Subspace identification of bilinear systems

Abstract

In this paper, we describe subspace algorithms for the identification of general bilinear MIMO systems. The results are a generalization of a previously presented paper that considered bilinear systems with white inputs [9]. They are also a direct extension of linear subspace identification. To illustrate the practical use of the method, it has been compared with linear subspace methods when applied to the same data obtained from a test setup of an industrial winding process. Keywords: bilinear systems, subspace identification 1 Introduction The interest of studying bilinear models is quite obvious. On the one hand they are more accurate than linear models while on the other hand they are still much easier to study in an analytical way than general nonlinear systems. Moreover there exist fare going analogies between linear and bilinear system theory [1]. The state space equations of a general MIMO bilinear system are given by: x k+1 = Ax k +Nu k\\Omega x k +Bu k + w k (1) y k = Cx...