IEEE Robotics & Automation Magazine - December 2020 - 18

A design but with the legs extending slowly (~50 mm/s).
Finally, in configuration E, we connected SILVER2, without
the arm, directly to the handle with a not-extensible wire. SIL-
VER2's movements were monitored by an operator, and the
arm movement was controlled with a camera mounted on the
SILVER2 platform itself.
Workspace Protocol
The workspace protocol focused on measuring the enlarge-
ment of the manipulator workspace when it was mounted
on SILVER2. In this protocol, SILVER2 and the manipula-
tor grasped an object from a lower platform and released it
in a target area at a height of 37 cm [Figure 5(b)]. The test
consisted of two steps: first, SILVER2 was placed in front of

the platform, and the soft manipulator grasped the object
from the lower platform. Second, SILVER2 extended its
legs without walking, and the manipulator executed several
movements (~50 mm/s) to release the object at the target
location. Specifically, the legs were elongated to gain
approximately 20 cm in the vertical direction. Releasing the
object on the target could be considered a success. The
error was measured as the distance between the center of
the object and that of the red, circular target. We also tested
the performance of the soft manipulator without SILVER2.
We placed the soft manipulator alone in the water tank and
performed the same soft manipulator routine: moving an
object from the bottom of the target area. Then, we mea-
sured the maximum height of the target area to which the
soft manipulator could reach and
release the object.

Silver2

Target
Area
r2

e
lv
Si

Handle
Force
Sensor

Object
(b)

(a)

Object
Target Area

Silver2

(c)

Object

Object
Silver2
Silver2

(d)

(e)

Figure 5. The experimental setup for different protocols conducted on the seafloor
(depicted here as a dry environment for demonstration purposes). The (a) force protocol,
(b) workspace protocol, (c) mobile pick-and-place protocol, and (d) and (e) confined
environment protocol. The robot's initial position is represented by a labeled square.
Scale bar: 15 cm.

18

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

DECEMBER 2020

Mobile Pick-and-Place Protocol
This protocol [Figure 5(c)] was used to
measure the performance of the whole
system in a typical mobile manipula-
tion task. It enabled us to estimate the
precision and effectiveness of the com-
bined SILVER2-manipulator arm sys-
tem. The test began with the arm
grabbing the target object (a shell) and
lifting it from the platform. Then, SIL-
VER2 initiated a lateral walking gait
(~150 mm/s) to reach the target zone,
which was 2 m from the starting point.
Once the target was within the work-
space of the arm, the arm was actuated
to place and release the object onto the
target plate. The system's performance
was evaluated by measuring the distance
between the center of the object and that
of the red circle for each trial. Releasing
the object on the target plate could be
considered a success.
Collecting Litter and Objects
With this protocol, we aimed to
qualitatively assess the system's per-
formance during a typical collection
mission, either to perform seabed
cleaning or retrieve a fragile biologi-
cal specimen. The selected test
objects were a hard silicone seashell;
an eggshell, which simulated a frag-
ile biological specimen; and three
pieces of common trash found on
the seabed: a plastic shopping bag, a
plastic bottle (0.5 L), and a nylon
fishing net (400 cm2). Additionally,
the net was tested in an entangled
condition: a small rock (0.35 kg) was
IEEE Robotics & Automation Magazine - December 2020

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2020
