IEEE Robotics & Automation Magazine - December 2015 - 130

manipulation can occur with any number of fingers, which
means that the most independent measure of strength would
be finger strength, and 2) there can be inherent variability in
finger strength across different fingers, even in cases where
they are mechanically equivalent. Finger strength is a function of the hands actuator capabilities, motion controllers,
mechanical design, and finger-object configuration.
Of the previously listed dependence, only the finger-object
configuration is used as a test variable. While using an extended
(near singular) finger configuration, position the finger just
above the force sensor and verify a zero force reading (see Figure 7). Under the position control, command the finger to close
completely, which should induce control saturation. This configuration was chosen as it is easily replicated across a diverse set of
robot hands and results in a maximum moment arm at the
point of contact that generates smaller contact forces when compared with more curled finger states. Record force sensor data
throughout the test. The use of a three-axis reference force sensor is encouraged to accurately capture the total contact force.
The particular performance measure regarding fingertip contact force magnitude, Ffinger, should be computed by the L 2 norm
for each set of force readings given by the external force sensor.
Next, the contact force magnitude from the quasistatic (settled)
force region (see Figure 8) should be extracted for each load cycle
(several cycles or repetitions should be conducted), and then avFfinger

Force Sensor
Figure 7. The robotic finger exerting maximum fingertip force as
measured by a reference force sensor.

Cycle 1

Cycle 2

Fcontact

Dynamic
Quasistatic

Figure 8. The depiction of dynamic and quasistatic force regions
during load cycles.

Experiments
The experimental results were obtained in this category by
commanding each finger (one finger per test) of a robotic
hand to close completely while on a collision course with an
object attached to an external six-axis load cell. Again, the
robotic hand and load cell were positioned so that contact
took place at the fingertip when the finger was fully extended
and perpendicular to the palm, as shown in Figure 9. Finger-
object contact was established and removed 32 times per finger for each hand while recording interaction forces via the
load cell to generate performance distributions. The results
from these experiments are shown in Table 3. Hand 1 exhibits Ffinger,max values of 8.24-14.29 N, while Hand 2 exhibits
Ffinger,max values ranging from 30.44 to 33.89 N.
Grasp Strength
Metric and Test Method
Grasp strength is a kinetic measure of the maximum internal
force a robotic hand can impose on an object. This measure will
yield information regarding the particular payload capabilities
for various object sizes as well as its limits in resisting pulling or
pushing forces during a grasp operation. Grasp strength is a
function of the hands actuator capabilities, motion controllers,
mechanical design, grasp configuration, and object size.
Of the previously listed dependence, only the grasp configuration and object size are assumed controllable. For this particular
test, a wrap grasp is issued on cylinders with different diameters
with embedded force sensing (see Figure 10). The wrap grasp is a
typical power grasp that should yield maximum grasp strength
performance. The cylinder diameters are varied to capture grasp
strength performance across differently sized artifacts that induce
changes in the settled grasp configuration. The artifact-embedded force sensors are used as the reference sensors for resolving
cross-sectional internal force transmission. In this particular
design, since the embedded force sensing can only resolve internal forces in one direction, measuring force transmissions in the
Table 3. The mean and 95% confidence intervals
of the maximum fingertip contact force magnitude
exerted by each finger for Hand 1 and Hand 2
measured using a load cell.
Hand 1 Ffinger, max (N )

Quasistatic
Grasp Force

Time

130

eraged to yield the maximum finger strength, Ffinger,max . Collect
maximum forces for several cycles and compute the mean and
95% confidence intervals to estimate finger strength.

IEEE ROBOTICS & AUTOMATION MAGAZINE

DECEMBER 2015

Hand 2 Ffinger, max (N )

95% Confidence
Finger Mean Interval
Mean

95% Confidence
Interval

14.29

[13.66, 14.73]

30.44

[29.39, 30.95]

8.24

[8.00, 8.80]

30.84

[30.63, 31.04]

10.64

[10.41, 10.83]

33.89

[33.21, 34.73]

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2015