Research on interactive links between actions and visual recognition for handling soft objects

2014 Fiscal Year Final Research Report

• Project/Area Number: 22240019
• Research Category: Grant-in-Aid for Scientific Research (A)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: Perception information processing/Intelligent robotics
• Research Institution: National Institute of Advanced Industrial Science and Technology
• Principal Investigator: KITA Yasuyo 独立行政法人産業技術総合研究所, 知能システム研究部門, 上級主任研究員 (00356875)
• Co-Investigator(Kenkyū-buntansha): KITA Nobuyuki 独立行政法人産業技術総合研究所, 知能システム研究部門, 主任研究員 (90356874) ; UESHIBA Toshio 独立行政法人産業技術総合研究所, 情報技術研究部門, 主任研究員 (20356546) ; KANEHIRO Fumio 独立行政法人産業技術総合研究所, 知能システム研究部門, 主任研究員 (70356806) ; MORISAWA Mitsuharu 独立行政法人産業技術総合研究所, 知能システム研究部門, 研究員 (00392671)
• Project Period (FY): 2010-04-01 – 2015-03-31
• Keywords: ロボットビジョン / 柔軟物ハンドリング / モデル駆動型画像処理 / アクション連動 / ヒューマノイド / ステレオビジョン / 広視野視覚 / 変形モデリング

Outline of Final Research Achievements

It is becoming increasingly important for robots to autonomously handle daily necessities, such as clothes folding. In this project, we have studied methods for robustly handling a clothing item by densely linking handling actions and visual recognition via a deformable model of the item. As a practical task, folding an item of clothing from an arbitrary shape into a fixed form was considered. We treat“recognition of clothing state”as a turning point in the total process. While the clothing state is unknown, actions which reshape the item so as to produce more informative observations are proactively used. Once the state of the item becomes known, the desirable clothing shape is positively led and traced by iterating handling actions and deformable-model based visual recognition. Experimental results using a humanoid system showed a good prospective of the proposed strategy.

Free Research Field

コンピュータビジョン