A Superspace Method for Discrete-Time Bilinear Model Identification by Interaction Matrices

Abstract

This paper presents a method for discrete-time bilinear state-space model identification from a single set of input–output data. The initial state can be unknown. By extending the interaction matrix formulation from linear to bilinear state-space models, the bilinear system state is expressed in terms of input–output measurements. This relationship is used to convert the bilinear model to an Equivalent Linear Model (ELM) which can be identified by any linear identification method such as a superspace method presented here. The bilinear state-space model is then extracted from the identified ELM.