IEEE Robotics & Automation Magazine - December 2018 - 53

Laboratory Experiments
To determine an estimate of the end-effector grasping force, laboratory experiments with the estimation algorithm (see the section "Grasp Force Estimation") have been carried out. After
calibration, different (stiff or soft) objects are grasped and
released by the system inside a water tank. The current residual
resulting from these tasks is recorded and reported. An example
of the grasp force experiment process and its resulting current
residual are presented in Figure 7, where the expected current
variations between objects of different stiffness (also shown) can
be seen. The experiment is performed for both terminal devices.
A rubber foam block (Young modulus - 0.015 GPa ) is
used as a soft specimen, while a small aluminum bar (Young
modulus - 69 GPa ) represents the stiff specimen. The shape
and dimensions of the two objects are very similar. During the
grasp phase, the SoftHand yields a mean current residual of
- 250 mA for the soft object and - 320 mA for the stiff one.
Regarding the soft gripper, the same quantities result in
- 400 mA and - 470 mA, respectively. As expected, the
current residual is proportional to
1) the grasped object's stiffness, i.e., the total grasp force exerted by the end effector
2) the grasp Jacobian of the contact, which is larger for the
soft gripper by virtue of its longer middle phalanges.
Moreover, laboratory experiments of the developed fast tool
change system have been carried out: a complete tool change
procedure is shown in Figure 5.
The manipulation of large objects, difficult to grasp with
the SoftHand, has been simulated employing the soft gripper
controlled by a human operator. Such tasks are shown in the
first two rows of Figure 9.

Pole-Slipping Detection

2,000
Current (mA)

grasping an object, e.g., to the grasped object's stiffness. This
behavior is shown in Figure 7. For the same closure reference
and object dimension, the current residual is larger when the
grasped object is stiffer, denoting a larger interaction force,
and reverts to its empty closed-hand value ^- 0h if the
object is pulled out.
To exploit the effects of pole slipping, experiments similar to those presented in the "Laboratory Experiments" section were conducted, applying a load torque demand
beyond the capability of the MC (which could happen if an
overaggressive control action is applied to the motor). Figure 8 shows the resulting I m with a dead zone threshold.
The excessive load torque occurs when I m - 1, 500 mA is
applied at t - 14 s. As expected, a discontinuity on I m is
observed. Then, the motor position remains constant until
t - 20 s, when the hand returns to open. Now, however, the
I m results are quite different compared to its open-hand
original value (which is set to be - 0 mA after a calibration
procedure), i.e., before the pole slipping occurred. If we
compare Figure 8 with Figure 7, where I m is shown in the
case of successful grasps, it is clear that an uncoupling can
be diagnosed through the current monitoring. In this manner, a simple fault-detection algorithm without any sensor
on the load side can be implemented.

PoleSlipping Event

1,500
1,000
500
0
-500

15
Time (s)

Measured Current Residual
Saturated Current Residual
Figure 8. The fault detection experiment with the soft gripper.
The excessive motor absorbed current I m that provokes the poleslipping condition at t = 14 s results in about 1,500 mA. Both
the measured (in blue) and saturated (for analysis simplification,
in red) motor currents are reported. A saturation threshold of
! 200 mA is depicted as a black dashed line. Compare this figure
with Figure 7, where the same experiment is performed without
excessive torque demand.

Finally, we performed a pressure test to better assess the
pressure-tolerant behavior of our mechanical structure. The
two terminal devices underwent a static depth test in a pressure
chamber (provided by a company that specializes in underwater robotic research and development, Graal Tech S.r.l. in
Genoa, Italy). We inspected them following the test, and they
showed no sign of degradation or damage after being subjected
to a pressure of 50 bar (- 500-m depth) . Their performance,
too, appeared to be unaffected by the pressure test because,
immediately after removing the devices from the chamber, we
were able to grasp even
complex objects with the
SoftHand one (thanks also
The devices' modular
to the fast tool change system). Video footage of
nature and the custom tool
the test is presented in the
supplementary material
change system open up the
available for this article on
IEEE Xplore.
possibility for simple use
Field Experiments
by an ADS operator or an
Underwater experimental validation of the sysROV/AUV robotic arm.
tem has been carried on
at various depths with
the two terminal devices depicted in Figure 3. The manipulation system was controlled by a human operator, as in
[11]. Figures 9 and 10 present field experiments employing the SoftHand and soft gripper device, respectively, at a
depth of - 1 m.
In these experiments, focused on the SoftHand terminal
device, archeological recovery [Figure 10(a)-(j)] and biological
sampling [Figure 10(k)-(t)] operations were simulated, along
with a force operation [Figure 10(u)-(y)]. The end effector
successfully grasped complex objects like stones, vase shards,
coins, reproduced coral pieces, and reproduced aquatic plant
december 2018

IEEE ROBOTICS & AUTOMATION MAGAZINE

IEEE Robotics & Automation Magazine - December 2018

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2018