Robot Acquisiton of Everyday Grasping Skills

Abstract

This study aims to map human grasp skills to a robot hand using “Learning from Observation”. The task domain, “motions of a Tea Ceremony”, is described by using “Kamakura’s Taxonomy” as grasp primitives, and realized by adding a multi-fingered hand with tactile sensors to our humanoid robot platform. As a result, everyday grasp skills are mapped to the robot.