Transfer function matrix identification

Abstract

A method is presented for identifying the transfer function matrix of a multi-input multi-output finite-dimensional continuous linear time invariant system from input-output observations in the presence of arbitrary initial conditions. A simple and efficient scheme is provided to handle these unknown initial conditions. The method avoids signal differentiation through the use of multiple integration. A set of equations, linear in the unknown parameters of the system transfer function matrix, are formed using the integrated values of inputs and outputs for different values of time. These equations are modified so as to yield an efficient method for computing the system parameters in the presence of additive measurement noise.