Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  

Parametric Methods for Nonlinear System Identification

Back
Abstract:
NARMAX (Nonlinear AutoRegressive, Moving Average eXogenous) models are parametric representations of non-linear systems. Parametric descriptions, while theoretically attractive, have been difficult to apply since basic questions in parametric identification have been left open. Specifically, these open questions have been: (1) how to estimate the order of the input-output lag and nonlinearity (model order selection) (2) how to select which parameters to include in the model (structure detection) and, (3) how to model and estimate parameters of "hybrid" or "multimode" systems. To tackle the above NARMAX identification problems I have developed (1) a bootstrap model order selection (BMOS) algorithm, (2) a bootstrap structure detection (BSD) algorithm and (3) a modified extended least squares (MELS) algorithm to estimate the coefficients of hybrid or multimode systems. In this presentation, I will demonstrate that the bootstrap is a useful statistical tool for identification of parametric nonlinear systems, and that the developed bootstrap model order selection and bootstrap structure detection algorithms are robust methods for selecting the order and structure of NARMAX models, and are resistant to noise. Moreover, application of the MELS algorithm to the vestibulo-ocular reflex (VOR) system will demonstrate that this algorithm is a robust method for estimating the coefficients of hybrid systems.

Type of Seminar:
Public Seminar
Speaker:
Dr. Sunil Kukreja
McGill University, Montréal, Canada
Date/Time:
Apr 13, 2000   10:45
Location:

ETH-Zentrum, ETL K 25, Physikstrasse 3, 8006 Zurich
Contact Person:

Dr. Saso Jezernik
No downloadable files available.
Biographical Sketch:
Sunil L. Kukreja received his B.S. degree in Electrical Engineering in 1993 from the Johns Hopkins University in Baltimore, U.S.A. In 1996 he received his M.Eng. degree in Biomedical Engineering from McGill University, Montréal, Canada and is currently finishing his Ph.D. degree at McGill University in Biomedical Engineering. His main interests are in the areas of linear and nonlinear system identification of complex biological systems, signal processing, simulation & modeling, and hybrid systems.