Neural networks output feedback controllers using nonlinear parametric wavelet functions

Abstract

Purpose -The purpose of this paper is to propose an adaptive output feedback controller using wavelet neural networks with nonlinear parameterization for unknown nonlinear systems with only system output measurement. Design/methodology/approach -An error observer is used to estimate the tracking errors through output measurement information, and the wavelet neural networks are utilized to online approximate an unknown control input by adjusting their internal parameters. Findings -The controller integrates an error observer and wavelet neural networks with nonlinear parameterization into adaptive control design and is derived in accordance with implicit function and mean value theorem. The adjustment mechanism for the parameters of the wavelet neural networks can be derived by means of mean value theorem and Lyapunov theorem, and the stability of the closed-loop system can be guaranteed. Originality/value -This paper utilizes the nonlinear parametric wavelet neural networks with estimate state inputs to obtain the adaptive control input for nonaffine systems with only system output measurement, and the nonlinear wavelet parameters can be adjusted efficiently.

RLSESN‐based PID adaptive control for a novel wearable rehabilitation robotic hand driven by PM‐TS actuators

“…However, in practice, it is difficult to obtain a satisfactory performance when applying a PID controller in a high‐precision trajectory tracking task. With the development of adaptive robust control (Boukattaya et al , 2011), fuzzy control (Chang et al , 2009; Jain et al , 2010), neural networks (Liu et al , 2008; Leu and Hong, 2010) and some other artificial intelligence approaches (Zhong, 2008), precise tracking control problems are partially solved by using these theories and technologies. Because ANNs have self‐organization, self‐learning, adaptive ability, the intelligent control approaches based on ANNs have become one of the most important ways to get small tracking error and good robustness.…”