This paper presents a formulation to identify the input-output maps of discrete- time bilinear state-space models. The new method uses input-output data from one experiment, and the input data need not be of special types such as extended pulses. The initial conditions can be non-zero and unknown. Recent methods, on the other hand, require data from multiple experiments, specialized input signals, and zero initial conditions. In the present method, a Nonlinear Auto-Regressive model with eXogenous input (NARX) is identified and used to predict the bilinear system response directly without returning to the bilinear state-space format. Numerical examples are provided to illustrate the identification method.
展开▼