Algorithms in system identification

Abstract

A first set of FORTRAN programs was written that generates noisy input output data from a multi-input, multi-output, linear, time-invariant system. A random number generator is used to produce the output noise. The system parameters are identified from the noisy input-output data using an adaptation of the least squares method in the second set of programs. The third set presents a new algorithm for system parameter identification using the concept of the supremal [F, G]-invariant subspace in ker[H], and some results from optimal control theory. Most of the results are calculated and tabulated for a specific two-input, two-output, third order system.