Analysis of the Motion Pattern and the Learning Mechanism for Manipulating Objects by Human Fingers.

Abstract

It is important to clarify the motion pattern and learning mechanisms for manipulating objects by human fingers in order to develop a hardware of multi-fingered robot hand and a redundant-multi-DOF learning control system. We measured the trajectory of five fingertips of human hand. The task was to rotate a circular cylinder as a simple example. Then, we analyzed the trajectory. The analyzing methods were based on statistics analysis, first Fourier transformation, spectrum analysis and chaos analysis. As a result, coordinated movement pattern of five fingers were measured. It is found that thumb and forefinger stick whereas other fingers slip during contact period. It is also found that the movement changes to a specific pattern when the task is learned. We imply that the motion control method changes from feedback control to feedforward control after learning.