Stable sequential identification of continuous nonlinear dynamical systems by growing radial basis function networks
- 1 September 1996
- journal article
- research article
- Published by Taylor & Francis in International Journal of Control
- Vol. 65 (1) , 53-69
- https://doi.org/10.1080/00207179608921686
Abstract
This paper presents a sequential identification scheme for continuous nonlinear dynamical systems using neural networks. The nonlinearities of the dynamical systems are assumed to be unknown. The identification model is a gaussian radial basis function neural network that grows gradually to span the appropriate state space and of sufficient complexity to provide an approximation to the dynamical system. The sequential identification algorithm for continuous dynamical nonlinear systems is developed in the continuous-time framework instead of in discrete time. The approach, different from the conventional methods of optimizing a cost function, attempts to ensure stability of the overall system while the neural network learns the system dynamics. The stability and convergence of the overall identification scheme are guaranteed by parameter adjustment laws developed using the Lyapunov synthesis approach. To ensure that the modelling error can be reduced arbitrarily, a one-to-one mapping is proposed so that the states and inputs of the system are transferred into compact sets. The operation of the sequential identification scheme is illustrated using simulated experimental results.Keywords
This publication has 20 references indexed in Scilit:
- A Function Estimation Approach to Sequential Learning with Neural NetworksNeural Computation, 1993
- Application of feedforward neural networks to dynamical system identification and controlIEEE Transactions on Control Systems Technology, 1993
- A perceptron network for functional identification and control of nonlinear systemsIEEE Transactions on Neural Networks, 1993
- Gaussian networks for direct adaptive controlIEEE Transactions on Neural Networks, 1992
- A Resource-Allocating Network for Function InterpolationNeural Computation, 1991
- Regularization Algorithms for Learning That Are Equivalent to Multilayer NetworksScience, 1990
- Backpropagation through time: what it does and how to do itProceedings of the IEEE, 1990
- Non-linear system identification using neural networksInternational Journal of Control, 1990
- Fast Learning in Networks of Locally-Tuned Processing UnitsNeural Computation, 1989
- A Learning Algorithm for Continually Running Fully Recurrent Neural NetworksNeural Computation, 1989