Next: Nonlinear System Identification Up: SYSTEMS AND CONTROL Previous: Large Deviation Laws

Nonlinear System Identification

Authors: [tex2html_wrap4422]D. Levanony, N. Berman (Dept. of Mechanical Eng., Ben Gurion University, Israel)

Investigator username: levanony

Category: systems and control theory


Recursive identification of non-linear systems via (truncated) Volterra series models is investigated. Besides a detailed analysis of asymptotic properties of a stochastic gradient algorithm (a generalization of linear results), the implications of the degree of the polynomial model on the resulting asymptotic properties are closely examined. This study indicates that, in some cases of interest, an attempt to improve estimation accuracy by increasing the degree of the polynomial model, might lead to poorer asymptotic estimates.