Abstract
This paper presents a procedure to identify inhomogeneous bilinear systems in the distinguishable state space representations from sequences of input and noise contaminated output measurements. The identification procedure is the following; at first, some distinguishable description models are derived from the zerostate response of the systems, then the zerostate input-output maps in the model are estimated by using the baise compensated least square method. Next, the internal discription model is realized by using the previously estimated maps.
The realization algorithm is based on the singular value decomposition of the reachability and distinguishability matrices. Realized the distinguishable canonical forms, the minimal realization systems are obtained to reduce to unreachable states from these models.
Therefore, the proposed methods do not require the factorization of the generalized Hankel matrices. The presented methods are effective as also shown by some numerical examples.
