Neural network interval observer for nonlinear descriptor systems with the optimized hidden node distribution

Abstract

For nonlinear descriptor systems, this article presents a novel neural network interval observer. The framework of this interval observer is based on the combination of monotone systems theory and radial basis function neural networks, as the expansion of the definition of the general interval observer. Additionally, the K-means algorithm is used in the design process of the neural network interval observer to achieve the optimal distribution of hidden nodes, thus avoiding the drawbacks of traditional neural network methods. The number of hidden nodes is significantly decreased, and the persistence of excitation conditions is secured. Also, the interval estimation and convergence performance of the observer can be obtained. Eventually, the validity of the proposed method is illustrated by a numerical example.