Thesis Identificacion de sistemas estocasticos de tiempo continuo utilizando funciones base de Kautz a partir de datos muestreados
Loading...
Date
2023
Authors
Journal Title
Journal ISSN
Volume Title
Program
Magíster en Ciencias de la Ingeniería Electrónica
Campus
Campus Casa Central Valparaíso
Abstract
El modelado de sistemas de tiempo continuo se destaca como un área de gran interés en diversas disciplinas, incluyendo la identificación de sistemas y el control. En algunos escenarios, la estructura del modelo de tiempo continuo de un sistema puede ser definida a priori mediante conocimientos específicos.
No obstante, existen aplicaciones en las cuales la estructura del modelo es desconocida, y en tales casos, se recurre a funciones base para obtener una aproximación del sistema.
En esta tesis, se enfrenta el desafóo de identificar un sistema estocástico de tiempo continuo. El enfoque propuesto implica la utilización de bases de funciones ortonormales de tiempo continuo, con una preferencia específica por la base de funciones de Kautz para aproximar el sistema real. Es importante destacar que, para el proceso de identificación, solo se cuenta con datos muestreados del sistema.
Para abordar el problema de estimación, se ha adoptado un algoritmo de identificación en el dominio de frecuencia centrado en la función de verosimilitud de Whittle. A través de este algoritmo, se busca determinar los parámetros asociados a las funciones base de Kautz que son responsables de la aproximación del sistema.
Con el objetivo de demostrar la efectividad de esta propuesta, se presentan resultados de simulaciones numéricas. Este algoritmo aprovecha datos muestreados derivados del sistema en tiempo continuo, destacándose por su capacidad para identificar con precisión los parámetros subyacentes relacionados con las funciones base de Kautz. Además, se compara este método con la técnica de máxima verosimilitud tradicional en el dominio de la frecuencia. Finalmente, se ilustran los beneficios de esta propuesta mediante su aplicación a un sistema de óptica adaptativa.
Continuous-time system modeling stands out as an area of great interest in various disciplines, including system identification and control. In some scenarios, the structure of the continuous-time model of a system can be predefined through specific knowledge. However, there are applications in which the model structure is unknown, and in such cases, basis functions are employed to obtain an approximation of the system. In this thesis, we tackle the challenge of identifying a continuous-time stochastic system. Our proposed approach involves the use of continuous-time orthonormal basis functions, with a specific preference for the Kautz basis functions to approximate the real system. It is important to note that, for the identification process, we rely exclusively on sampled data of the system. To address the estimation problem, we have adopted a frequency domain identification algorithm centered on the Whittle likelihood function. Through this algorithm, we aim to determine the parameters associated with the Kautz basis functions, responsible for the system’s approximation. In order to demonstrate the effectiveness of our proposal, we present results from numerical simulations. This algorithm leverages sampled data derived from the continuous-time system, standing out for its ability to accurately identify the underlying parameters related to the Kautz basis functions. Additionally, this method is compared with the traditional maximum likelihood technique in the frequency domain. Finally, we illustrate the benefits of our proposal through its application to an adaptive optics system.
Continuous-time system modeling stands out as an area of great interest in various disciplines, including system identification and control. In some scenarios, the structure of the continuous-time model of a system can be predefined through specific knowledge. However, there are applications in which the model structure is unknown, and in such cases, basis functions are employed to obtain an approximation of the system. In this thesis, we tackle the challenge of identifying a continuous-time stochastic system. Our proposed approach involves the use of continuous-time orthonormal basis functions, with a specific preference for the Kautz basis functions to approximate the real system. It is important to note that, for the identification process, we rely exclusively on sampled data of the system. To address the estimation problem, we have adopted a frequency domain identification algorithm centered on the Whittle likelihood function. Through this algorithm, we aim to determine the parameters associated with the Kautz basis functions, responsible for the system’s approximation. In order to demonstrate the effectiveness of our proposal, we present results from numerical simulations. This algorithm leverages sampled data derived from the continuous-time system, standing out for its ability to accurately identify the underlying parameters related to the Kautz basis functions. Additionally, this method is compared with the traditional maximum likelihood technique in the frequency domain. Finally, we illustrate the benefits of our proposal through its application to an adaptive optics system.
Description
Keywords
Estocástico, Whittle, Datos muestreados