An improved adaptive identifier for discrete multivariable linear systems

It is shown that the Kudva-Narendra method for recursive identiffication of discrete-time linear multivariable systems may be modified to allow possibly improved convergence during the initial stages, finite convergence, or the use of a priori information about the parameter estimates. The modification entails orthogonal projection of the vector of system states and inputs used for excitation of the adaptation equations and does not impair the stability of the original identifier. Examples illustrating various aspects of the modification are given.