论文信息 - On a New Type of Neural-Network-Based Input-Output Model: The ANARMA Structure

On a New Type of Neural-Network-Based Input-Output Model: The ANARMA Structure

Abstract In this paper, we present the Additive Nonlinear AutoRegressive Moving Average (ANARMA) structure as an excellent choice for neural-networks-based inputoutput models. The advantage of the ANARMA model is that the time-steps in the argument are pair-wise decomposed, which allows the ANARMA model to be realized in state-space, and to linearize the model via dynamic output feedback. Results of recursive training and feedback linearization of such NN-ANARMA models are presented.

Ülle Kotta | Fahmida N. Chowdhury | Sven Nomm

[1] Ülle Kotta,et al. Two Approaches for State Space Realization of NARMA Models: Bridging the Gap , 2002 .

[2] Fahmida N. Chowdhury. Input-output modeling of nonlinear systems with time-varying linear models , 2000, IEEE Trans. Autom. Control..

[3] Ton J.J. van den Boom,et al. Predictive control based on neural network models with I/O feedback linearization , 1999 .

[4] S. Billings,et al. The Practical Identification of Systems with Nonlinearities , 1985 .

[5] Ülle Kotta,et al. Output Feedback Linearization of Nonlinear Discrete Time Systems , 2000 .