Page 113 - Human Inspired Dexterity in Robotic Manipulation

Approaching Human Hand Dexterity Through Highly Biomimetic Design 111

operator through our custom-made data glove. As shown in Fig. 6.19, the
two principal components of the thumb motions, namely the flexion/exten-
sion and the abduction/adduction, were tested separately. The teleoperation
results in more scattered data points compared to the ones collected from the
preprogrammed motions of ring and little fingers (see Fig. 6.18).

6.5.2 Object Grasping and Manipulation

To further evaluate the overall performance of our robotic hand, we con-
1
ducted grasping and manipulation experiments using 31 objects from the
prioritized list [22]. During the test a human operator hands different objects
to the robotic hand with his right hand, meanwhile, using his left hand to
teleoperate the digits of the robotic hand to grasp/manipulate the object
via the data glove (see Fig. 6.16). This process is known as telemanipulation
during which the movement of the robotic hand is both controlled and
guided by the same human operator’s hand motion and visual feedback. This
experiment can be seen as a series of cooperative grasping tasks between the
human operator and the robotic hand in which the former could clearly
monitor the status of the grasped object without the occlusion issue at
the grasping site (see our video submission for details).
During the object grasping task, we observed that different grasping pos-
tures can be naturally transferred from the human operator to our biomi-
metic robotic hand, particularly from the motion of the thumb. This is
because our biomimetic robotic hand successfully preserves the important
biomechanics of the human hand that essentially determines the hand kine-
matics. The resulting grasps cover most of the grasping types defined by
human hand taxonomy [23], except for the ones that require independent
control of the ring and little fingers (see Fig. 6.20).
Last but not least, we tested the in-hand manipulation ability of our bio-
mimetic robotic hand. As shown in Fig. 6.21, the whiteboard eraser was suc-
cessfully regrasped from a horizontal to vertical position through a series of
continuous hand motions involving the use of all the digits. It is interesting
to observe that complicated in-hand manipulation tasks can be accomplished
without any force feedback. This again suggests that matching the kinemat-
ics of the robotic hand to its human counterpart is important for the success
of teleoperation tasks.

1
Details about the tested objects can be found in [21].