| Autor: |
Yann-Michaël De Hauwere, Peter Vrancx, Ann Nowe |
| Přispěvatelé: |
Tumer, Kagan, Yolum, Pinar, Sonenberg, Liz, Stone, Peter, Computational Modelling, Vrancx, Peter, Knudson, Matt, Grzes, Marek, Informatics and Applied Informatics |
| Jazyk: |
angličtina |
| Rok vydání: |
2011 |
| Předmět: |
|
| Zdroj: |
Vrije Universiteit Brussel |
| Popis: |
Recent research has demonstrated that considering local interactions among agents in specific parts of the state space, is a successful way of simplifying the multi-agent learning process. By taking into account other agents only when a conflict is possible, an agent can significantly reduce the state-action space in which it learns. Current approaches, however, consider only the immediate rewards for detecting conflicts. This restriction is not suitable for realistic systems, where rewards can be delayed and often conflicts between agents become apparent only several time-steps after an action has been taken. In this paper, we contribute a reinforcement learning algorithm that learns where a strategic interaction among agents is needed, several time-steps before the conflict is reflected by the (immediate) reward signal. To do this, we make use of statistical information about the future returns and the state information of the agents. This allows the agent to determine when it should expand its state representation with information on the other agents and when it can safely rely on its own state information. We apply our method to a set of representative grid world problems and show that with our approach, agents successfully manage to expand their state information to solve delayed coordination problems. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
