The Robotic Seabed Cleaning Platform: An

Underwater Cable-Driven Parallel Robot for

Marine Litter Removal

Marc Gouttefarde1 , Mariola Rodriguez2 , Cyril Barrelet1 , Pierre-Elie Hervé2 ,

Vincent Creuze1 , Jose Gorrotxategi2 , Arkaitz Oyarzabal2 , David Culla2 ,
Damien Sallé2 , Olivier Tempier1 , Nicola Ferrari3 , Marc Chaumont1 , and
Gérard Subsol1
1
LIRMM, Univ Montpellier, CNRS, Montpellier, France
[email protected]
2
TECNALIA, Basque Research and Technology Alliance (BRTA)
3
Servizi Tecnizi Srl, Italy

Abstract. As a contribution to the development of new techniques to

remove marine litter from the seabed of sees and oceans, the Robotic
Seabed Cleaning Platform has been designed, built and experimented
in the framework of the European Union project MAELSTROM. It es-
sentially consists of a floating platform that supports the base elements
of a 6 degree-of-freedom cable-driven parallel robot actuated by eight
winches. The mobile platform of this robot can work underwater and is
equipped with sensors to control its underwater motions and to detect &
identify marine litter. To achieve efficient and selective litter removal, an
aspiration system and a gripper are installed on the CDPR underwater
mobile platform.

Keywords: Marine litter, underwater robots, cable-driven parallel robots.

1 Introduction

It is estimated that about 70% of marine debris sinks to the seabed where it will
fragment in micro and nano plastics. On overall, tens of million tons of waste
and debris are estimated to lie on the seafloors of seas and oceans worldwide.
Significant efforts to reduce sea and ocean pollution are thus of fundamental
importance. Besides, developing new techniques for the removal of marine litter
is needed. In this context, the European Union project MAELSTROM [1] aims
at designing and integrating technologies to identify, remove, sort and transform
all types of collected marine litter into valuable raw materials. In this project,
two systems are considered for marine litter removal. The first one is developed
by The Great Bubble Barrier and consists in catching plastic pollution in rivers
before it reaches the ocean, using bubbles. The second one is presented in the
present paper: The Robotic Seabed Cleaning Platform. It mainly consists of
a floating platform that supports the base elements of a 6-DOF Cable-Driven
2 M. Gouttefarde et al.

Parallel Robot (CDPR) actuated by eight winches. The mobile platform of this
CDPR can work underwater and is equipped with sensors to control the under-
water mobile platform motion and detect & identify marine litter to be removed.
To achieve efficient and selective removal, an aspiration system to suck up smaller
litter and a gripper to grasp and remove larger items, e.g. tires, are also installed
on the CDPR underwater mobile platform.

Existing solutions for removal of marine litter lying on the seabed can be
highly harmful to the marine ecosystem, e.g., dredges and bottom trawl nets.
Human divers can also remove litter from the seabed but this work can put them
in danger and their actions are limited in time and litter weight. Another option
for marine litter removal is to use robotic systems. For instance, there exist
autonomous surface vehicles to remove floating pollution such as the low cost,
zero greenhouse emissions WaterShark [4] and the multi-purpose Jellyfishbot
[3]. Besides, to remove litter from the seabed, underwater robots [5] can be
used. Both autonomous underwater vehicles and Remotely Operated Vehicles
(ROV) allow to reach seabed at depth of several tens or hundreds of meters.
For example, the SeaClear project [2] aims at deploying a system consisting of
an unmanned surface vehicle, an unmanned aerial vehicle, a small ROV and a
larger ROV. The unmanned surface vehicle is the mothership that deploys the
aerial vehicle and the two underwater robots. The aerial vehicle can search litter
from the air (in clear water) and helps the navigation of the surface vehicle
while the small ROV scan the seabed at close range. The larger ROV removes
litter with a gripper and suction device. The SeaClear system should thus be
able to operate autonomously over large distances but can remove rather small
lightweight objects from the seafloor. Moreover, ROV may take time to stabilize
before being able to grasp a litter, thus impairing efficiency. Hence, one advantage
of the Robotic Seabed Cleaning Platform introduced in this paper is its capability
to efficiently and selectively remove relatively large and heavy objects, as well
as smaller ones, since it is based on a CDPR which constitutes a (relatively)
accurate and fast positioning system.

Numerous previous papers on underwater cables exist in the literature, from

their physical modeling, e.g. [6, 7], to their use as ROV umbilical, e.g., [13, 15],
where the cables are usually used in low tension contrary to the case of CDPRs
where tensed cables transmit motion and force. In fact, few previous works deal
with marine and underwater CDPRs. In [9], the idea of a 6-DOF underwater
mobile platform connected to surface ships by more than 6 cables is presented
but there is no specific underwater modeling, design, control or experiments in-
troduced in this paper. In [11, 12], the workspace, dynamic performances and
various cable configurations of a Gough-Stewart type floating cable-driven ma-
rine platform subjected to waves are studied. The winches are placed on the
surface mobile platform and the cables anchored on the seabed (mooring ca-
bles). Computer simulations are reported but no prototype and experiments are
presented. Moreover, the concept of underwater robot actuated in a hybrid fash-
ion by thrusters and cables is introduced in [8] and further studied in numerical
 The Robotic Seabed Cleaning Platform 3

Fig. 1. The main components of the RSCP.

simulations in [14], but no experiment or prototype has been presented at the

time of writing the present paper.
Hence, the contribution of this paper is to introduce the Robotic Seabed
Cleaning Platform (RSCP) design, control and experimental testing to remove
marine litter lying on the seabed of Venice lagoon. To the best of the authors’
knowledge, it is the first time an underwater CDPR is designed and used. The
paper is organized as follows. In Section 2, an overview of the RSCP is pre-
sented. Then, in Sections 3 and 4, the RSCP design and control are described,
respectively. Finally, Section 5 presents the results of the first RSCP cleaning
campaign in Venice lagoon while Section 6 concludes the paper.

2 Presentation of the Robotic Seabed Cleaning Platform

The RSCP consists mainly of a CDPR installed and operated from a floating
barge. The mobile platform of the CDPR can work underwater and is equipped
with selective cleaning tools that allow the removal of legacy small (micro-plastics
> 5 mm) and large items on the seabed, as well as floating plastics in the water
column, in a beneficial way for the marine ecosystem. The main components of
the RSCP are presented in Fig. 1 where the mobile platform is shown outside
of the water. In the CAD view of Fig. 2, the mobile platform is shown working
underwater. The 6-DOF mobile platform of the RSCP is driven by eight cables
in a suspended configuration similar to the one of the CoGiRo CDPR [10]. Four
base frame posts are secured to the floating platform (floating barge), the latter
being made of several modules (pontoons) assembled together. Each post has two
winches and pulleys that route the cable from the winch to the top of the post
where a swiveling output pulley directs the cable toward the mobile platform.
One of the pulley located near the winch at the bottom of the post is equipped
with a load pin and thereby allows the measurement of the cable tension. The
mobile platform can move down in the water below the floating barge which
has a rectangular hole (inner pool) in its middle. An aspiration system to suck
4 M. Gouttefarde et al.

Fig. 2. A CAD view of the RSCP with the CDPR mobile platform working underwater.

up smaller litter in the water column and seabed below the hole in the floating
barge, as well as a gripper to remove larger items, are installed on the CDPR
underwater mobile platform. A control room hosts the electronic and control
cabinets, the control PCs and the human operator(s) of the RSCP.
A picture of the RSCP during first tests on the ground in Tecnalia’s facilities
in Spain is shown in Fig. 3. The mobile platform is the beige structure in the
middle of the picture. After its setting up in Tecnalia’s facilities, the RSCP
was disassembled and moved to Venice in Italy for its first underwater cleaning
campaign (cf Section 5), as shown in Fig. 4.

3 Design of the Robotic Seabed Cleaning Platform

The RSCP components have been designed to be used in a harsh outdoor mar-
itime environment. The maximum depth of operation of the underwater mobile
platform is directly correlated to the size of the floating barge. Besides, the rigid-
ity of the barge depends on the size of its inner pool. A compromise has been
found in the design of the floating barge to reach the required depth of opera-
tion (between 15 to 20 m) while maximizing its rigidity. The CAD of the floating
barge final design is shown in Fig. 5.
The floating barge, designed for both still waters and sea operations, is com-
posed of several pontoons. Two metallic posts are fixed on two steel bases located
on the same pontoon to avoid relative displacements between the swiveling out-
put pulleys of the CDPR. The two other posts are also mounted on another
single pontoon. During the experiments reported in Section 5, the relative dis-
placement between two opposite posts mounted on different pontoons has been
estimated by means of measurements with a total station. When middle size
 The Robotic Seabed Cleaning Platform 5

Fig. 3. The RSCP during tests in Tecnalia’s facilities.

waves are shaking the floating barge, the relative movement is around +/- 5
mm, which was deemed to be an acceptable value. The floating barge is main-
tained in position, depending on the depth of the cleaning, by anchor feet or
wire anchors to the seafloor. The floating barge is displaced from a cleaning spot
to another one by a tugboat.
As shown in Fig. 6, each cable goes out of the winch and is guided by pulleys
to a movable carriage. This carriage has a vertical translation to move the upper
swiveling pulleys from a the lower position to an upper position, the latter being
2.5 m higher than the former. On the one hand, when the carriage is at the top
of the post (parking position), this reconfiguration capability allows the CDPR
mobile platform to be out of the water for assembly and maintenance operations
and for moving the floating barge from spot to spot. On the other hand, when
the carriage is at the bottom of the post (working position), it allows having
the robot inside the water for marine litter removal operations with increased
operational workspace. The marine litter removal operations are also possible
in parking position, but the CDPR workspace size is reduced. The design of
the supporting structure of the winches and pulleys is suitable for any kind of
pontoons and floating barge and is independent of the height of the pontoon.
The location of the posts on the floating barge has been defined by optimizing
the workspace of the CDPR with methods well known in the state of the art.
The calibration of the positions of the swiveling pulleys (points Ai ) is made
with the help of a total station that measures reflective targets closed to the
pulleys as shown in Fig. 7. For practical reasons, it is not possible to measure
directly the positions of the cable-platform connection points Bi as they are
virtual points at the center of rotation of universal joints. Hence, as shown in
Fig. 8, a 3D printed plastic part installed around point Bi is used to measure
6 M. Gouttefarde et al.

Fig. 4. The RSCP in Venice lagoon.

Fig. 5. CAD of the final design of the floating barge.

2 reflective targets. The position of point Bi is at the middle of the measured

targets. The plastic part is removed from point Bi once it has been measured and
installed to the next point Bi . Besides, three reflective targets are fixed on the
mobile platform to measure its pose. The positions of these targets are measured
just before or after the position measurements of the points Bi . This platform
pose measurement is required to calibrate the initial pose of the platform.
The underwater mobile platform, shown in Fig. 9, is a classic steel welded
structure, protected by a marine resistant paint. The cable attachment points
(points Bi ) are made with stainless steel universal joints. All bearings at points
Ai and Bi are Igus plastic bearings to avoid corrosion. The cables are non-
rotating steel wires of diameter 12 mm. They are generously greased with a bio
sea water compatible grease, to prevent corrosion.

4 Sensors and control system

Various sensors are located on the floating floating barge including a pressure sen-
sor to compensate for the atmospheric pressure and two RTK GPS placed at two
 The Robotic Seabed Cleaning Platform 7

Fig. 6. Final CAD design of the metallic posts (left subfigure) and the swiveling output
pulleys (right subfigure).

different locations on the floating barge to estimate the position and orientation
(around the vertical axis) of the floating barge based on position measurements
from a global navigation satellite system. It thus allows to position the RSCP
on top of a previously generated bathymetry map of the seabed environment,
allowing to identify ”hotspots”, to position the barge inner pool above one of
these hotspots and to make the CDPR mobile platform dive to reach it. This
capability turned out to be a key enabling feature in the highly turbid water of
Venice lagoon. Moreover, the CDPR has eight force sensors placed at fixed (non
swiveling) routing pulleys to measure the cable tensions as well as encoders in
the winch motors.
The sensors located inside the underwater platform shown in Fig. 9 comprise
five IP cameras which enable the human operator located in the control room on
the floating barge to see the surrounding of the underwater platform (provided
that water turbidity is not too high). Several other sensors, listed below, are also
integrated into the so-called ”smart camera” system, shown in Fig. 10, which
is fixed to the mobile platform on one of its edge. The smart camera mainly
consists of the following components. A camera used for visual servoing: The
camera enables a marine litter to be seen by the operator who can click on
the litter in the camera image so that the CDPR mobile platform approaches
the litter automatically. One depth (pressure) sensor to measure the depth of
the mobile platform (distance with respect to the sea surface). One IMU used
mainly to measure the orientation of the underwater mobile platform. A Doppler
Velocity Log (DVL) which is a hydro-acoustic sensor integrating four acoustic
beams to measure the distance of the mobile platform to the seabed and its
velocity with respect to the seabed. The DVL estimates velocity relative to
the sea bottom by sending acoustic waves from four angled transducers and
then measure the frequency shift (Dopplers effect) from the received echo. By
combining the measurements of all four transducers and the time between each
8 M. Gouttefarde et al.

Fig. 7. Calibration: Determination of the positions of the points Ai from the target
measurements.

Fig. 8. Calibration: 3D plastic parts used to measure the positions of points Bi .

acoustic pulse, it is possible to estimate the speed and direction of movement.

Since the DVL also indicates the altitude with respect to the seabed (from the
range measurements achieved by the four beams), it can be used to build a
local map of the seabed slope (sea ground surface) such as illustrated in Fig. 12.
To this end, the pressure sensors on the floating barge and on the underwater
mobile platform are also used to determine the depth of the mobile platform
with respect to the floating barge and the altitude of the floating barge with
respect to the seabed. Subsequently, the two GPS on the floating barge can be
used to locate the local map in a global earth coordinate system. This allows
either estimating previously located underwater litter positions and/or to build
a map of the (removed) litters.
The control system of the CDPR of the RSCP used in the experiments re-
ported in Section 5 consists of the CDPR teleoperation control and the percep-
tion system for underwater vision. Regarding the CDPR teleoperation control,
the pilot (human operator) uses estimated underwater mobile platform pose as
well as the cameras located on the mobile platform to drive the underwater mo-
bile platform using a joystick. The joystick commands issued by the pilot are
 The Robotic Seabed Cleaning Platform 9

Fig. 9. Close-up view of the underwater mobile platform including 3 reference points
on the platform (reflective targets).

Fig. 10. The waterproof underwater capsule of the smart camera system (left) and the
smart camera system description with reference frames fixed to it (right).

interpreted as desired operational velocities, i.e., linear and angular velocities

of the underwater mobile platform. These velocities are integrated to provide
desired positions and orientations of the mobile platform that are converted into
desired winch motor positions by means of the inverse kinematics of the CDPR.
These desired joint motor positions are sent on the fly to the drive setpoint con-
trollers which are in charge of the feedback position control of the motors. The
Human-Machine Interface (HMI) of the CDPR is shown in Fig 11. It allows the
human operator (pilot) to monitor various sensor values and to choose between
different control modes. The human operator can see the videos from the IP
cameras placed on-board the underwater mobile platform in a dedicated screen.
Regarding the perception system for underwater vision, the camera in the
smart camera system is used as the underwater perception system. It notably
enables visual servoing: Once a marine litter is seen by the human operator (at
relatively closed range from the camera because of water turbidity), the latter
can click on the litter in the camera image and the CDPR mobile platform
approaches the litter automatically. Figure 12 shows the HMI of the underwater
perception system. On the left, the depth map built with the DVL measurement
is shown. In the middle, the human operator can see the image from the camera
10 M. Gouttefarde et al.

Fig. 11. HMI of the CDPR: 1. Winch axis information area; 2. Secondary commands
for individual or combined motion of the cables, motion command to calibration posi-
tion, and enabling post vertical carriage linear movements using the joystick; 3. Mobile
platform position and orientation information area; 4. Manual commands area includ-
ing initialization (calculation of the Cartesian pose computed using the winch motor
positions by means of forward kinematics), manual move with joystick enabling &
speed override setting, and robot status; 5. Automatic command area including point-
to-point motion command (actual to destination pose) and CNC trajectory program
execution; 6. Cable tension monitoring area: Force sensor values in Kg.

of the smart camera system (here shown in turbid water with very little visibility
range). The operator can click on a point in this image, where a possible litter
to be removed is located, so that the mobile platform moves toward this point.
On the right of the HMI, the bathymetry map is shown. It is centered on the
RSCP inner pool (the orange rectangle). The red rectangle is the safe working
zone where no collision between the RSCP and the CDPR cables can happen. In
the HMI bottom part, various sensor values and other information are displayed.

5 Experiments in Venice

The RSCP was tested in September 2022 in Venice lagoon in Italy where it per-
formed a cleaning campaign. Despite highly turbid water, the cleaning campaign
was a success since many marine litters of several different types were removed
from the seabed, as shown in Fig. 13. To deal with the very low visibility due to
the turbidity of the lagoon water, real-time image enhancement was used to im-
prove the perception of the marine litter with the cameras on-board the CDPR
underwater mobile platform.4

4
As shown in the following videos of the RSCP experiments in Venice:
https://youtu.be/1EVQm-0yyRY and https://youtu.be/16k3-Bp4FCI
 The Robotic Seabed Cleaning Platform 11

Fig. 12. HMI of the underwater perception system.

Fig. 13. The RSCP removing a tire from the water (left) and various litters collected
by the RSCP during the cleaning campaign in Venice lagoon.

6 Conclusion

This paper presented the Robotic Seabed Cleaning Platform. It mainly consists
of a floating platform that supports the base elements of a 6-DOF CDPR actu-
ated by eight winches. The mobile platform of this CDPR can work underwater
and is equipped with sensors to control the underwater mobile platform motion
and detect & identify marine litter to be removed. To achieve efficient and se-
lective removal, an aspiration system to suck up smaller litter and a gripper to
grasp and remove larger items, e.g. tires, are also installed on the CDPR un-
derwater mobile platform. To the best of the authors’ knowledge, it is the first
time an underwater CDPR is designed and used. The Robotic Seabed Cleaning
Platform was successfully experimented in Venice lagoon where many marine
litters of several different types were removed from the seabed.
12 M. Gouttefarde et al.

7 Acknowledgments

This work was supported by the European Union’s H2020 Program (H2020-FNR-
2020) under the grant agreement No. 101000832 (MAELSTROM project).

References
1. European union’s H2020 project MAELSTROM. https://www.maelstrom-
h2020.eu (2023)
2. European union’s H2020 project SeaClear. https://seaclear-project.eu/ (2023)
3. Jellyfishbot: Surface robot for the cleaning and the depollution of water surfaces.
https://www.jellyfishbot.io/ (2023)
4. Wasteshark: Autonomous surface vessel designed by RanMarine.
https://www.ranmarine.io/products/wasteshark-3/ (2023)
5. Antonelli, G.: Underwater Robots: Motion and Force Control of Vehicle-
Manipulator Systems. In: Springer Tracts in Advanced Robotics. Second edition,
Springer (2006)
6. Boyer, F., Nayer, G., Leroyer, A., Visonneau, M.: Geometrically exact kirchhoff
beam theory: Application to cable dynamics. ASME Journal of Computational
and Nonlinear Dynamics 6 (Oct 2011)
7. Buckham, B., Nahon, M.: Formulation and validation of a lumped mass model for
low-tension rov tethers. International Journal of Offshore and Polar Engineering
11(4), 282–289 (2001)
8. El Ghazaly, G., Gouttefarde, M., Creuze, V.: Hybrid cable-thruster actuated un-
derwater vehicle manipulator systems: A study on force capabilities. In: Proc.
IEEE/RSJ Int. Conf. Intelligent Robots and Systems (IROS). Hamburg, Germany
(2015)
9. Ghaffar, A., Hassan, M.: Study on cable based parallel manipulator systems for
subsea applications. In: Proc. 3rd International Conference on Mechanical Engi-
neering and Mechatronics. Prague, Czech Republic (2014)
10. Gouttefarde, M., Collard, J.F., Riehl, N., Baradat, C.: Geometry selection of a
redundantly actuated cable-suspended parallel robot. IEEE Trans. on Robotics
31(2), 501–510 (2015)
11. Horoub, M., Hassan, M., Hawwa, M.: Workspace analysis of a gough-stewart type
cable marine platform subjected to harmonic water waves. Mechanism and Machine
Theory 120, 314325 (2018)
12. Horoub, M., Hawwa, M.: Influence of cables layout on the dynamic workspace of a
six-dof parallel marine manipulator. Mechanism and Machine Theory 129, 191201
(2018)
13. Laranjeira, M., Dune, C., Hugel, V.: Catenary-based visual servoing for tether
shape control between underwater vehicles. Ocean Engineering 200 (2020)
14. Sacchi, N., Simetti, E., Antonelli, G., Indiveri, G., Creuze, V., Gouttefarde, M.:
Analysis of hybrid cable-thruster actuated rov in heavy lifting interventions. In:
Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems
(IROS). Kyoto, Japan (2022)
15. Viel, C.: Self-management of the umbilical of a ROV for underwater exploration.
Ocean Engineering 248 (2022)