ERK'2021, Portorož, 163-166 163
Open-hardware open-source low-cost underwater ROV
Vid Rijavec, Danijel Skoˇ caj
UniversityofLjubljana,FacultyofComputerandInformationScience
E-mail: vid.rijavec@gmail.com,danijel.skocaj@fri.uni-lj.si
Abstract
As monitoring coastal regions becomes ever more topical
due to unforeseen changes in ecosystems there still has
not been a major breakthrough in marine robotics when
it comes to mass accessibility. In this work we present
our low-cost prototype of an underwater remotely oper-
ated vehicle (ROV) for coastal exploration and monitor-
ing. We present the mechanical, electrical and software
components of the developed ROV and describe the re-
sults of field tests. We made all the data, schematics,
software, and documentation needed for construction of
a such low-cost vessel publicly available.
Keywords
ROV , submarine, low-cost, Raspberry Pi
1 Introduction
Robotics has been making major strides in worldwide
adoption, especially in unmanned ground and aerial vehi-
cles (UGVs and UA Vs) where prices have been dropping
and with household robots entering mass production. The
same cannot be said regarding autonomous underwater
vehicles (AUVs) and remotely operated underwater vehi-
cles (ROVs). The potential for such vessels includes con-
trolling and monitoring large underwater areas without a
large crew of divers and by users without diving certifi-
cation. A number of species are already on the decline
due to climate change [1], and those are being replaced
by others, affecting ecosystems in unpredictable ways.
As such, having access to more effective and low-cost
AUVs and ROVs could help in detecting changes sooner,
allowing for faster response and potential food chain col-
lapse prevention. Similar applications involving monitor-
ing coastal regions by utilising such an underwater robot
are numerous.
There are many commercial options for buying a re-
search ROV . High-end versions include the iBubble
EVO [2] or the Blue Robotics BlueROV2 [3], with projects
like the Blue Dot Model C [4] presenting a more afford-
able but less capable option. Given that even the low-
cost options still aren’t truly affordable most enthusiasts
still prefer to construct their own do-it-yourself (DIY) re-
search vessel to customize it to their needs. However,
most of these attempts tend to still be bulky and analog.
Our vessel, shown in Fig. 1, is designed as a combi-
nation of the classical AUV torpedo-like design and the
typical boxy ROV designed for high movement accuracy,
by combining the hydrodynamic hull shape with multi-
ple control thrusters to allow for more degrees of free-
dom than a singular rear propeller. Our intention was to
construct a fast-moving ROV that can explore a coastline
from shore, one that’s controlled by a pilot using a tablet
computer but offers advanced stabilization functions and
an autopilot for ease of use. As such our design is a com-
promise between the two underwater vehicle types, giv-
ing the vessel greater speed at the cost of some maneu-
verability.
Figure 1: ROV CAD Model. The visible components of the
vessel are: a PVC pressure hull, external LED lights, modified
bilge pump thrusters (allowing movement in 4-DoF), rear stabi-
lizers, a top handle, and a transparent camera dome.
The paper outlines the system and the specific com-
ponents we’ve developed for it, with the rest of the code,
schematics and documentation being available on Github
[5].
2 Mechanical design
Figures 2, 3 and 4 present an overview of the ROV and its
component parts, along with the tablet computer used by
the pilot. The entire vessel is approximately 74 cm long
and weighs just over 10 kg with batteries.
Building a sturdy and resistant pressure hull requires
a rigid and non-permeable material that can be effectively
glued to other parts. Out of all the tested low-cost compo-
nents, the most cost-effective parts were determined to be
PVC-U sewer pipe sections. We’ve used a cleaning sec-
tion of pipe to make up the majority of the hull, as it has

164
an external hatch which is practical for accessing internal
electronics.
The front of the vessel is capped with an acrylic dome
with a thickness of 3 mm, to allow the internal camera
to see the underwater environment. It’s attached to the
sewer pipe hull using epoxy glue and silicone caulking.
The curvature of the dome later proved to be detrimental
to the video quality as it added unwanted distortions and
reflections. A more optimal design would involve the use
of a thicker but flat (but smaller) front window. The video
is then processed and sent over a length of rugged CAT
5E cable.
Figure 2: Top view of the vessel. 1. PLA flashlight mount. 2.
Depth control thrusters. 3. Horizontal movement thrusters. 4.
Zip ties. 5. PLA thruster nacelles. 6. CAT 5E cable.
Figure 3: Internal cutaway of the vessel. 7. Transparent dome
containing the Pi camera and IMU. 8. Cytron MDD10A motor
drivers. 9. Dual purpose top fin and handle. 10. Steel structural
rail. 11. Motor wires. 12. Diving flashlights with diffusers. 13.
V oltage divider and ADC. 14. Access hatch. 15. Single board
computer. 16. 2x 5 Ah 3S LiPo batteries. 17. 5 V 10 Ah Power
bank. 18. Water pressure sensor.
Submersible DC bilge pump motors intended for clear-
ing out any water that seeps into a ship’s bilge are a com-
mon choice for DIY ROV construction. The main ap-
peals are the low cost, ease of modification and full water
ingress protection. A simple improvement is to remove
the pump’s impeller and housing, replacing them with
a more efficient propeller [6] to achieve higher (and re-
versible) thrust. Our prototype uses nylon boat propellers
sized at 56 mm. These motors do have a drawback -
the o-ring seal around the shaft which exerts increasing
amounts of drag with increasing pressure [7]. The seals
fully block the shaft from rotating at around 25 m, which
limits the sub’s operational depth. As such, using fully
flooded brushless motors may be a better fit if greater
depths are needed.
Rear wings aid with stabilization when moving at
speed and were also chosen to be constructed out of PVC
due to its light weight. The tail itself is not pressurized
(typically called an outer or light hull which allows for
water ingress) and houses the water pressure sensor, the
tail fins, and protects the power/data cables along with
providing better hydrodynamic performance.
3 Hardware and electronics
The electronics system as seen on the cutaway in Fig. 3 is
centered around a single board computer, specifically the
Raspberry Pi 2 Model B+, due to its low cost and lower
TDP than later versions. The key features were GPIO pin
access and the RJ45 socket which allows for an Ethernet
link between the vessel and the control app on the sur-
face. The surface component is a low-cost Lenovo tablet
computer encased in a 3D printed housing that connects
to the Ethernet link using an external USB dongle as seen
in Fig. 4.
Figure 4: Tablet-based remote control system. 19. Rein-
forced CAT 5E cable (70 m total). 20. Cable clamp. 21. 3D
printed splash-proof housing. 22. M3 screws. 23. Micro USB
connector. 24. Ethernet network card. 25. Lenovo Tab 3.
3.1 Propulsion
As the ROV uses DC brushed motors (with a current rat-
ing of 5 A), we chose low cost Cyron MDD10A motor
drivers to connect them to the Pi. Since the current rat-
ing of the drivers is twice the required maximum we can
count on lower internal heating.
3.2 Sensors
The ROV is equipped with various sensors: an IMU, a
water pressure (depth) sensor, a battery voltage sensor,
an internal air pressure and a temperature sensor. We’ve
also used a Pi Camera Module v2, and later the D160 fish
eye modification for video recording.
We’ve used the MPU9250 IMU, which offers an ac-
celerometer, gyroscope and magnetometer. The expan-
sion board used is the GY-91 which also contains a BMP-
280 pressure and temperature sensor, so that is used to log
the state of internal air as well.
For the water pressure sensor we’ve selected a low-
cost transducer intended for measuring the pressure in-
side hoses with a capability of 0.8 MPa (which equates

165
to roughly 80 m depth of seawater). As the sensor is
analog and the Pi has no analog reading pins we’ve also
added a 10 bit 4-channel ADC to link the two together.
As the ADC has spare channels we can also measure bat-
tery level using a voltage divider circuit.
4 Software
We’ve based our software around a socket.io link, which
allows communication between the C# Unity app running
on the remote control tablet, and the node.js server run-
ning on the submarine itself, which controls most of the
functionality. To avoid having a rooted system on the
tablet and a static IP, the Raspberry Pi instead acts as
a DHCP server and assigns a known IP to any Ethernet
connected device.
Not all of the system can be managed with node.js
directly, as the libraries that allow full Pi Camera con-
figuration have only been provided for Python and C++.
Since the system requires the usage of GPU splitter ports
to simultaneously stream and record video while taking
pictures, the camera control part ended up being imple-
mented in Python, which is then launched and managed
from the node.js server.
4.1 Stabilization
We implemented two automatic functions: depth hold
and heading hold. Establishing proportional control be-
tween the detected depth and target depth turned out to
be enough to archive a satisfactory depth hold.
A more advanced system was required for holding a
specific heading. An AHRS IMU fusion system [8] was
added to get a good internal belief state regarding the
submarine’s orientation. We’ve then implemented a 360-
degree-capable proportional rotational system, which al-
lows the pilot to input a single forwards or backwards
speed while the vessel holds orientation.
Full PID controllers did not turn out to be required
for adequate stabilization due to high drag and inherent
damping and were also infeasible to properly tune during
operation.
The system manages the aforementioned sensors and
streams their data to the tablet, uses them internally for
stabilization and also records them into a log file for later
use.
4.2 Remote control app
We’ve implemented a front-end interface for controlling
and configuring the submarine from a tablet computer,
which was developed using Unity. The video arriving
from the camera in a mjpeg format is displayed using an
adapted version of the Mjpegprocessor library [9], while
the rest of the telemetry and commands get routed through
the socket.io library [10]. Fig. 5 shows the GUI while
streaming the image of a calibration board, which indi-
cates the camera field of view.
The main functionality is remote control, which is
done either with manual joysticks or by selecting a di-
rection using the bottom compass bar and a depth using
a vertical depth indicator. In order to provide as much
Figure 5: Layout of the tablet interface, showing the relevant
telemetry and video feed.
on-site debugging power, the app can set all stabilization
and video recording parameters and allow for rudimen-
tary terminal access.
4.3 Visualization and analysis
We built a Godot engine visualizer to attempt to pro-
cess the recorded log file data and reconstruct the ROV’s
odometry to some degree as seen in Fig. 6.
The logged data gives us two absolute measurements:
rotation (based on the fused IMU data) and depth (cal-
culated from water pressure readings). It’s not possible
to completely determine the movement path using just
these two values, however, we also logged the individual
thrusters’ speeds which were used to infer movement ve-
locity. While it’s not possible to compensate for drift due
to waves and currents the resulting movement path is still
good enough in most cases.
While the data was logged every 100 ms, the visual-
ization ran at a higher frequency (16 ms for 60 frames
per second) so the data had to be linearly interpolated for
smooth movement. A synced on-board video of a dive
with the telemetry visualization which can be viewed on
Youtube [11].
Figure 6: The ROV being rendered in Godot, visualizing a
recorded dive and drawing a trail behind the vessel.
5 Evaluation and testing
The finalized ROV shown in Fig. 7 performed as re-
quired and is capable of exploring a coastal area while
recording clear video as shown in Fig. 8. An edited

166
compilation of video recordings that sums up research,
construction and testing can be viewed on Youtube [12].
Figure 7: The ROV with an action camera mounted on the top
rail.
To determine the effects of different buoyancy on the
system’s operation, tests were done in both salt and fresh
water. The difference ended up being minor, however,
the depth calculation can be off by 2.5 % due to the extra
mass of salt water. We therefore added parameters that
can be changed in the mobile app to compensate for the
changing environment.
We’ve used two different action cameras to record ex-
ternal video during testing: the Apeman A80 and later a
GoPro Hero 9 Black combined with a red filter as the pre-
vious setup proved unsatisfactory. Having a red filter al-
lows us to do an initial colour correction by restoring the
lower wavelengths which get absorbed far more in water
and lead to over-saturation of green and blue.
The recorded data shows that the air pressure in the
ROV stays relatively constant regardless of depth, which
would mean that there is minimal flex in the pressure
hull as intended. The internal air temperature also stayed
close to the external water temperature. It would appear
the lack of isolation in the hull provides for excellent
cooling and would allow for higher TDP computers to
work without issues.
Figure 8: Fan mussles which were found in large numbers when
testing in Izola, some of them decayed due to parasite infesta-
tion, confirming news sources [13].
Another addition to the system would be a sonar sys-
tem for sea floor and obstacle detection, however, there
are no affordable sensors that are capable of being inte-
grated into a custom system at this time. One would need
to construct a custom driver circuit for a commercial fish
finder sonar transducer (e.g. TL88E, XF02), as conven-
tional sonar sensors designed for AGV/AUV atmospheric
usage lack the required power and do not operate in ef-
fective frequencies, which lowers the range and accuracy
under water to unusable levels.
6 Conclusion
In this paper, we described the process of building a low-
cost underwater DIY ROV with open-source hardware
and software, useful for light research and monitoring
of shallow coastal regions. We’ve achieved this goal by
choosing a crush depth of 25 m and used less resistant
and subsequently cheaper materials, as well as using 3D
printed components to fill the gap between other parts.
The total cost of the system ended up being around 400 C,
excluding the tablet computer and action cameras.
The vessel is stable enough to record useful video
due to the included IMU and sensor fusion, while be-
ing able to utilize its high top speed to rapidly traverse
a chosen area and locate points of interest. The external
mechanical design is robust enough to handle expected
conditions, however the electrical internal setup could be
further improved by designing a dedicated printed circuit
board.
7 Acknowledgements
The authors would also like to thank professors Dr. Niko-
laj Zimic and Dr. Matjaˇ z Vidmar for their help and con-
tributions in the research phase of the project.
References
[1] Marine Food Webs Are on the Brink of Collapse Because
of Climate Change, https://futurism.com/marine-food-
webs-brink-collapse-because-climate-change
[2] iBubble - Your Personal Underwater Cameraman,
http://ibubble.camera/product/ibubble-autonomous-
underwater-drone
[3] Blue Robotics, http://bluerobotics.com
[4] Blue Dot Underwater Drones, https://www.bluedotrov.com
[5] Source code, schematics and documentation,
https://github.com/MoffKalast/RaspberryUnderwaterROV
[6] Homebuilt ROVs - Seafox Retrofit Thrusters,
https://www.homebuiltrovs.com/seafoxretrofitthrusters.html
[7] Homebuilt ROVs Mayfair 750 GPH Bilge Pump Thruster
Testing,
http://www.homebuiltrovs.com/mayfair750test.html
[8] ahrs - npm, https://www.npmjs.com/package/ahrs
[9] SampleUnityMjpegViewer,
https://github.com/DanielArnett/SampleUnityMjpegViewer
[10] socket.io-unity,
https://github.com/floatinghotpot/socket.io-unity
[11] ROViz visualization,
https://www.youtube.com/watch?v=M0sli2o5RJY
[12] Construction and testing video,
https://www.youtube.com/watch?v=DufHhX7p4Xk
[13] Slovenian sea researchers uncover a number of deceased
fan mussles, https://www.sta.si/2797119