IDENTIFICATION OF ECONOMICALLY PARAMETRIZED SYSTEMS.

Abstract

The problem of identifying a partially known linear, time invariant system is considered where the unknowness is that associated with a limited number of physical components comprising the system or with physical parameters affecting part of the system. This unknowness translates to structural conditions on the system transfer function or a state variable representation of the system, and the associated identification problem is multilinear in the unknown parameters. Algorithms which use measurements of the input and output and knowledge of the polynomial coefficients of the multilinear combinations of the system parameters are then described. Persistence of excitation conditions on the input for computability of these algorithms are derived, and uniform asymptotic stability under a variety of settings established.