Appl. Math. Inf. Sci. 8, No. 6, 3055-3062 (2014) 3055
Applied Mathematics & Information Sciences
An International Journal
http://dx.doi.org/10.12785/amis/080646
Discrete Adaptive Sliding Mode Control via Wavelet
Network for a Class of Nonlinear Systems
Xiaoyu Zhang
∗
College of Electronic and Information Engineering, North China Institute of Science and Technology, Yanjiao Box206, 101601 Beijing,
China
Received: 16 Nov. 2013, Revised: 14 Feb. 2014, Accepted: 15 Feb. 2014
Published online: 1 Nov. 2014
Abstract: Unmodelled dynamics and perturbations are always immeasurable. In this paper, an adaptive sliding mode control (ASMC)
based on wavelet network (WN) for a class of non-affine multi-variable nonlinear discrete systems is presented in order to compensate
them. Wavelet network which parameters are tuned on-line is adopted to realize the equivalent control, and hitting controls are added
in order to satisfy reaching conditions. By combining the adaptive WN with SMC strategy, the constructed control law has many
advantages such as robustness, adaptive characters, and the precise mathematic models of controlled plants are not required. Finally,
experiment on an inverted pendulum control system based on the proposed control design method is given to verify its effectiveness
and performance.
Keywords: Discrete, nonlinear system, sliding mode control, adaptive, wavelet network
1 Introduction
In practical control systems, highly unknown
uncertainties, disturbances and nonlinearities always
exist. Many efforts on this problem have been made by
researchers of robust control, adaptive control and
intelligent control etc. In recent years, wavelet network
(WN) is used as a powerful tool for signal and data
processing, time-series analysis and the approximation of
arbitrary unknown functions such as literatures
[
1],[2],[3],[4] and [5]. Using WN for function
approximation and identification of nonlinear systems has
been studied by literatures[
1],[2],[3] etc.
Adaptive neural network control has been widely
investigated by many researchers such as literatures
[
6],[7],[8],[9] and [10]. The parameters of neural network
(NN) are tuned on-line to approximate the unknown
nonlinear dynamic. However the precision depends on the
structure selection which is a difficult problem at present.
Inspired by the theory of adaptive NN, adaptive wavelet
network methods are reported dealing with on-line
application in the control problem of dynamic nonlinear
systems, it refers to literatures [
11]-[18]. WN can be
regarded as a class of NN, but it has its special
characteristics such as the linearity in parameter space
and the orthonormality. These make WN is suitable for
on-line estimating, and there is not the problem of
structure selection in adaptive wavelets networks.
Therefore, successful application of adaptive WN to
nonlinear systems is researched[
17]. Based on the
conception of multi-resolution approximation (MRA),
WN is a three-layer network consisting of orthonormal
father wavelets and mother wavelets. Because of the
orthonormal property, it is possible to regulate the
network structure and parameters on-line. Moreover, the
multi-resolution approximation ensures that the
approximation precision can be improved quickly as
resolution increases. Although the precision can be
improved arbitrarily, there exist many perturbation and
disturbance that impact on the system performance such
as stability, steady-state error and so on.
Sliding mode control (SMC) theory has been proved
to be an effective way to control nonlinear dynamic
system with strong robustness[
19][20]. If we combine
SMC theory into the adaptive wavelet network, the
designed controller will possess many advantages. The
wavelets neural network control (WNNC) based on SMC
control theory and adaptive theory for the linear motor
and induction motor drive has been studied by [
15][18].
Nevertheless, their research works aim at special plants.
∗
Corresponding author e-mail:
ysuzxy@aliyun.com
c
2014 NSP
Natural Sciences Publishing Cor.