Login

 


Plenary Lecture

Neuro-Controller Using Simultaneous Perturbation

Professor Yutaka Maeda
Faculty of Engineering Science
Kansai University
Japan
E-mail: maedayut@kansai-u.ac.jp

Abstract: The simultaneous perturbation optimization method is a stochastic gradient method which uses only values of an objective function to find a optimal point of the function. The optimization method was introduced by J. C. Spall. Y. Maeda also independently proposed a learning rule using the simultaneous perturbation for artificialneural networks and reported a feasibility of the learning rule. At the same time, the merit of the learning rule was demonstrated in the hardware implementation of neural networks.
The important advantage of the simultaneous perturbation method is its simplicity. The simultaneous perturbation can estimate the gradient of a function using only the two values of the function. Therefore, it is relatively easy to implement even for recurrent neural networks, compared with the back-propagation learning rule. Moreover, when we use a neural network as a controller, without using Jacobian of an objective plant, it is possible to apply the learning scheme directly.
This paper presents a trajectory control for a SCARA robot using a recurrent neural network.We adopt the simultaneous perturbation optimization method as a learning rule of the recurrent neural network. This neuro-controller learns the inverse dynamics of the SCARA robot. We describedetails of the control scheme. Moreover, we consider an example for the circular path control. Someresultsfor an actual SCARA robot are shown.

Brief Biography of the Speaker: Yutaka Maeda received the B.E., M.E. and D.E. (Doctor of Engineer) degrees in Electronic Engineering from Osaka Prefecture University in 1979, 1981 and 1990, respectively. He joined KANSAI University, Faculty of Engineering in 1987, where heis a Professor of the Faculty of Engineering Science, and a Vice president of Kansai University.
He was a Visiting Researcher in Electrical and Computer Engineering Department, University of California at Irvine, USA in 1995.Hehas established the Electronic Control Laboratory in Kansai University. The laboratoryis producing promising graduates for many industrial fields. Recent research interests in this laboratory are in the areas of soft computing; artificial neural networks, fuzzy theory for robot control, moreover, he is also interesting in the control theory and signal processing related to the simultaneous perturbation optimization. He is also author of about 80 papers in international journals and conference proceedings, and book chapters.

Bulletin Board

Currently:

The Conference Program is online.

The Conference Guide is online.

The paper submission deadline has expired. Please choose a future conference to submit your paper.


WSEAS Main Site


NAUN Main Site

Publication Ethics and Malpractice Statement