| Project/Area Number |
09450106
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent mechanics/Mechanical systems
|
|---|
| Research Institution | Kobe University |
|---|
Principal Investigator |
UEDA Kanji Kobe University, Faculty of engineering, Professor, 工学部, 教授 (50031133)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
OHKURA Kazuhiro Kobe University, Faculty of engineering, Research Associate, 工学部, 助手 (40252788)
MANABE Keishi Kobe University, Faculty of engineering, Research Associate, 工学部, 助手 (90209677)
ヤリ ワーリオ ノキアジャパン, ノキアリサーチセンタ, 研究室長
VAARIO Yari Nokia Research Center, Japan, Senior Research Manager
SVININ Mikha 神戸大学, 工学部, 助教授 (90274125)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Project Status |
Completed (Fiscal Year 1998)
|
| Budget Amount *help |
¥4,800,000 (Direct Cost: ¥4,800,000)
Fiscal Year 1998: ¥1,300,000 (Direct Cost: ¥1,300,000)
Fiscal Year 1997: ¥3,500,000 (Direct Cost: ¥3,500,000)
|
|---|
| Keywords | Reinforcement learning / Autonomous adaptation system / Autonomous locomotive robot / Specialization of functions / Self organization / 知的ロボット / 自律ロボット / 適応システム / 自己創出 / 人工生命 / 進化システム |
|---|
| Research Abstract |
In this research, to develop self-emergent autonomous adaptation systems, we develop two autonomous robot systems as follows :
(1) To obtain a gait of a robot systems with four legs, we developed controllers of the robot systems with four controllers assigned to each legs as an asynchronous autonomous agent, and applied CSCG (Continuous Space Classifier Generator) to each agent. To evaluate the controllers, we applied the robot systems to the task that the robot systems should be moved to a light source, and confirmed that the robot systems can be obtain a gait through trial and error.
(2) To develop a control system for two AGVs connected by a link, we developed the control systems with two autonomous agents of each AGV and try to obtain control rules for each AGV by using reinforcement learning. In general, since it is difficult to apply the dynamic problems just like the problem in this research, we combined the predictor by using SLA (Stochastic Learning Automata) with the reinforcement learning. To evaluate the controller, we apply the controller to the task that the two AGVs move along the wall in a room, and confirm the cooperative behavior of the two AGVs by using the controllers proposed in this paper.
|
