RESEARCH ON INTELLIGENT CONTROL AND ITS SAFETY
| Project/Area Number |
03650336
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
計測・制御工学
|
|---|
| Research Institution | THE UNIVERSITY OF TOKYO (1992-1993)
University of Tsukuba (1991) |
|---|
Principal Investigator |
SHIN Seiichi THE UNIV.OF TOKYO, DEPT.ENG., ASSOC.PROF., 工学部, 助教授 (20134463)
|
|---|
| Project Period (FY) |
1991 – 1993
|
| Project Status |
Completed (Fiscal Year 1993)
|
| Budget Amount *help |
¥2,300,000 (Direct Cost: ¥2,300,000)
Fiscal Year 1993: ¥400,000 (Direct Cost: ¥400,000)
Fiscal Year 1992: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 1991: ¥1,100,000 (Direct Cost: ¥1,100,000)
|
|---|
| Keywords | INTELLIGENT CONTROL / SAFETY / ARTIFICIAL INTELLIGENCE / ADAPTIVE CONTROL / NEURAL NETWORK / ROBUSTNESS / LOGIC CIRCUIT / CONTINUOUS ANALOG / 神経回路網 / 人工知能. / ロバスト / 論理図路 |
|---|
| Research Abstract |
The intelligent control system considered here consists of the sequential part and the feedback part. The objective of this research is to seek design and analyzing methods of Intelligent control system from the viewpoint of the continuous analog of logic, which is used in Fuzzy and neural network systems.
We devise the mechanism which estimates the inner states of VLSI circuit from the input and the output pins with a structure of the observer used in the control theory, the steepest decent algorithm used in the neural network technology and the continuous analog which is a key technology in Fuzzy system.
Parallel with this research, we reconsider the learning law used in the neural network, which is a sort of the steepest decent algorithm, from the viewpoint of the theory of robust adaptive control, which is one of the modern control theory. With Taylor expansion, we divide a nonlinear function into a linear part and a remainder part. Since the remainder part is bounded, we improve the steepest decent algorithm with the s modification and a dead zone, which are used in the robust adaptive control. We show boundness of estimates used in these learning law, whereas it is not shown with the ordinary steepest decent learning law.
These results show effectiveness of the control theory and the continuous analog to design and analyze the intelligent system and also show a total way to design and analyze these system. These results contribute to safety intelligent system and to engineers who should treat these complex systems, since there were no way to treat these system theoretically.
|
Report
(4 results)
Research Products
(7 results)
