Research on a method to transfer high-level skills extracted from sensing information to robots

2019 Fiscal Year Final Research Report

• Project/Area Number: 17K06471
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Measurement engineering
• Research Institution: Chukyo University
• Principal Investigator: Hashimoto Manabu 中京大学, 工学部, 教授 (70510832)
• Co-Investigator(Kenkyū-buntansha): 秋月 秀一 慶應義塾大学, 理工学部(矢上), 助教 (40796182)
• Project Period (FY): 2017-04-01 – 2020-03-31
• Keywords: 作業動作分析 / 熟練作業者 / 視線計測 / 動作計測 / 深層学習 / LSTM / 3次元物体認識 / スキル分析

Outline of Final Research Achievements

In this research, we have studied a method that can extract "skill" of an expert person automatically and a method that can transfer the extracted "skill" to the robot system. There are two major research results as follows. (1)We proposed a technique of sensing, description and analysis that treat both 3D objects and human hand motions. Concretely, a method based on information about view-direction and working-time, a method based on co-occurrence between view-direction and hand-motion, and a method based on spacial temporal information about view-direction and hand-motion. (2)We also proposed a curvature-based local features, a model-less object recognition method by utilizing approximating objects to prepared primitive shapes such as hexahedron, cylinder and sphere, and an object recognition method by using rough shape models.

Free Research Field

画像情報処理

Academic Significance and Societal Importance of the Research Achievements

本研究の成果は，熟練作業者の「技」という，従来は暗黙知とされてきた情報を，視線や手の動きという物理計測可能な情報の分析によってデータ化し，熟練者の技が具体的にどのように数値的に表現できるかを示した．この点に最も大きな学術的意義がある．また，年々少子化が進み，労働力人口の減少を余儀なくされている日本の社会において，人に代わって高度な製品を製造するロボットの実現に向けた有用な情報を提供できた点に，社会的な大きな意義があると考えている．