Page 153 - Human Inspired Dexterity in Robotic Manipulation

150 Human Inspired Dexterity in Robotic Manipulation

8.1 INTRODUCTION

Hand manipulation is one of the most important technologies in robotics.
Industrial robots installed in a mass-production line use simple grippers to
repeatedly execute a pick-and-place operation. The grippers usually do
not adequately have the degree of freedom (DOF) to manipulate the grasped
object. The shape of the gripper jaws is designed to fit the shape of a selected
grasped object, and the particular jaw shape limits a variety of graspable
objects. Therefore, grasp and manipulation ability of the grippers are limited.
Recently, work in cell production and logistics require robots to grasp and
manipulate various objects, and multifingered robotic hands are expected to
perform complex work.
Grasping is a basic function of multifingered hands and can be roughly
classified into power and precision grasps [1]. In a power grasp, the fingers
make contact with an object with multiple points, and the DOF of the fin-
gers is totally lost. Thus, the hand can firmly grasp large and heavy objects
with the palm and fingers, but the fingers cannot manipulate the object.
Meanwhile, precision grasp is a method to grasp an object only with the fin-
gertips. The hand still has a remaining DOF to manipulate a grasped object if
an appropriate controller is designed.
Power grasp is an inherent reflexive skill that a new-born baby can realize
with tactile sensing, and, alternately, a baby takes about 9 months to acquire
precision grasp [2]. Furthermore, a physiological fact noted during the dis-
charge of neurons in the brain of a macaque monkey during a precision grasp
was stronger than that during a power grasp [3] that indicated that a precision
grasp is a more dexterous skill than a power grasp. In robotics, it is well
known that a power grasp can be implemented with an underactuated
mechanism controlled by a simple strategy. On the other hand, manipula-
tion during a precision grasp is a difficult skill because a hand needs to main-
tain contact between the fingertips and the object during manipulation.
This paper introduces a control method of the precision grasp and
manipulation inspired by the thumb opposability and its implementation.
This method is based on a passivity of the system that is useful in decreasing
prescribed information and object sensing.

8.2 STRUCTURE OF HAND CONTROL SYSTEMS

Fig. 8.1 shows the basic structure of control systems of multifingered robotic
hands. The control systems for robotic hands have three different roles; task