Study of coordinated motions of the human hand for robotic applications

Abstract

This paper presents an acquisition method that comprehensively looks for the mimic configurations of the human hand. The data obtained through this process is further analyzed, transformed, and then used to synthesize a reduced configuration space of a robot anthropomorphic hand. The method rely on a dimensionality reduction technique that provides a new basis of the full configuration space, from which one can select a subset of the vectors forming that basis, and finally obtaining a simpler configuration subspace. These vectors are called Principal Motion Directions, and represent the coordinated motions captured by a sensorized glove on a human hand and transferred to the robot hand. The characteristics and limitations of the subspace are discussed, as well as its application in several scenarios within robotics such as the motion planning of robot hands, where the subspace has been successfully implemented and executed.