| Project/Area Number |
22500131
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Kyoto Institute of Technology |
|---|
Principal Investigator |
IIMA Hitoshi 京都工芸繊維大学, 工芸科学研究科, 准教授 (70273547)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KUROE Yasuaki 京都工芸繊維大学, 工芸科学研究科, 教授 (10153397)
|
|---|
| Project Period (FY) |
2010 – 2012
|
| Project Status |
Completed (Fiscal Year 2012)
|
| Budget Amount *help |
¥4,290,000 (Direct Cost: ¥3,300,000、Indirect Cost: ¥990,000)
Fiscal Year 2012: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2011: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2010: ¥2,600,000 (Direct Cost: ¥2,000,000、Indirect Cost: ¥600,000)
|
|---|
| Keywords | 強化学習 / PSO / 群知能 / Particile Swarm Optimization / Particle Swarm Optimization |
|---|
| Research Abstract |
We proposed swarm reinforcement learning methods based on particle swarm optimization (PSO) for acquiring optimal policies rapidly, and applied the proposed methods to some complicated reinforcement learning problems such as ones with continuous state-action space. In the proposed method, multiple sets of an agent and an environment, which are called learning worlds, are prepared, and agents in each learning world learn not only by individually using a usual reinforcement learning method but also through exchanging information among the learning worlds by using the update equations of PSO.
|
