Coherent Behavior in Multiagent System Based on Reinforcement Learning
Anton Kabysh, Vladimir A. Golovko · 2010
This paper covers area of Collective Reinforcement Learning. We introduce and describe new simple approach to Collective Reinforcement Learning named Related Temporal Difference. This approach can supports coherence of agent’s behavior in distributed and structurally complicated multi*agent system. We construct a decentralized Multi*Agent system which describes behaviors of multi*joint robot. Given experiments show, that system of local learning procedures in complex system can be much faster than learning system on the whole.