voltage variations), are strongly needed. Therefore, this thesis deals with control problems of DC-DC power converters represented by hybrid models.Considering the variable structure of these switched systems, a hybrid model describes more precisely the converter’s dynamics in its operating domain. From this perspective, a PWA (piecewise affine) approximation is used in order to model the DC-DC converters. Based on the developed hybrid models, first we have designed a stable piecewise linear state-feedback controller using piecewise quadratic (PWQ) Lyapunov functions, and secondly, we have implemented an explicit predictive control law taking into account constraints on the control input. The hybrid modeling technique and the proposed control strategies were applied on two different topologies of converters: a buck converter, in order to have a thorough knowledge of the controllers’ tuning, and a flyback converter with an input filter. This last topology, allowed us to manage different control problems (non-minimum phase behavior) encountered in the majority of topologies of DC-DC power converters. The controllers’ performances were validated in simulation on both considered topologies and also experimentally on buck converter.

O presente trabalho apresenta um método de projeto de controlador com alocação de pólos em sistemas fuzzy utilizando funções de Lyapunov por partes e contínuas no espaço de estado. A idéia principal é utilizar controladores chaveados no espaço de estado para obter uma resposta transitória satisfatória do sistema, obtida pela localização dos pólos. A modelagem fuzzy Takagi-Sugeno é utilizada para representar um sistema não-linear em diversos pontos de linearização através de uma aproximação por vários modelos locais lineares invariantes no tempo. A análise de estabilidade e o projeto de sistemas de controle podem se formulados em termos de desigualdades matriciais lineares (em inglês, linear matrix inequalities (LMIs)), as quais são resolvidas por técnicas de programação convexa. Na análise de estabilidade ou na síntese de um controlador em sistemas fuzzy é necessário resolver um número determinado de LMIs de acordo com o número de modelos locais. Encontrar uma função de Lyapunov comum a todos os modelos locais pode ser inviável, especialmente quando se impõem critérios de desempenho, que aparecem como restrições no contexto de LMIs. A proposta de uma função de Lyapunov por partes objetiva diminuir o conservadorismo na busca de um controlador que leve os pólos de malha fechada à uma região desejada. Resultados de análise e síntese da teoria de sistemas lineares por partes contribuíram para a construção do resultado apresentado. Exemplos com simulação ilustram o método proposto. / This work presents a controller design method for fuzzy dynamic systems based on piecewise Lyapunov functions with constraints on the closed-loop pole location. The main idea is to use switched controllers to locate the poles of the system to obtain a satisfactory transient response. The pole placement strategy allows to specify the performance in terms of the desired time response of the feedback system. The Takagi-Sugeno fuzzy model can approximate the nonlinear system in several linearization points using linear time invariant systems. Thus, a global fuzzy model can be obtained from a fuzzy combination of these linear systems. Stability analysis and design of fuzzy control systems can be efficiently carried out in the context of linear matrix inequalities (LMIs). If the fuzzy system is described by many local models, the resulting set of LMIs may be infeasible. The search for a Lyapunov function in the fuzzy pole placement problem may be easier to be satisfied in a piecewise framework. Some results from piecewise linear systems theory have contributed to the development of the presented technique. Some examples are given to illustrate the proposed method.

Supervised learning of piecewise linear models is a well studied problem in machine learning community. The key idea in piecewise linear modeling is to properly partition the input space and learn a linear model for every partition. Decision trees and regression trees are classic examples of piecewise linear models for classification and regression problems. The existing approaches for learning decision/regression trees can be broadly classified in to two classes, namely, fixed structure approaches and greedy approaches. In the fixed structure approaches, tree structure is fixed before hand by fixing the number of non leaf nodes, height of the tree and paths from root node to every leaf node of the tree. Mixture of experts and hierarchical mixture of experts are examples of fixed structure approaches for learning piecewise linear models. Parameters of the models are found using, e.g., maximum likelihood estimation, for which expectation maximization(EM) algorithm can be used. Fixed structure piecewise linear models can also be learnt using risk minimization under an appropriate loss function. Learning an optimal decision tree using fixed structure approach is a hard problem. Constructing an optimal binary decision tree is known to be NP Complete. On the other hand, greedy approaches do not assume any parametric form or any fixed structure for the decision tree classifier. Most of the greedy approaches learn tree structured piecewise linear models in a top down fashion. These are built by binary or multi-way recursive partitioning of the input space. The main issues in top down decision tree induction is to choose an appropriate objective function to rate the split rules. The objective function should be easy to optimize. Top-down decision trees are easy to implement and understand, but there are no optimality guarantees due to their greedy nature. Regression trees are built in the similar way as decision trees. In regression trees, every leaf node is associated with a linear regression function. All piece wise linear modeling techniques deal with two main tasks, namely, partitioning of the input space and learning a linear model for every partition. However, Partitioning of the input space and learning linear models for different partitions are not independent problems. Simultaneous optimal estimation of partitions and learning linear models for every partition, is a combinatorial problem and hence computationally hard. However, piecewise linear models provide better insights in to the classification or regression problem by giving explicit representation of the structure in the data. The information captured by piecewise linear models can be summarized in terms of simple rules, so that, they can be used to analyze the properties of the domain from which the data originates. These properties make piecewise linear models, like decision trees and regression trees, extremely useful in many data mining applications and place them among top data mining algorithms. In this thesis, we address the problem of supervised learning of piecewise linear models for classification and regression. We propose novel algorithms for learning piecewise linear classifiers and regression functions. We also address the problem of noise tolerant learning of classifiers in presence of label noise. We propose a novel algorithm for learning polyhedral classifiers which are the simplest form of piecewise linear classifiers. Polyhedral classifiers are useful when points of positive class fall inside a convex region and all the negative class points are distributed outside the convex region. Then the region of positive class can be well approximated by a simple polyhedral set. The key challenge in optimally learning a fixed structure polyhedral classifier is to identify sub problems, where each sub problem is a linear classification problem. This is a hard problem and identifying polyhedral separability is known to be NP complete. The goal of any polyhedral learning algorithm is to efficiently handle underlying combinatorial problem while achieving good classification accuracy. Existing methods for learning a fixed structure polyhedral classifier are based on solving non convex constrained optimization problems. These approaches do not efficiently handle the combinatorial aspect of the problem and are computationally expensive. We propose a method of model based estimation of posterior class probability to learn polyhedral classifiers. We solve an unconstrained optimization problem using a simple two step algorithm (similar to EM algorithm) to find the model parameters. To the best of our knowledge, this is the first attempt to form an unconstrained optimization problem for learning polyhedral classifiers. We then modify our algorithm to find the number of required hyperplanes also automatically. We experimentally show that our approach is better than the existing polyhedral learning algorithms in terms of training time, performance and the complexity. Most often, class conditional densities are multimodal. In such cases, each class region may be represented as a union of polyhedral regions and hence a single polyhedral classifier is not sufficient. To handle such situation, a generic decision tree is required. Learning optimal fixed structure decision tree is a computationally hard problem. On the other hand, top-down decision trees have no optimality guarantees due to the greedy nature. However, top-down decision tree approaches are widely used as they are versatile and easy to implement. Most of the existing top-down decision tree algorithms (CART,OC1,C4.5, etc.) use impurity measures to assess the goodness of hyper planes at each node of the tree. These measures do not properly capture the geometric structures in the data. We propose a novel decision tree algorithm that ,at each node, selects hyperplanes based on an objective function which takes into consideration geometric structure of the class regions. The resulting optimization problem turns out to be a generalized eigen value problem and hence is efficiently solved. We show through empirical studies that our approach leads to smaller size trees and better performance compared to other top-down decision tree approaches. We also provide some theoretical justification for the proposed method of learning decision trees. Piecewise linear regression is similar to the corresponding classification problem. For example, in regression trees, each leaf node is associated with a linear regression model. Thus the problem is once again that of (simultaneous) estimation of optimal partitions and learning a linear model for each partition. Regression trees, hinge hyperplane method, mixture of experts are some of the approaches to learn continuous piecewise linear regression models. Many of these algorithms are computationally intensive. We present a method of learning piecewise linear regression model which is computationally simple and is capable of learning discontinuous functions as well. The method is based on the idea of K plane regression that can identify a set of linear models given the training data. K plane regression is a simple algorithm motivated by the philosophy of k means clustering. However this simple algorithm has several problems. It does not give a model function so that we can predict the target value for any given input. Also, it is very sensitive to noise. We propose a modified K plane regression algorithm which can learn continuous as well as discontinuous functions. The proposed algorithm still retains the spirit of k means algorithm and after every iteration it improves the objective function. The proposed method learns a proper Piece wise linear model that can be used for prediction. The algorithm is also more robust to additive noise than K plane regression. While learning classifiers, one normally assumes that the class labels in the training data set are noise free. However, in many applications like Spam filtering, text classification etc., the training data can be mislabeled due to subjective errors. In such cases, the standard learning algorithms (SVM, Adaboost, decision trees etc.) start over fitting on the noisy points and lead to poor test accuracy. Thus analyzing the vulnerabilities of classifiers to label noise has recently attracted growing interest from the machine learning community. The existing noise tolerant learning approaches first try to identify the noisy points and then learn classifier on remaining points. In this thesis, we address the issue of developing learning algorithms which are inherently noise tolerant. An algorithm is inherently noise tolerant if, the classifier it learns with noisy samples would have the same performance on test data as that learnt from noise free samples. Algorithms having such robustness (under suitable assumption on the noise) are attractive for learning with noisy samples. Here, we consider non uniform label noise which is a generic noise model. In non uniform label noise, the probability of the class label for an example being incorrect, is a function of the feature vector of the example.(We assume that this probability is less than 0.5 for all feature vectors.) This can account for most cases of noisy data sets. There is no provably optimal algorithm for learning noise tolerant classifiers in presence of non uniform label noise. We propose a novel characterization of noise tolerance of an algorithm. We analyze noise tolerance properties of risk minimization frame work as risk minimization is a common strategy for classifier learning. We show that risk minimization under 01 loss has the best noise tolerance properties. None of the other convex loss functions have such noise tolerance properties. Empirical risk minimization under 01 loss is a hard problem as 01 loss function is not differentiable. We propose a gradient free stochastic optimization technique to minimize risk under 01 loss function for noise tolerant learning of linear classifiers. We show (under some conditions) that the algorithm converges asymptotically to the global minima of the risk under 01 loss function. We illustrate the noise tolerance of our algorithm through simulations experiments. We demonstrate the noise tolerance of the algorithm through simulations.

In this dissertation, a new non-global regression model - the partial linear threshold regression model (PLTRM) - is proposed. Various issues related to the PLTRM are discussed. In the first main section of the dissertation (Chapter 2), we define what is meant by the term “non-global regression model”, and we provide a brief review of the current literature associated with such models. In particular, we focus on their advantages and disadvantages in terms of their statistical properties. Because there are some weaknesses in the existing non-global regression models, we propose the PLTRM. The PLTRM combines non-parametric modelling with the traditional threshold regression models (TRMs), and hence can be thought of as an extension of the later models. We verify the performance of the PLTRM through a series of Monte Carlo simulation experiments. These experiments use a simulated data set that exhibits partial linear and partial nonlinear characteristics, and the PLTRM out-performs several competing parametric and non-parametric models in terms of the Mean Squared Error (MSE) of the within-sample fit. In the second main section of this dissertation (Chapter 3), we propose a method of estimation for the PLTRM. This requires estimating the parameters of the parametric part of the model; estimating the threshold; and fitting the non-parametric component of the model. An “unbalanced penalized least squares” approach is used. This involves using restricted penalized regression spline and smoothing spline techniques for the non-parametric component of the model; the least squares method for the linear parametric part of the model; together with a search procedure to estimate the threshold value. This estimation procedure is discussed for three mutually exclusive situations, which are classified according to the way in which the two components of the PLTRM “join” at the threshold. Bootstrap sampling distributions of the estimators are provided using the parametric bootstrap technique. The various estimators appear to have good sampling properties in most of the situations that are considered. Inference issues such as hypothesis testing and confidence interval construction for the PLTRM are also investigated. In the third main section of the dissertation (Chapter 4), we illustrate the usefulness of the PLTRM, and the application of the proposed estimation methods, by modelling various real-world data sets. These examples demonstrate both the good statistical performance, and the great application potential, of the PLTRM. / Graduate

Estudamos uma classe de campos de vetores seccionalmente lineares no plano denotada por X. Tais campos aparecem frequentemente em modelos matemáticos aplicados à engenharia. Baseados no trabalho de J. Sotomayor e R. Garcia [SG03], impondo condições sobre as singularidades, órbitas periódicas e separatrizes, definimos um conjunto de campos de vetores que são estruturalmente estáveis em X. Provamos que esse conjunto é aberto, denso e tem medida de Lebesgue total em X, o qual é um espaço vetorial de dimensão finita. / We study a class of piecewise-linear vector fields in the plane denoted by X. These vector fields appear often in mathematical models applied to Engineering. Based on Jorge Sotomayor and Ronaldo Garcia paper [SG03], we impose conditions on singularities, periodic orbits and separatrices, to define a set of vector fields structurally stable in X. We give a proof that this set is open, dense and has full Lebesgue measure in X, that is a finite dimensional vector space.

Nesta tese estudamos um dos principais problemas na teoria qualitativa das equações diferenciais planares: o problema de determinar a bacia de atração de um ponto de equilíbrio. Damos uma prova rigorosa de que para sistemas lineares por partes de costura com duas zonas no plano, definidas por matrizes Hurwitz o único ponto de equilíbrio na reta de separação é globalmente assintoticamente estável. Por outro lado, provamos que nesta classe de sistemas, podemos ter um ponto de equilíbrio instável na origem quando uma curva poligonal separa as zonas, levando a um resultado contra-intuitivo do comportamento dinâmico de sistemas lineares por partes no plano. Além disso, estudamos os ciclos limites em perturbações suaves por partes de centros Hamiltonianos. Neste cenário, é comum adaptar resultados clássicos de sistemas suaves, como funções de Melnikov, para sistemas não-suaves. No entanto, existe pouca justificativa para este procedimento na literatura. Ao utilizar o método de regularização damos uma prova que suporta o uso de funções de Melnikov diretamente do problema não-suave original. / In this thesis we study one of the main problems in the qualitative theory of planar differential equations: the problem of determining the basin of attraction of an equilibrium point. We give a rigorous proof that for planar sewing piecewise linear systems with two zones, defined by Hurwitz matrices the unique equilibrium point in the separation straight line is globally asymptotically stable. On the other hand, we prove that sewing piecewise linear systems with two zones in the plane, defined by Hurwitz matrices can have one unstable equilibrium point at the origin allowing a broken line to separate the zones, leading to counterintuitive dynamical behaviors of simple piecewise linear systems in the plane. Furthermore, we study limit cycles in piecewise smooth perturbations of Hamiltonians centers. In this setting it is common to adapt classical results for smooth systems, like Melnikov functions, to non-smooth ones. However, there is little justification for this procedure in the literature. By using the regularization method we give a proof that supports the use of Melnikov functions directly from the original non-smooth problem.

Le but de cette thèse est d'étudier les structures symplectiques dans la catégorie des variétés linéaires par morceaux (PL). La question centrale est de déterminer si toute variété symplectique lisse $(M,omega)$ peut être triangulée de manière symplectique, au sens où il existe une variété linéaire par morceaux $K$ et une triangulation $h :K -> M$ telle que $h^*omega$ est une forme symplectique constante par morceaux. Nous étudions d'abord un problème plus simple, qui consiste à trianguler les formes volumes lisses. Étant donnée une variété lisse $M$ munie d'une forme volume $Omega$, nous montrons qu'il existe une triangulation lisse $h :K -> M$ telle que $h^*Omega$ est une forme volume constante par morceaux. En particulier, les variétés symplectiques lisses de dimension 2 admettent donc des triangulations symplectiques. Étant donnée une variété symplectique fermée $(M,omega)$, nous montrons ensuite que pour certaines triangulations lisses $h :K -> M$, on peut, par une modification arbitrairement petite du complexe $K$, supposer que la forme $h^*omega$ est de rang maximal le long de tous les simplexes de $K$. Ce résultat permet d'approximer arbitrairement bien toute variété symplectique fermée par une variété symplectique PL. Nous nous intéressons finalement au cas d'une sous-variété symplectique $M$ d'un espace ambiant qui admet lui-même une triangulation symplectique. Nous montrons qu'il est possible de construire un cobordisme entre la sous-variété $M$ considérée et une approximation lisse par morceaux de celle-ci, triangulée par un complexe symplectique. / In this thesis, we study symplectic structures in a piecewise linear (PL) setting. The central question is to determine whether a smooth symplectic manifold can be triangulated symplectically, in the sense that there exists a triangulation $h :K -> M$ such that $h^*omega$ is a piecewise constant symplectic form on $K$. We first focus on a simpler related problem, and show that any smooth volume form $Omega$ on $M$ can be triangulated. This means that there always exists a triangulation $h :K -> M$ such that $h^*Omega$ is a piecewise constant volume form. In particular, symplectic surfaces admit symplectic triangulations. Given a closed symplectic manifold $(M,omega)$, we then prove that there exists triangulations $h :K -> M$ for which the piecewise smooth form $h^*omega$ has maximal rank along all the simplices of $K$. This result allows to approximate arbitrarily closely any closed symplectic manifold by a PL one. Finally, we investigate the case of a symplectic submanifold $M$ of an ambient space which is itself symplectically triangulated, and give the construction of a cobordism between $M$ and a piecewise smooth approximation of $M$, triangulated by a symplectic complex. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished