On Gradient-Based Search for Multivariable System Estimates

Authors:	Wills Adrian, University of Newcastle, Australia Ninness Brett, University of Newcastle, Australia Gibson Stuart, University of Newcastle, Australia
Topic:	1.1 Modelling, Identification & Signal Processing
Session:	Identification of Multivariable Systems
Keywords:	system identification, parameter estimation, state-space models

Abstract

This paper addresses the design of gradient search based algorithmsfor multivariable system estimation. In particular, the work hereconsiders so-called `full parametrization' approaches, and establishesthat the recently developed `Data Driven Local Coordinate' (DDLC)methods can be seen as a special case within a broader class oftechniques that are designed to deal with rank-deficient Jacobians.This informs the design of a new algorithm that, via a strategy ofdynamic Jacobian rank determination, is illustrated to offer enhancedperformance.