Title :
State variable approach to neuro-identification
Author :
Tanomaru, Julio ; Takahashi, Yoshizo
Author_Institution :
Fac. of Eng., Tokushima Univ., Japan
Abstract :
In this paper, a nonlinear systems identifier consisting of an association between a recurrent multilayer neural network (MNN) and a feedforward one is proposed. The basic idea is to use feedforward MNNs to emulate the nonlinear mappings of the extended state variable model of the system to be identified. A simple training algorithm which do not require additional memory is developed. Preliminary simulation results illustrate the effectiveness of the proposed neuro-identifier, which potentially has wide applicability in the signal processing and control fields, especially due to the fact that a quantity characterizing the "state" of the system is produced as a by-product of the identification process.
Keywords :
feedforward neural nets; identification; learning (artificial intelligence); nonlinear systems; recurrent neural nets; feedforward neural networks; learning algorithm; neural control; neuro-identification; nonlinear mappings; nonlinear systems identifier; recurrent multilayer neural network; signal processing; state variable model; Artificial neural networks; Cities and towns; Electronic mail; Multi-layer neural network; Neural networks; Nonlinear control systems; Nonlinear systems; Process control; Recurrent neural networks; Signal processing algorithms;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.717010
