IEEE Robotics & Automation Magazine - March 2017 - 38

Alignment

Grasp Hold Test

Grasp Acquisition Test

Figure 9. A summary of the open-loop grasp evaluation test. The grasp acquisition tests whether or not the hand can successfully pick
an object off the table with open-loop control. The grasp hold test determines whether the object remains within the grasp in different
orientations.

continue to move until either all phalanxes have made contact
or a joint limit is reached. This results in a high degree of passive adaptability, as shown in Figure 6. In contrast, the Model
T42's two independently actuated fingers allow for stable precision grasping but still retain a high degree of adaptability
(Figure 7). The Model O can switch between multiple base
configurations, shown in Figure 8, facilitating the grasp of a
wider range of object geometries.
Dexterous manipulation primitives have also been demonstrated with these underactuated designs. The passive reconfiguration in underactuated fingers can enable robust
surface-constrained precision grasping. This same reconfiguration, in combination with the joint elasticity in the fingers,
also enables in-hand manipulation, assuming the hand operates in the subset of actuation space that maintains no-slip
contact conditions.
Open-Loop Grasping Evaluation
Optimal grasp planning is beyond the scope of this article, but
a rudimentary, open-loop grasping test was performed to
help elucidate the grasping capabilities of these underactuated
hand designs. Similar to tests run in [18] and [24], the hands

were directed to perform surface-constrained grasping. In
each trial, the hand was initialized with its palm directed
downward at the table and oriented such that it was aligned
with the object's principal axis. The hand was then lowered
toward the object until either the palm touched the object or
the fingers would otherwise be obstructed by the table during
grasping closure. Regardless of the object geometry, the same
open-loop command was given to the hand to close and then
attempt to lift the object. After lifting, if the object remained
in the grasp, the hand was reoriented by r/2 radians such that
its palm axis vector now pointed outwards. The hand was
then rotated about its palm axis by r radians and then back
to help determine the grasp quality and its ability to maintain
a hold on the object under the influence of gravity in different
orientations. Figure 9 summarizes the procedure in these
grasping tests. The tests were performed for all three
OpenHand designs and both types of fingers for the Model T
and T42. For the Model O, the fingers were oriented such that
the static finger directly opposed the other two in a powergrasping configuration.
Table 3 details the results for a selection of household objects
with different geometries. Grasp acquisition by itself was quite

Table 3. The results of open-loop grasp testing.

Object

Coffee Cup

Mustard Bottle (Full)

Spatula

Cheez-It Box (Full)

Weight (g)

118

432

104

453

Size (mm)

89 × 89 × 83

38 × 76 × 178

38 × 102 × 356

64 × 161 × 229

Grasp Acquisition Test (Hold Test)

38

*

T (Pivot)

5/5 (5/5)

5/5 (4/5)

5/5 (5/5)

5/5 (3/5)

T (Flexure)

5/5 (1/5)

5/5 (5/5)

5/5 (5/5)

4/5 (0/5)

T42 (Pivot)

5/5 (5/5)

5/5 (4/5)

4/5 (3/5)

5/5 (0/5)

T42 (Flexure)

5/5 (4/5)

5/5 (2/5)

3/5 (2/5)

5/5 (0/5)

O

5/5 (5/5)

5/5 (2/5)

5/5 (5/5)

5/5 (2/5)

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

March 2017
Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2017
