1995 Fiscal Year Final Research Report Summary
Research on Control System based on Petri Network Brain Model
| Project/Area Number |
06452256
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
計測・制御工学
|
|---|
| Research Institution | KYUSHU UNIVERSITY |
|---|
Principal Investigator |
HIRASAWA Kotaro Kyushu University, Faculty of Engineering, Professor, 工学部, 教授 (70253474)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
OHBAYASHI Masanao Kyushu University, Faculty of Engineering, Assistant, 工学部, 助手 (60213849)
MURATA Junichi Kyushu University, Faculty of Engineering, Assistant Professor, 工学部, 助教授 (60190914)
|
|---|
| Project Period (FY) |
1994 – 1995
|
| Keywords | Petri Network / Brain Science / Control System / Learning / Self Organization / Neural Network / R.B.F. / Fuzzy |
|---|
| Research Abstract |
Functions of the large-scale complicated control systems are similer to the functions of the brain in the points of recognizing and controlling environments.
The purpose of this research is to establish the control methodology which has the ability of human's judgment using the knowledge of recently developed brain science.
Functions of the brain are distributed. First, the fundamental model for controlling large-scale complicated systems is studied considering the functions distribution of the brain and introducing the ability of learning.
The model is named LPN (Learning Petri Network) because it is based on Petri Network and the learning ability is attached to it.
It is the most important feature that the specific learning algorithms with teacher and self-orgnizing scheme are introduced in LPN in order to realize the functions distribution.
Next, simulations are carried out to investigate the peformance of LPN for the problem of pattern recognition of nonlinear discontinuous functions, system identification and control of nonlinear dynamics systems.
From simulations, it has been shown that LPN has the ability of selecting appropriate routes in the network depending on the value of inputs, in other words, the ability of the functionsdistribution. It has been also shown that LPN is superior in performance to Neural Network in the point of forming nonlinear discontinuous functions and identifying and controlling nonlinear dynamic systems.
Therefore, a new control methodology based on the functions distribution and the learning ability of human brain has been established.
This new control methodology can be applied to not only conventioned process control systems but also large-scale complicated control systems which require sophisticated and advanced control functions.
|
