Future University Hakodate Academic Archive >
研究者 >
複雑系知能学科 >
三上 貞芳 >

このアイテムの引用には次の識別子を使用してください: http://hdl.handle.net/10445/8452

タイトル: Acquisition of Self-Recovery Actions for Leg and Wheeled Autonomous Rover by Reinforcement Learning
著者: Horikawa, Masatoshi
Wakahara, Takumi
Ikeda, Kazunori
MIKAMI, Sadayoshi
アブストラクト: This paper proposes an adaptation method for an autonomous leg-and-wheeled robot to enable stuck-fee movement under a variety of terrains. Targeting hardware is one of the anticipated configurations for exploration tasks, which has 6 legs and a motor-driven wheel on top of each leg. The method is based on an on-line combinatorial search for both gait and wheel movement according to a performance measurement. The problem is its huge search space, and the proposing system cuts down state and action space to minimal amounts. Experiments were conducted by using physics simulation, and the results show that the proposed system could gain smooth gait/wheel control under simulated rugged ground for a variety of learning coefficients.
研究業績種別: 国際会議/International Conference
資料種別: Conference Paper
査読有無: あり/yes
単著共著: 共著/joint
発表雑誌名,発表学会名など: 12th International Symposium on Advanced Intelligent Systems
開始ページ: 315
終了ページ: 318
年月日: 2011年9月30日
出現コレクション:三上 貞芳


ファイル 記述 サイズフォーマット
2011-09-30-ISIS2011-Horikawa-315p.pdf851.93 kBAdobe PDF見る/開く



Powered by DSpace Software Copyright © 2002-2007 MIT and Hewlett-Packard