American Journal of Mechanical Engineering, 2019, Vol. 7, No.

3, 116-128
Available online at http://pubs.sciepub.com/ajme/7/3/2
Published by Science and Education Publishing
DOI:10.12691/ajme-7-3-2

Using a High-Speed Mini-PC to Control

an Autonomous Underwater Vehicle
Ali. Jebelli1,*, Mustapha C.E. Yagoub1, Balbir S. Dhillon2
1
School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, Canada
2
Department of Mechanical Engineering, University of Ottawa, Ottawa, Canada
*Corresponding author: ajebelli@uottawa.ca

Received May 23, 2019; Revised July 02, 2019; Accepted July 14, 2019
Abstract In this paper, the authors present the design of a mini-PC board to fast processing of an AUV
(Autonomous Underwater Vehicle) with two 360° rotational thrusters. This rotation capability in the thrusters, along
with an embedded mass shifter, allows for long-time immersive motor maneuvers for this AUV. Also, the conditions
for moving the device at the water surface and depth are provided by changing the angle of the thrusters and moving
the mass shifter along with the motor speed change. All the commands related to the motor angles, proper position
of the thrusters, mass shifter, and speed, are sent to an Engine board via a high-speed Mini PC. The high-speed Mini
PC sends decisions to the Engine board after receiving and processing all sensors data as well as data from
embedded cameras. This integrated system provides the ability to move safely based on accurate analysis of the
environment by sensors and images received by cameras from the surroundings, with minimal energy consumption
for the underwater AUV. The results of various tests demonstrated the high performance of this AUV.
Keywords: autonomous underwater vehicle, PTFE, Mini-PC
Cite This Article: Ali. Jebelli, Mustapha C.E. Yagoub, and Balbir S. Dhillon, “Using a High-Speed Mini-PC
to Control an Autonomous Underwater Vehicle.” American Journal of Mechanical Engineering, vol. 7, no. 3
(2019): 116-128. doi: 10.12691/ajme-7-3-2.

cameras and sensor information. Thus, the board defines

the position of the mass shifter, speed, and angle of the
1. Introduction thrusters. To reduce weight and optimize size in the AUV,
we designed two-level electronic boards that allow
Nowadays, the research to enhance the capabilities of reducing communication cables. Finally, a prototype of
Autonomous Underwater Vehicles (AUVs) is mainly to the designed AUV was successfully tested under real
reach uncertain and harsh environments that could be very conditions (in both pool and ocean environments).
difficult to access to humans. Such AUVs can indeed allow
exploring the sea and monitoring underwater structures,
where prolonged exposure to humans is dangerous or even, 2. AUV Description
sometimes, impossible. Reducing the size of these AUVs
will increase their availability and maneuverability. However, To successfully design the body, one should establish a
it is very difficult and challenging to embed electrical and primary plan of the body structure with the exact location
mechanical parts inside these AUVs, thus leading to high of each sub-part to be embedded, knowing that it will be
costs as well as hard challenges for AUV designers [1-5]. made of one piece with cameras on the top and bottom,
In this paper, the authors introduced a small Teflon and engines attached on the sides (Figure 1 to Figure 3).
(PTFE) underwater AUV, which has a very high As seen in Figure 2, the curving all around the body
maneuverability capability with only two 360-degree minimizes the drag force, as detailed later. The valve #1 in
rotational thrusters. Reducing the number of thrusters can Figure 3 is used for placing charger cables, microcontrollers
save energy and thus give the AUV a higher working time and processor data cables. It, therefore, enables the user to
autonomy. To this end, a user will provide the desired have access to the whole system without having to open
depth and direction to the AUV, and this later will move the whole body, thus preserving the body sealing. It is to
toward the predefined arrival point with minimal energy be noted that the rear of the body has a cubic shape to
consumption because of a mass shifter implemented in the precisely allow more space to this access valve. The holes
AUV and a control loop between the thrusters and this #2 and #3 in Figure 3 are for the pressure sensor and the
mass shifter. In this AUV, we used a high-speed mini-PC antenna, respectively.
unit that simultaneously captures/analyses data from The body was designed with two dolly blocks (#4 in
embedded cameras and sensors, and sends the appropriate Figure 3), in which servomotors have been placed to
commands to the Engine board by matching images from change the thruster angle.
 American Journal of Mechanical Engineering 117

Valve #5 is for the parts that are connected to the middle

sheet: the mass shifter and its motor. Finally, valve #6 is
for the main camera space as well as for the sensor board.
Figure 7 shows the placement of the driver board, the bottom
camera, the sensor board, the mass shifter, and the main
camera.

Figure 1. Isometric scheme of the designed AUV Figure 4. Six designed valves

Figure 2. Different views of the designed AUV

Figure 5. Location of the sensors

Figure 3. Access valve (1), Antenna holes (2), Pressure sensor (3), and
Dolly block of servo motor (4)
Figure 6. The Bottom openings: (1) Bottom camera, (2) and (3) Sensors
The thickness of the body (made of Polytetrafluoroethylene
- PTFE) is of 1 cm. There are two cameras, one above
(main camera) and one on the bottom. Considering the
thickness of the body, some extra tools may be attached to
the body if deemed necessary. Therefore, adding more
thrusters or an arm would be possible to further enhance
the AUV functionality.
Beside the main access valve at the rear, five other valves
have been also designed to embed the electrical/mechanical
parts. Such valves have been isolated from the rest of the Figure 7. Locating internal components: (1) Sensors, (2) Bottom camera,
body by walls to avoid any damage due to possible (3) Mini-PC, (4) Batteries (5) Mass Shifter, (6) Main Camera
leakage. In Figure 4, valve #1 is a valve under the access
valve to access the batteries. The engine driver board is
placed in the space provided by valve #2. Valve #3 has 3. Main Processor
been dedicated to locate the Mini PC and valve #4 to
initially locate the sensors. However, due to their high All AUV processes including camera/sensor data, detecting
sensitivity, they have been relocated as in Figure 5: the obstacles, controlling motor operation, etc., should be
circular space near the cape has been used for placing the managed via a main processing board. Considering the
Inertial Measurement Unit (IMU) and the compass sensors, amount of data to manage in real time, the AUV operation
as in Figure 6. In this Figure, #1 is devoted to the bottom needs a high-speed processor to efficiently receive and
camera while #2 and #3 are for the sensors. Note that send commands. With its high processing speed, supporting
these parts have holes at the bottom covered by clear and USB ports, and 802.11ac dual band wireless network, we
thinner-than-body lens, the aim being to weaken the selected the mini PC board Giada i200-BG000, Celeron
earth's electromagnetic waves inside the body. [6]. Figure 8 shows the block diagram of the main processor.
118 American Journal of Mechanical Engineering

Because there is a linear relation between water

pressure and water depth, we used a pressure sensor to
measure the device depth (fixed to a maximum of 8 m).
The selected pressure sensor is the MS5803-14BA [9], a
high-resolution pressure sensor with I2C interface.
A high-resolution temperature output allowed the
implementation of a depth measurement system and
thermometer without any additional sensor. A gel
protection and antimagnetic stainless-steel cap has been
included to protect against overpressure waterproof [1].
To measure humidity and temperature, we used the
SHT-11 digital sensor with I2C protocol. A unique
Figure 8. Block diagram of main processor
capacitive sensor element is used to measure relative
humidity while the temperature is measured by a band-gap
4. Sensors and Sensor Boards sensor [1].
The aim of the sensor board is to collect data from the
In this work, we used different sensors based on the above sensors. Data will be sent to the main processing
specific functionalities assigned to the designed AUV. The board via a serial port. This board is responsible for
common factors in selecting these sensors are their measuring data from all the sensors and sending them to
relatively low cost, high performance, small size, high the main processor board. This board also receives the
sensitivity, resistance against heat and low current required commands from the main processing board
consumption, as well as their ability to work with the I2C and send them to the motor driver board. It is based
communication protocol. As mentioned above, due to on the microcontroller ATMEGA64 [10], which has two
their high sensitivity, the sensors and their electronic Universal Asynchronous Receiver/Transmitter (UART)
board have been embedded in a separate container in order communication ports: one for the main processor and one
to reduce the communication error with the main for the motors driver board. The mass shifter circuit,
processor as well as to minimize the risk of damage and composed of two switches at both ends of a rod, as well as
calibration errors in case of accidently hitting an obstacle. the LED lighting control circuit are mounted in this board.
Figure 9 shown the hardware and electronics connections. A programmer socket, a reset circuit, a battery voltage
detection socket, and a LDR photocell sensor used to
detect changes in the resistance of the circuit board have
been also mounted on this board along with sensor sockets
of the above mentioned HMC6343, MPU9250, MS5803
and SHT11 (two). A vacant socket (for future use) is also
part of this board. Note that due to the large space required
by all fittings and connectors, this board has been
designed in two levels. The circuits are located in the
bottom level while the sockets and connectors are located
on the top level of the board.

5. Transceiver
As mentioned, the AUV’s function is to work into
Figure 9. Hardware & Electronics connections water at a maximum depth of 8 m. The Mini PC (Giada
The selected IMU sensor is the MPU-9250 with a i200-BG000, Celeron) usually operates with Wi-Fi and
3-axis gyroscope, a 3-axis accelerometer, a 3-axis compass, Bluetooth systems. However, due to underwater antenna
and an internal temperature sensor [7]. The selected issues in terms of cost and power supply, we decided to
compass sensor is the HMC6343, a fully integrated use two transceiver modules in order to evaluate the
compass module [8]. According to the available space and system performance during the AUV mission and to
the sensors involved, two electronic boards were designed control the AUV in emergency mode. For this purpose, we
to connect the sensor board with an IDE cable (Figure 10). retained two RF7020 (ADF7020) modules: one external to
be connected to the external controller and one internal to
be connected to the Mini-PC through a USB port [11,12,13].

6. Cameras
In recent years, demands for underwater tasks, such as
exploration of ocean environment and resources or
inspection of underwater structures, have dramatically
Figure 10. Sensor board (top board of dimensions 67.4*48.4 mm2 and increased. We have then included two cameras in the AUV
bottom board of dimensions 69.8*49.8 mm2) design [1]: the first (WebCam C615) in a transparent
 American Journal of Mechanical Engineering 119

dome-like space at the top of the AUV with a capability of As main power supply, we used two rechargeable
a 360o rotation and the second on the AUV floor (720p batteries: a 20V/14Ah battery and a 12V/12Ah battery
USB Camera with Infrared LEDs) [1], both for observing (with regulators). Full battery specifications are reported
the surroundings and detecting obstacles (Figure 11). Also, in Table 2.
several LEDs have been added all around the cameras to
shed light into the water environment. A light board was Table 2. Battery Specifications
then designed, consisting of an LDR (Light Dependent
Resistors) photocell sensor [1] to measure the ambient
light and four LED for providing light on the top of the
camera.

8. Access Valve Board

An access valve board has been placed in the final part
of the AUV (in the access valve chamber). The aim of
designing such board is to turn the device on, access to
information stored in the device, and charge the battery
Figure 11. (a): The Main Camera Board and (b): The Down Camera after being drained. This board is made up of two
perpendicular boards so that the user can plug the network
cable into the charger cable.
7. Power Distribution Board and A micro-switch has been embedded in the horizontal
Batteries board for the user to directly turn-on or -off the main
processing board. There are also two input sockets and
Because of the variety of the devices to be embedded two output sockets for voltage as well. Input sockets are
in the AUV (Table 1), a power distribution board of connected to the battery and output sockets will be linked
dimensions 69.7*48.6 mm2 was designed with a µA 7805 to the power source. Two headers have been used for
regulator (Figure 12). connecting the input sockets to the on-board battery
Also, since the 12 V battery voltage should not fall charger on the vertical board. Figure 13 shows the Access
below 9 V and the 20 V battery voltage should not fall board, which includes two sockets for connecting chargers
below 15 V, two-voltage detection circuits have and a network socket for connecting to an external
been embedded. Their outputs are evaluated by computer.
the analog-to-digital converter of the sensor board
microcontroller. As a result, the outputs from these
circuits have been connected to the sensor board.
Table 1. Required Voltages

Figure 13. Access Valve Board with dimensions 52.4*87.4 mm2

9. Engine and Driver Board

A single driver is used for controlling all the engines.
This driver is responsible for receiving a series of orders
in a serial form from the main processor board via the
sensors board, and controlling the direction and speed of
all engines, which includes two BTN7971B motor driver
modules for control of DC motors [14].
This board produces the Pulse Width Modulation
(PWM) signals required by all engines via a mega64
microcontroller. Block diagram and schematic of the
engine’s driver board is shown in Figure 14-a and the
Figure 12. Power Distribution Board BTN7971B Motor Driver Module in Figure 14-b.
120 American Journal of Mechanical Engineering

horizontal movement) and depth (corresponding to the

device vertical movement), as well as the pitch angle and
the direction level.

10.1. Speed Controller

This controller adjusts the speed of the left and right
engines to provide the required speed depending on the
scheduled route. Thus, a proportional integral controller
(PI) [1] has been used, which formula can be set as:
T kpu × kiu ∫ ε u .
= (1)

with kpu the speed controller proportional coefficient and

kiu the speed controller integral coefficient. T is the
controller output in terms of engine revolution percentage
and εu the speed error, calculated as:
ε=
u ud − u (2)
where ud is the desired speed and u the speed measured by
the sensors.
Figure 14. Engine driver board with dimensions 157.5*12 mm2, (b)
BTN7971B driver module
10.2. Pitch Controller
In order to provide different types of movements to the
device we implemented: A mass moving system was designed to control the
- Two Buhler 1.13.044.023.03 24 V DC motors for pitch angle, i.e., the X body-axis towards the horizon.
moving the main body of the device [15]. The intended mass moves only towards the longitudinal
- One Moons hybrid stepper motor 16HS Series 1.8° direction of the body (the x-axis) and causes rotation
/12 V DC to move the ball screw and Mass Shifter, around the center of gravity, thus creating positive or
its dynamic torque curves is shown in Figure 15. negative angles towards horizon. Due to its inherent
- Two XQ-S4020D 21.8 kg waterproof digital stability, the device tends to keep the pitch angle close to
servomotors for the horizontal and vertical rotation zero. This implies that if the instrument is out of balance,
of thrusters. it returns to its steady state, the zero angle. Moreover, in
Note that all the above engines are directly connected to dive or rising, it is desirable that the angle of the nose can
the motor’s driver board. change upwards or downwards to decrease the body angle
with its path, thus decreasing the friction with the fluid,
and therefore, increasing speed.
For this purpose, the pitch controller adjusts the moving
poise close to the center of balance to make the nose angle
zero, when the device moves horizontally and if the device
tends to change its depth upwards or downwards it replaces
the moving poise backwards or forwards, respectively.
A proportional controller has been used to adjust the
nose angle as:

SGB=
B
kp p × ε p . (3)
2

Where SGB is the controller output, which includes the

2B
poise distance towards the center of balance. kpp is the
controller proportional coefficient and εp the nose angle
error between the desired pitch angle θd and the pitch
Figure 15. Dynamic torque curves of the Moons hybrid stepper motor angle θ measured by the sensors:
16HS Series 1.8° /12 V DC [1,16]
ε=
p θd − θ (4)

10. Controller Design Note that preliminary simulations showed that there is
no need for an integrator term.
The designed vehicle has five degrees of freedom
for moving namely, horizontal/vertical movements and 10.3. Depth controller
rotations around the x-axis, the y-axis, and the z-axis.
Different controlling algorithms were then implemented The depth controller maintains the device at the desired
to control the AUV’s speed (associated with the device depth. It has two modes, at rest and during movement:
 American Journal of Mechanical Engineering 121

 Depth Controller at rest: If the AUV is at rest, the

speed is zero and the angle θs relatively to the
horizon is set automatically by the servos to 90o.
The depth is maintained at the desired value by
adjusting the level of engines’ revolution. For this, a
proportional integral controller has been used,
T = kpD1 × ε D + kiD1 ∫ ε D . (5)

with kpD1 the depth controller proportional coefficient at

rest and kiD1 the depth controller integral coefficient at rest.
Figure 16. The final prototype of AUV in operation on water. At this
T is the controller output determined by the engines’ time, the user is sending it the commands to begin its mission
revolution in terms of percentage. εD is the depth error
calculated as:
ε=
D Dd − D. (6)
where Dd is the desired depth and D the measured depth
evaluated by the sensors.
 Depth controller during movement: when the speed
is not zero, the θs parameter is used to control the
depth by changing the engines’ angle towards
horizon and the required vertical force is provided.
For this purpose, a proportional integral controller
Figure 17. Internal view of the AUV
has been used as
Integrating so different blocks is indeed challenging. So,
ε s= kpD2 × ε D + kiD2 ∫ ε D . (7)
we first tested the interaction between blocks to
with kpD2 the depth controller proportional coefficient predefined commands before embedding them. Once
assembled and insulated, we can access the AUV through
during movement and kiD2 the depth controller integral the access board. At rest, the servomotors and the mass
coefficient during movement. shifter return to their default mode (the mass shifter return
to the center of gravity of the AUV and servomotors
10.4. Level Controller return to horizontal state or zero angle). Also, the rotating
body of the camera chamber is placed on the AUV nose in
The logic of controlling the AUV direction is based default mode.
on the difference between the left and the right servo's We first connected the Mini-PC to a computer to
angle. For example, if the desired rotation is towards perform initial verifications. Then, we activated the
right, the left servo is placed upper than the right servo wireless connection mode for the Mini-PC and removed
and becomes closer to horizon and if the desirable the cable network from the access board. Next, we placed
rotation of the device is towards left, the right servo the cap access valve and sealed it. Once placed on water,
comes up and becomes closer to horizon. To implement the AUV will float at the surface until we send the first set
this logic a proportional-integral controller has been of commands.
used as
θe = kp y × ε y + ki y ∫ ε y (8) 11.2. Performance Results

Here kpy is the proportional coefficient of the controller In this section, all the above controllers (speed, pitch,
and kiy its integral coefficient. θe is the controller output, depth and direction) were integrated into a single
showing the required difference between the servo motors, algorithm and the whole system behavior evaluated.
and εy is the direction error given by For the test, a predefined set of commands was sent to
the prototype in terms of desired speed (Figure 18), depth
ε=
y yd − y (9) (Figure 19) and direction angle (Figure 20). As shown in
Figure 21 to Figure 33, the AUV’s responses follow the
where yd is the desired direction angle and y the direction given commands, thus allowing validating the algorithm
angle calculated by the sensors. implementation.
In fact, while comparing Figure 21 and Figure 22, we
can notice that the evolution of the speed prototype to the
11. Experimental Tests speed command agrees with the vertical speed. In spots
where the depth change occurs, this later rises to balance
11.1. Experimental Setup the AUV’s moving.
This has been further confirmed by the acceleration
The above algorithms were first implemented and tested curves (Figure 23 to Figure 25). Indeed, when the AUV
separately. Then, the AUV was built with all required changes the angle of its pitch and rotates towards left and
mechanical and electrical blocks (Figure 16 and Figure 17). right, an acceleration is applied to the AUV in the y-axis
122 American Journal of Mechanical Engineering

direction of the body to closely follow its commands. Similarly, for the other acceleration components.

Figure 18. Desired speed

Figure 19. Desired depth.

Figure 20. Desired direction angle

 American Journal of Mechanical Engineering 123

Figure 21. AUV response (Black) to speed command (Blue)

Figure 22. Vertical speed

Figure 23. Acceleration towards x-axis

124 American Journal of Mechanical Engineering

Figure 24. Acceleration towards y-axis

Figure 25. Acceleration towards z-axis

Figure 26. AUV response (Black) to depth command (Blue)

 American Journal of Mechanical Engineering 125

Figure 27. AUV response (Black) to direction angle command (Blue).

Figure 28. Angular velocity around x-axis

Figure 29. Angular velocity around y-axis

126 American Journal of Mechanical Engineering

Figure 30. Angular velocity around z-axis

Figure 31. The status of the moving weight

Figure 32. Left servo angle

 American Journal of Mechanical Engineering 127

Figure 33. Right servo angle

Same conclusions can be made for the depth and pitch After receiving predefined-user commands regarding
angle (Figure 26 and Figure 27). the desired arrival point to reach at a certain depth and
We should note that to maintain the stability of the with a certain speed, the fabricated AUV prototype was
AUV, angular velocity around the x-axis occurred to keep able to perform the given task by efficiently changing its
it stable around its longitudinal axis (Figure 28). The driving motors’ speed and servomotors’ angle. This
variations of the y- and z-axis angular velocities as well as played an important role in preserving the energy
the servo angles and the corresponding moving of the resources of the AUV, thus increasing its autonomy.
mass shift weight confirm the good operation of the Finally, after more than 600 hours of testing under various
prototype (Figure 29 to Figure 33). operating conditions including both pool and ocean
environments, the designed AUV successfully responded
to all commands by taking appropriate decisions.
12. Conclusion
In this paper, we designed a lightweight autonomous References
underwater AUV with vision capabilities. The tasks
involved the design of various mechanical parts, [1] Jebelli, A., “Design of an Autonomous Underwater Vehicle
particularly two rotating thrusters and an interior mass with Vision Capabilities,” Ph.D Thesis, University of Ottawa,
2016.
shifter to efficiently control the device movement. Sensor [2] Jebelli, A., “Development of Sensors and Microcontrollers for
boards were also embedded to share information with a Underwater Robots,” Master’s Thesis, University of Ottawa, 2014.
Mini-PC board and thus, to assure proper AUV operation. [3] Jebelli, A., Yagoub, M.C.E, Dhillon, B.S., “Design and control
Control-based algorithms were developed to control the of a self-balancing autonomous underwater vehicle with vision
AUV operation and assure proper responses to external capabilities,” Journal of Marine Science: Research &
Development, Vol. 8, No. 1, pp. 1-7, 2018.
commands and/or surrounding conditions. Also, two [4] Jebelli, A., Yagoub, M.C.E, Dhillon, B.S., “Modeling of an
cameras were used to collect information about the autonomous underwater robot with rotating thrusters,” Advances
surrounding environment. in Robotics and Automation, Vol. 6, No. 1, pp. 1-10, 2017.
To do so, we started with the AUV body by selecting a [5] Zhao, Y., Wang, L. and Peng, S., “A Real-Time Collision
Avoidance Learning System for Unmanned Surface Vessels,”
shape and subdividing it into blocks in order to locate all
Neurocomputing, 182, pp. 255-266, 2016.
embedded mechanical and electrical components required [6] i200, 4th Gen. Intel® CoreTM i3/i5/i7 Processor Intel® HD
for proper operation under real environmental conditions; Graphics 4400, Intel, 2015.
the objective being to achieve a good trade-off between [7] MPU-9250 Product Specification, Document Number: PS-MPU-
size and weight. Then, by setting the simulated maximum 9250A-01, InvenSense Inc, 2014.
depth to 50 m, loading, strain, and stress analyses and [8] Three-axis Compass with Algorithms HMC6343, Honeywell’s
Magnetic Sensors, 2014.
simulations were performed to ensure the body safety [9] 126MS5803-14BA Miniature 14 bar Module, Measurement
while submitted to external forces (note that the effective Specialties Inc., 2012.
maximum depth was set to 8 m, well below the simulated [10] http://www.atmel.com/images/atmel-2490-8-bit-avr-
value of 50 m). Next, we included a mass shifter to control microcontroller-atmega64- l_datasheet.pdf, [Last access May 23,
2018].
the AUV movement and assure its stability. We also
[11] FSK/ASK Transceiver IC, Analog Devices, Inc., 2012.
designed a sensor board, a power board, an access board, [12] KeyStone Architecture Universal Asynchronous Receiver/Transmitter,
and a motor driver board as well as a Mini-PC to manage Texas Instruments, 2010.
the data exchange with the different mechanical and [13] http://www.dorji.com/docs/data/DRF7020D13.pdf, [Last access
electronic parts like the IMU, pressure, and compass sensors. May 25, 2018].
128 American Journal of Mechanical Engineering

[14] BTN7971B High Current PN Half Bridge NovalithIC, Infineon [16] http://www.motionusa.com.s3-website-us-east-
Techn., 2008. 1.amazonaws.com/moons/Hybrid
[15] http://www.buehlermotor.de/C12572D40025EAF8/vwContentBy _Stepper_Motors_16HS_Series.pdf, [Last access May 18, 2018].
Key/W274AHF53 91 WEBRDE/$File/DC-
Motor_51x73_1.13.044.0XX.pdf, [Last access May 27, 2018].

© The Author(s) 2019. This article is an open access article distributed under the terms and conditions of the Creative Commons
Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).