Multiple Model-Based Reinforcement Learning

We propose a modular reinforcement learning architecture for non-linear, nonstationary control tasks, which we call multiple model-based reinforcement learning (MMRL). The basic idea is to decompose a complex task into multiple domains in space and time based on the predictability of the environmental dynamics. The 1 

• Type: Journal
• Year: 2002
• Where: NECO
• Authors: Kenji Doya, Kazuyuki Samejima, Ken-ichi Katagiri, Mitsuo Kawato
• Keywords: Model-based Reinforcement, Modular Reinforcement, NECO 2002, Nonstationary Control Tasks