RECURSIVE-IDENTIFICATION USING FEEDFORWARD NEURAL NETWORKS

The paper is concerned with the identification of an unknown nonlinear dynamical system when only the inputs and outputs are accessible for measurement. Under certain assumptions it is shown that, generically, the system can be realized by a recursive input-output model. Furthermore, relying on the approximation properties of neural networks and the existence of effective training algorithms, it is demonstrated how an effective identification model can be constructed. Simulation results are presented to complement the theoretical discussions.