Page 107 - Human Inspired Dexterity in Robotic Manipulation

Approaching Human Hand Dexterity Through Highly Biomimetic Design 105

Table 6.1 The specifications of the dynamixel servos used in our biomimetic
robotic hand
Dynamixel servo model AX-12A MX-12W
Working voltage (V) 12 12
No load speed (RPM) 59 470
Stall torque (Nm) 1.5 0.2
Gear ratio 254/1 32/1
Resolution (degrees) 0.29 0.088
Range of motion (degrees) 300 360
Communication speed 7343 bps to 1 Mbps 8000 bps to 4.5 Mbps
Weight (g) 55 54.6
Dimensions (mm) 32 40 50 32 40 50

Notes: Two MX-12Ws and one AX-12A are used for controlling the thumb; five MX-12Ws are used for
the index and middle fingers; and two MX-12Ws are used for the coupled ring and little fingers.

straighten independently. But they also share an extra servo for a coupled
control at their MCP joints. We use three actuators to control the thumb.
One of them is an AX-12A Dynamixel servo that has a larger gear ratio
(254/1) than others (32/1) and is used for the extension/abduction of the
thumb. The other two servos of the thumb are assigned to control the flex-
ion and adduction motions, respectively. The diameter of the pulley that is
directly attached to the shaft of the servo is 20 mm.
The specifications of the two types of servos are listed in Table 6.1.

6.4.2 Teleoperation of Our Proposed Robotic Hand
Currently, no tactile or joint angle sensor was implemented into our pro-
posed robotic hand. But the designs we developed for the artificial skin
and fingertip sensor could all be modified to work with the existing robotic
hand prototype. For this proof-of-concept prototype, the next critical step is
to validate its functionality with teleoperation. To this end, a new type of
data glove was designed to achieve the one-to-one easy mapping from
the human hand to the robotic hand (see Fig. 6.15).
As our proposed anthropomorphic robotic hand closely mimics the
human hand, it is straightforward to consider extracting movement informa-
tion from the latter and sending it as control inputs for the former. This idea
is not new, in fact, most of the dexterous motions demonstrated by the exist-
ing anthropomorphic robotic hands were all realized by using this method
known as teleoperation. During this process, to map the motion from the
human hand to the robotic hand, joint angles are often collected by a data
glove, and then translated into commands that can be used to control the