Reinforcement Learning Using Chaotic Exploration in Maze World

Koichiro MORIHIRO; Nobuyuki MATSUI; Haruhiko NISHIMURA

doi:10.11499/sicep.2004.0_1_5

SICE Annual Conference 2004

DOI https://doi.org/10.11499/sicep.2004.0_1_5

会議情報

主催: The Society of Instrument and Control Engineers

共催: The IEEE Industrial Electronics Society, The IEEE Robotics and Automation Society, The IEEE Control Systems Society, The IEEE Systems, Man and Cybernetics Society, The Instrumentation, Systems, and Automation Society

Reinforcement Learning Using Chaotic Exploration in Maze World

*Koichiro MORIHIRO, Nobuyuki MATSUI, Haruhiko NISHIMURA

著者情報

キーワード: Reinforcement Learning, Q-learning, Chaos, Exploration, Shortcut Maze

会議録・要旨集フリー

p. 1

詳細

抄録

In this research, we propose an application of the random-like feature of deterministic chaos for the exploration generator in reinforcement learning. In a no stationary shortcut maze problem, the chaotic exploration generator based on the logistic map gives better performances than the stochastic random one. In order to understand the differences, we investigate the learning structures obtained from the two explorations.

責任著者(Corresponding author)

J-STAGEへの登録はこちら（無料）