1998 Fiscal Year Final Research Report Summary
NEUROCONTROLLER EVALUATED BY GAS WITH TIMEVARYING ROUGH EVALUATIONS
| Project/Area Number |
09650295
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent mechanics/Mechanical systems
|
|---|
| Research Institution | UNIVERSITY OF THE RYUKYUS |
|---|
Principal Investigator |
YAMAMOTO Tetsuhiko UNIVERSITY OF THE RYUKYUS,FACULTY OF ENGINEERING,PROFESSOR, 工学部, 教授 (20045008)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
NAKAZONO Kunihiko 琉球大学, 工学部, 助手 (80284959)
YAN Yuling 琉球大学, 工学部, 助教授 (00295297)
KINJO Hiroshi 琉球大学, 工学部, 助教授 (50211206)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Keywords | NEURAL NET / GENETIC ALGORITHMS / INVERTED PENDULUM / CRANE / SWING SUPPRESSION / ARTIFICIAL LIFE / ROUGH EVALUATION |
|---|
| Research Abstract |
This report is an investigation of a new controller design method that requires little knowledge of control theories (analysis and synthesis). Mathematical physics provides strict design methods, which ensures their application limits. However, much time and effort are required for ordinary engineers to master these control theories based on mathematical physics. This investigation reveals a very easy and convenient design method in which neurocontrollers are automatically generated and evolved via genetic algorithms (GAs). GAs require evaluations of the controllability of neurocontrollers to apply the principle of survival of the fittest. We adopted very simple evaluations for this aim. The utility of the proposed method is shown in several simulations : control of an inverted pendulum, control of a bicycle, up-swing and inverted control of a pendulum, and suppression of load swing of cranes. One of the cranes considered is a traverse type in which the rope length changes, and the other is a jib type with a floating base. In all cases, controllable neurocontrollers are generated in three to five minutes. The controllability of the neurocontrollers is superior to that of the linear quadratic regulator.
|
Research Products
(4 results)
