Auto-structuring fuzzy neural system for intelligent control

An auto-structuring fuzzy neural network-based control system (ASFNS), which includes the auto-structuring fuzzy neural network (ASFNN) controller and the supervisory controller, is proposed in this paper. The ASFNN is used as the main controller to approximate the ideal controller and the superviso...

Full description

Saved in:
Bibliographic Details
Published in:Journal of the Franklin Institute 2009-04, Vol.346 (3), p.267-288
Main Author: Cheng, Kuo-Hsiang
Format: Article
Language:English
Subjects:
Adaptative systems
Applied sciences
Computer science
control theory
systems
Computer systems and distributed systems. User interface
Control theory. Systems
Exact sciences and technology
Fuzzy neural network
Intelligent control
Lyapunov theorem
Modelling and identification
Parameter learning
Process control. Computer integrated manufacturing
Software
Structure learning
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:An auto-structuring fuzzy neural network-based control system (ASFNS), which includes the auto-structuring fuzzy neural network (ASFNN) controller and the supervisory controller, is proposed in this paper. The ASFNN is used as the main controller to approximate the ideal controller and the supervisory controller is incorporated with the ASFNN for coping with the chattering phenomenon of the traditional sliding-mode control. In the ASFNS, an automatic structure learning mechanism is proposed for network structure optimization, where two criteria of node-adding and node-pruning are introduced. It enables the ASFNN to determine the nodes autonomously while ensures the control performance. In the ASFNS, all the parameters are evolved by the means of the Lyapunov theorem and back-propagation to ensure the system stability. Thus, an intelligent control approach for adaptive control is presented, where the structure and parameter can be evolved simultaneously. The proposed ASFNS features the following salient properties: (1) on-line and model-free control, (2) relax design in controller structure, (3) overall system stability. To investigate the capabilities, the ASFNS is applied to a kind of nonlinear system control. Through the simulation results the advantages of the proposed ASFNS can be validated.
ISSN:0016-0032
1879-2693
DOI:10.1016/j.jfranklin.2008.09.003