このエントリーをはてなブックマークに追加


ID 33073
フルテキストURL
著者
Ito, Kazuyuki Okayama University
Matsuno, Fumitoshi Tokyo Institute of Technology
抄録

We consider a flexible autonomous system. To realize the system, we employ a hyper-redundant system (a flexible hardware system) and reinforcement learning controller "QDSEGA" (Q-learning with structuring exploration space based on genetic algorithm) which is a flexible software system. In this paper we apply QDSEGA for controlling of the hyper-redundant robot. To demonstrate the effectiveness, a task of acquisition of locomotion patterns is applied to a multi-legged formation and a snake-like formation, from which an effective locomotion is obtained.

キーワード
Hyper-redundant systems
Q-learning
Reinforcement learning
Genetic Algorithm
QDSEGA
備考
Published with permission from the copyright holder. this is the institute's copy, as published in SICE 2002. Proceedings of the 41st SICE Annual Conference, 5-7 Aug. 2002, Volume 3, Pages 1499-1504.
Publisher L:http://ieeexplore.ieee.org/search/wrapper.jsp?arnumber=1196529
Copyright © 2002 IEEE. All rights reserved.
発行日
2002-8
出版物タイトル
SICE 2002. Proceedings of the 41st SICE Annual Conference
3巻
開始ページ
1499
終了ページ
1504
資料タイプ
学術雑誌論文
言語
英語
査読
有り
DOI
Submission Path
mechanical_engineering/8