| Presentation | 2004/3/12 A belief-state reinforcement learning scheme for a multi-agent card game Hajime FUJITA, Shin ISHII, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In this report, we deal with the card game "Hearts", an instance of decision making problems in partially observable situations. We present the state estimation method based on a sampling method, and a reinforcement learning (RL) scheme for a multi-agent environment using the state estimation method. Since there are often a lot of unobservable cards in this game, RL is dealt with in the framework of a partially observable Markov decision process (POMDP). Using pessimistic observations, the learning agent focuses on an important domain of a large state space, estimates unobservable states based on a sampling method from such a subspace, and makes a decision by predicting the environmental behavior. Simulation results show that our model-based POMDP-RL method with sampling state estimation is applicable to this realistic multi-agent problem. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | sampling method / POMDP / estimation of unobservable state variables / multi-agent / reinforcement learning (RL) |
| Paper # | NC2003-205 |
| Date of Issue |
| Conference Information | |
| Committee | NC |
|---|---|
| Conference Date | 2004/3/12(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Neurocomputing (NC) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A belief-state reinforcement learning scheme for a multi-agent card game |
| Sub Title (in English) | |
| Keyword(1) | sampling method |
| Keyword(2) | POMDP |
| Keyword(3) | estimation of unobservable state variables |
| Keyword(4) | multi-agent |
| Keyword(5) | reinforcement learning (RL) |
| 1st Author's Name | Hajime FUJITA |
| 1st Author's Affiliation | Nara Institute of Science and Technology() |
| 2nd Author's Name | Shin ISHII |
| 2nd Author's Affiliation | CREST, Japan Science and Technology Agency |
| Date | 2004/3/12 |
| Paper # | NC2003-205 |
| Volume (vol) | vol.103 |
| Number (no) | 734 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |