论文信息 - Reinforcement Learning of Multi-Link Robot with Fuzzy ART Neural Networks for State-Space Segmentation - 字舞流文

Reinforcement Learning of Multi-Link Robot with Fuzzy ART Neural Networks for State-Space Segmentation

強化学習の状態判別器にFuzzy ARTを適用して漸増的に状態空間を拡張させることで学習速度が向上でき, さらにハードウエア資源を節約できることを示した. また, 新たな状態に対応するカテゴリーユニットを作成する際に, 類似した既存ユニットから状態価値を引き継ぐことで, 学習性能がより向上することをシミュレーションならびに実験によって示した.

Koichi Okuda | Masayuki Nunobiki | Syunsuke Maeda | K. Okuda | M. Nunobiki | S. Maeda

[1] Yoshihiro Takita,et al. An Investigation of Climbing up Stairs for Inchworm Robot. , 1998 .

[2] Stephen Grossberg,et al. Fuzzy ART: Fast stable learning and categorization of analog patterns by an adaptive resonance system , 1991, Neural Networks.

[3] Koichi Okuda,et al. Returning Motion from the State of Falling Sideways for Articulated Mobile Robot , 2004 .

[4] Yoshihiro Takita,et al. Study on the gaits of an inchworm robot through a narrow path , 1998, Adv. Robotics.

[5] T. Horiuchi,et al. An Incremental State-Space Construction Based on the Notion of Contradiction for Reinforcement Learning , 2002 .

[6] Koichi Okuda,et al. Study of Mobility of an Articulated Mobile Robot in Narrow Passage , 2003 .