Identification of Linear Stochastic Plants from Incomplete Noisy Measurement of the State-Vector

An observable performance criterion is proposed permitting the use of the adjustable model approach in identification of a transition matrix identification problem when only incomplete noisy measurement data on the current state of the noise-disturbed plant are available. The only requirement to actual use of the method is prior knowledge of the observation matrix. A practical identifiability condition is given for transition matrix entries, which does not call for transforming the matrix into companion or block-companion form and the experimental results are presented.