Computing the structural indices in an on-line algorithm for multivariable system identification

An on-line identification software for multi input-multi output, stochastic, linear systems is presented. It combines an extension of the modified maximum likelihood method for S. I. S. O. systems[1] to perform the parameter estimation, and an algorithm, based on the Range -- Error Test[2] for the structural identification of the system. The program developed can be used as an on-line facility because it determines the structure without any a priori knowledge about the plant. It also checks the validity of the structure of the actual model being identified and recomputes the structural indices if necessary.