DETERMINISTIC LEARNING OF NONLINEAR DYNAMICAL SYSTEMS
- 1 April 2009
- journal article
- research article
- Published by World Scientific Pub Co Pte Ltd in International Journal of Bifurcation and Chaos
- Vol. 19 (04) , 1307-1328
Abstract
In this paper, we investigate the problem of identifying or modeling nonlinear dynamical systems undergoing periodic and period-like (recurrent) motions. For accurate identification of nonlinear dynamical systems, the persistent excitation condition is normally required to be satisfied. Firstly, by using localized radial basis function networks, a relationship between the recurrent trajectories and the persistence of excitation condition is established. Secondly, for a broad class of recurrent trajectories generated from nonlinear dynamical systems, a deterministic learning approach is presented which achieves locally-accurate identification of the underlying system dynamics in a local region along the recurrent trajectory. This study reveals that even for a random-like chaotic trajectory, which is extremely sensitive to initial conditions and is long-term unpredictable, the system dynamics of a nonlinear chaotic system can still be locally-accurate identified along the chaotic trajectory in a deterministic way. Numerical experiments on the Rossler system are included to demonstrate the effectiveness of the proposed approach.Keywords
This publication has 24 references indexed in Scilit:
- Learning From Neural ControlIEEE Transactions on Neural Networks, 2006
- Direct adaptive NN control of a class of nonlinear systemsIEEE Transactions on Neural Networks, 2002
- System Identification1999
- Stability and approximator convergence in nonparametric nonlinear adaptive controlIEEE Transactions on Neural Networks, 1998
- Identifying chaotic systems via a Wiener-type cascade modelIEEE Control Systems, 1997
- Stable sequential identification of continuous nonlinear dynamical systems by growing radial basis function networksInternational Journal of Control, 1996
- Gaussian networks for direct adaptive controlIEEE Transactions on Neural Networks, 1992
- Identification and control of dynamical systems using neural networksIEEE Transactions on Neural Networks, 1990
- Interpolation of scattered data: Distance matrices and conditionally positive definite functionsConstructive Approximation, 1986
- Detecting strange attractors in turbulence1981