DEGREE PROJECT IN VEHICLE ENGINEERING,

SECOND CYCLE, 30 CREDITS

STOCKHOLM, SWEDEN 2021

Design and development of a

submersible UAV
Design, build and testing of an amphibious UAV
capable of vertical takeoff

ADAM OLSSON

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ENGINEERING SCIENCES
Abstract

Using autonomous underwater vehicles is a popular method of collecting samples

and conducting surveys, but the transportation of this information is not always easy.
The underwater vehicle might be unable to transmit the information wirelessly, and
samples may be required to be transported a long distance. A possible solution to
these problems is a hybrid unmanned aerial vehicle, accompanying said underwater
vehicle. After a submerged deployment, this vehicle could transport the information
over long distances, or conduct other operations at different locations in air or water.
While quadcopters are an increasingly popular type of vehicle, conventional fixed­wing
planes are still superior when it comes to range. This thesis designs, builds and tests
such a vehicle, with the goal of a submerged deployment, performing vertical takeoff,
and then transitioning to fixed­wing flight. To minimize the drone’s impact on the
vehicle which it accompanies, it is nearly buoyancy neutral by flooding the hull with
water, which enters and exits the vehicle rapidly during dive and egress. To manage
the pressure at the underwater vehicle’s operating depth, it utilizes a bladder rather
than having a heavy rigid compartment. It floats as a tailsitter at the surface, using
two motors in a tractor configuration to pull itself out of the water. The vehicle built
proved capable of being submerged and taking off vertically, however there were no
fixed­wing flights attempted.

ii
Sammanfattning

Autonoma undervattensfordon är en populär metod för att samla in prover och

göra undersökningar, men det är inte alltid uppenbart hur denna information
ska transporteras. Undervattensfarkosten kanske inte har kapaciteten att sända
informationen trådlöst, och prover kan krävas att transporteras en lång sträcka. En
möjlig lösning på dessa problem är en hybrid obemannad flygande farkost som följer
med undervattensfordonet. Efter en undervattensstart kan detta fordon transportera
informationen över långa avstånd, alternativt genomföra andra operationer på olika
platser. Quadcoptrar är en alltmer populär typ av fordon, men konventionella flygplan
är fortfarande överlägsna när det gäller räckvidd. Denna avhandling konstruerar,
bygger och testar ett sådant fordon, med målet att efter en undervattenstart, genomföra
vertikal start från vattenytan och sedan övergå till horisontell flygning. För att
minimera drönarens inverkan på fordonet som den transporteras med har den nästan
neutral flytkraft, detta sker genom att kroppen fylls med vatten, som snabbt kommer
in och ut ur fordonet under dykning och vid flygning. För att hantera trycket på
undervattensfordonets operationsdjup använder den en blåsa i stället för att ha en
styv behållare. Den flyter med svansen ner på ytan, och använder två motorer i
en traktorkonfiguration för att dra sig ur vattnet. Fordonet som byggts visade sig
kunna sänkas ned under vattnet och lyfta vertikalt, men det gjordes inga horisontella
flygningar.

iii
Contents

1 Introduction 1

2 Background 2
2.1 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Design and prototyping 6

3.1 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.1 Water flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.2 Buoyancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.3 Nosecone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2.4 Electrical compartment and bladder . . . . . . . . . . . . . . . . 11
3.2.5 Electrical Components . . . . . . . . . . . . . . . . . . . . . . . 12

4 Testing 14
4.1 Flight test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2 Test videos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

5 Discussion 17
5.1 Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.1.1 Buoyancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.1.2 Water problems . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.1.3 Electronics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.2 Design flaws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.3 Thesis remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

6 Conclusions 23

iv
CONTENTS

6.1 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

References 25

v
Chapter 1

Introduction

Unmanned aerial vehicles (UAVs) have become increasingly popular, with their
potential usages expanding. One such area is how they can be used together with and
in support of submarines, both for civilian and military purposes. Military submarines
can utilize UAVs for communication purposes, as transmitting information while
submerged is difficult, but they can also be used for surveillance. Blackwing[1] is one
present UAV for this purpose, providing both communications and aerial surveillance,
while launched from a submarine. This also represents a special case of UAVs,
capable of being launched while the submarine is still submerged. There are similar
applications in the civilian world, for instance supplementing autonomous underwater
vehicles (AUVs) doing research. The UAV can transport samples or information
gathered by the AUV, which may be limited in its ability to transfer this information.
It can also be used for other research­oriented missions, such as scouting a different
location while airborne, or landing elsewhere for some sensor­oriented mission. One
benefit of using an UAV instead of an AUV is the ability to cover great distances in
shorter periods of times, and the ability to use different types of sensors. This thesis
explores the concept of such a vehicle.

1
Chapter 2

Background

This thesis presents the development of the Maribot Petrel in its third iteration. Petrel
is a submersible UAV, designed to be capable of vertical takeoff from water and
transition to fixed­wing flight. The mission for which Petrel is designed is sketched
in Fig. 2.1.1. The starting point of the mission is riding with an AUV at some depth
below the surface, waiting to be deployed. After a submerged deployment, the drone
should float up to the surface where it should be able to communicate with a control
station and eventually be able to tak eoff vertically. Once airborne, it should transition
to fixed­wing flight and perform some type of mission.This mission could entail flying
a long distance, performing observations, landing in water for measurements, or flying
”home”. It should have the capability of doing all of this autonomously.

Previous projects have attempted different parts of this mission with varying success.
As the full mission is very broad, the focus of this thesis was to develop, prototype
and test a UAV capable of riding with an AUV, deploying, VTOL from the surface and
fixed­wing transition. The AUV the drone was designed to accompany, LoLo [3], is a
3.8 m long and 1.05 m wide submarine which is being developed at the KTH Maritime
Robotics Laboratory.

This thesis originally started as a group project, but was split up down the line. The
focus of this report is mainly the design, testing and development. To read more about
how the different parts of the UAV were manufactured, see [8].

2
CHAPTER 2. BACKGROUND

2.1 Related Work

The second iteration of Petrel, carried out in 2019 [5] showed that a tilt­rotor
configuration could be a viable option for vertical takeoff. This was done using
an existing airframe which floated flat at the surface as a bellysitter, with a tractor
configuration mounted on a rotating mechanism. While it proved capable of taking
off, it showed the fragility of such a system and the need for a computer system to
control the rotating mechanism.

Figure 2.1.1: Mission profile.

Further development of the tilt­rotor configuration was carried

out in the Kungsfiskare[4] project at KTH. This was done using the same airframe
as the previous project, but instead using three motors in a tricopter setup, as seen
in Fig. 2.1.2c. The front two motors tilt for fixed­wing flight, while the rear motor is
only used for vertical flight. This showed excellent vertical takeoff and landing (VTOL)
capabilities and hovering maneuverability, but also the limitations due to the added
weight from the tilt mechanism. It proved capable off taking of from the water surface
and transitioning to fixed­wing flight.

In 2017 Caruccio et al[2] designed a similarly capable UAV, showing the advantages of
having a tailsitter vehicle with a single puller motor mounted in the nose. A tailsitter
floats with the tail submerged at the surface, with the nose pointing straight up. While
this project had issues with underwater communication and taking off, it showed the
potential capabilities of a very simple and light VTOL design.

[7] designed and built a cross­domain capable vehicle, also utilizing the tail­sitter

3
CHAPTER 2. BACKGROUND

(a) Eagleray. (b) CX − 2

(c) Kungsfiskare. (d) Petrel v2.

Figure 2.1.2: Tail­sitter configurations.

configuration. This vehicle showed the capabilities of both underwater and flying
locomotion, and the transition between the two mediums. It presented a way to rapidly
drain the wings of water during egress, by utilizing extra spoilerons to open the trailing
edge. During egress it also demonstrated the need for ailerons to be mounted in the
propeller stream to have control during low flight speeds, this to avoid spinning out of
control during takeoff from the torque generated by the propeller.

2.2 Requirements
The requirements of the drone are specified below, which also includes the limitations
imposed by the carrying AUV.

The drone

1. shall perform multiple preplanned aerial missions in fixed wing flight, including
VTOL, from and to the water surface.

2. shall withstand and operate in typical archipelago conditions (<5 m/s wind speed

4
CHAPTER 2. BACKGROUND

and related sea conditions).

3. shall be submersible to a maximum depth of 10 m below sea level.

4. shall be net positively buoyant at all times.

5. shall be attached to LoLo while LoLo is moving at a maximum speed of 1 m/s at

a maximum depth of 10 m below sea level.

6. should not produce more than 40N of drag onto LoLo at any time.

7. should not produce more than 20N of vertical force onto LoLo at any time.

8. shall be detachable from LoLo by a release mechanism operated by and mounted

on LoLo.

9. should be detachable up to a speed of 1 m/s, at a depth of 10 m below sea level.

10. shall always be prepared to be released and initiate a mission while attached to
LoLo.

11. should perform a minimum of one aerial mission after a submerged start from
LoLo.

5
Chapter 3

Design and prototyping

One of the first design decisions was whether it should be a tailsitter or bellysitter.
While [4] showed that tilting the motors was a viable option with excellent control, it
also showed the difficulties that comes with it. [6] explored the option of having a quad
motor setup for VTOL and a separate motor for fixed­wing flight, but concluded that
the most efficient and simplest solution was a tail­sitter. Another issue with the belly­
sitter configuration is that the state which it floats at the surface becomes crucial. It
is difficult to control the orientation as it floats up from being submerged, and it may
end up with the propellers facing down. The tail­sitter has the benefit of orientation
not mattering, as long as it is floating vertically. Using a tail­sitter configuration,
there is still the option of having motors which can tilt for more control. But as [7]
demonstrated vertical takeoff capabilities with fixed motors, tilting motors seemed
excessive for the mission requirements.

While it is beneficial to reuse proven technology, such as an already flown airframe, this
was not possible here. Most of the available airframes in this size (≈2 m wingspan) are
made out of foam, and as there is a strict limit to how buoyant the drone is allowed to
be, these are not usable. Therefore a new airframe needs to be designed and built.

From the list of requirements a few key criterion can be taken out as the starting points
of the design.

• The outer dimensions of the drone are limited by the size of the AUV which the
drone would be riding with, as it should not hinder or limit said vehicle in any
way. The span was limited to 1.4 m, which in turn limits the weight and stall
speed.

6
CHAPTER 3. DESIGN AND PROTOTYPING

• As the drone is required to be near buoyancy neutral, the inside of the plane needs
to be flooded with water. During egress this water has to be drained from the
vehicle rapidly. To reduce the amount of water which needs to flow in and out,
the total enclosed volume should be minimized.

• A waterproof compartment for the electrical components is required, which

imposes a limit to the size of the fuselage as the box needs to fit inside.

• It should be a fixed­wing.

• It should be able to take off vertically as a tailsitter, which means it needs to be

floating nose up at the surface.

3.1 Concepts
Several different concepts were considered, ranging from more advanced designs to
simple wings. A few of these are shown in Fig. 3.1.1. As [6] demonstrated problems
due to the torque generated from the propeller during takeoff, two motors were used
with counter­spinning propellers to counteract each other. A rough size of the fuselage
was estimated based on the size of the box used for the electrical components. As there
is no commercially available servos which can handle a water depth of 10 m, the servos
needs to be placed in a water­proof container. The size of this box was in turn estimated
from the necessary flight equipment, such as batteries, flight computer and servos.
The weight was estimated as 1 kg of electrical components, and 1.5 kg for the hull and
miscellaneous things. Using OpenVSP1 different concepts were looked at, see appendix
A.

In Fig. 3.1.1a a design using folding wings is shown, the idea being that if the entire
drone could be fit inside the AUV, there is no problem with the forces that comes from
moving a wing through water. The only remaining force impacting the AUV would be
the buoyancy, this was however scrapped early on due to the complexity that comes
with folding a wing, in particular at small scale under water. Instead the drone needs
to be attached on the outside of the AUV, with some other mechanism for dealing with
the forces from the wing. Fig. 3.1.1b shows a more classical design, however it proved
difficult to find a compartment for electronics which could fit inside a circular fuselage.
Having a v­tail also makes the mechanical joints from the servos more complicated,
1
http://openvsp.org/

7
CHAPTER 3. DESIGN AND PROTOTYPING

(a) Folding wings. (b) Tube & wing.

(c) Long nose. (d) Smaller body.

Figure 3.1.1: Different concepts.

and adds a limited gain with thrust vectoring. Fig. 3.1.1c shows a slightly different
approach, with a larger fuselage to have enough space to fit the electronics box. This
design also uses two extra long motor pins to make sure the motors are outside of the
water for takeoff. Fig. 3.1.1d shows a refined version, using a more pointy nose to
make sure the GPS and radio is above the water surface before takeoff. It is also more
compact to minimize the internal volume and enclosed water. The motor pins were
also extended and reused as tail­booms, however this also came with the issue of servo
actuation from the center.

3.2 Design

Fig. 3.2.1a shows the final design with its dimensions together with the first prototype
in Fig. 3.2.1b. The center section consists of a modified B­29 airfoil to ensure enough
space is provided for the electronics box, and some additional margin for operations
such as opening the box. The airfoil used for the wing is SD7080, providing high
lift while keeping a smooth stall behavior. There is no dihedral or twist to ensure
flight stability, but rather this is done by the flight computer in order to simplify the

8
CHAPTER 3. DESIGN AND PROTOTYPING

(a) CAD drawing. (b) First prototype.

Figure 3.2.1: Design of drone.

manufacturing process. As [2] hypothesized to have had problems from a buoyant tail
during egress, the tail here is just made as a flat plate with no buoyant pieces.

The all­up weight (AUW) of the model is 3.3 kg, almost 1 kg heavier than planned.
Much of the extra weight comes from the shell being heavier than originally estimated,
but also some of the equipment being different than planned. The motors are placed
close to the center to generate enough airflow over the elevator during egress, and
at low flight speeds to provide pitch control. Similarly the ailerons are placed far in
towards the center to be within the prop wash for roll stability at low speeds. Yaw
control comes from having two motors, together with small fins for stability when
gliding.

3.2.1 Water flow

The previously mentioned requirements for buoyancy implies that the whole plane has
to be filled with water as not to be too buoyant. The enclosed volume of the shell is 12
l, meaning that the weight of the drone filled with water is nearly four times as much as
empty. During egress this water rapidly has to exit the drone to make vertical takeoff
possible, and similarly it also needs to enter the drone quickly when mounted to the
AUV and diving, as it otherwise functions as a large balloon. To minimize the need
for mechanical actuators a passive drainage solution was sought after, which does not
require any extra components. [7] drained the plane by having extra spoilerons which
could be deflected during egress to let water out, but as any mechanical actuation
needs to be linked to the servos in the box, this turns into complicated mechanical
joints.

9
CHAPTER 3. DESIGN AND PROTOTYPING

(a) Water flow. (b) Leading edge foam.

Figure 3.2.2: Foam And water outlets.

The solution is a wing which has open wing­tips, as well as ailerons placed with a slight
gap to the wing2 , shown in Fig. 3.2.2a. While placing the ailerons further out reduces
the efficiency, this was a fair trade to simplify the system. The hatch seen in Fig. 3.2.2a
is detachable for operations, but is also split in half. The front part is attached with
clips, while the rear part is hinged to the front and attached with a spring to the fuselage.
The spring is just strong enough to help keep the hatch closed during fixed­wing flight,
but during egress it opens due to the water inside. During egress air flows in through
the wing­tips, while water flows out through the tips, aileron gaps and hatch. To be able
to do consecutive starts, the drone has to sink down to a tail­sitter position once landed
or having been put into the water. The nose acts as an air cushion as it is completely
sealed, preventing the drone from sinking down. To make sure the drone moves into
correct position for takeoff, several extra air holes needs to be drilled in the nose and
fuselage to let air out.

3.2.2 Buoyancy

The buoyancy of the plane is regulated by two things. The first is the limitation to
vertical forces when attached to LoLo, meaning that it cannot be too buoyant. The
other one is that it has to float vertically with its tail down at the surface, while also
floating high enough up so that the nose is above the water­line for GPS reception. The
only part which is buoyant is the electronics box, which has a volume of 1l, and as the
drone weighs 3.3 kg additional buoyancy is needed. To generate enough buoyancy for
the drone to float additional foam is added along the leading edge of the wing as shown
in Fig.3.2.2b. This foam also acts as structural support as the wing is simply a shell with
2
The ailerons placement in Fig. 3.2.2a is extreme, not actually placed this far out.

10
CHAPTER 3. DESIGN AND PROTOTYPING

no spars or ribs.

The position of the electronics box is such that the center of it is forward of the center of
gravit (CG), to make sure the center of buoyancy (CB) is far enough forward to balance
the drone with the nose pointing up. The placement of the box is limited due to the
shape of the nose and it is placed as far forward as possible. When the drone rests at the
surface, the motors are above the waterline causing a pitching moment. To minimize
this moment caused by the lever arm from the motor pins, the pins are shortened as
much as possible to keep the motors close to the water surface. The length of the pins
is limited by the need to clear the propellers of the nose.

3.2.3 Nosecone

The entire shell and tail is made of carbon fiber, this means that there is no
communication possible from within the shell and the electronics box for radio
communications and GPS. Furthermore the GPS has to be above the waterline, and
since it works like a tail­sitter it has to be placed far forward in the nose. To solve both
of these problems the nose of the plane is replaced with a more pointy one, made out
of glass fiber. This makes it possible to transmit and receive signals, as well as pushes
both the GPS and radio above the waterline when the drone is resting at the surface. A
layup of how the GPS & Radio is positioned is shown in Fig. 3.2.3a.

3.2.4 Electrical compartment and bladder

A plastic box was used to keep the electrical components dry and as the mission
requires the drone to be submerged down to 10 m depth, the box also needs to be strong
enough to withstand the pressure. One option is to use a compartment with a strong
enough structure to support the pressure on its own, such as [7] using a metal cylinder.
However using a box which is strong enough comes with the addition of extra weight.
To avoid this, a regular plastic food­box from Ikea was used, attached using screws
and rubber pieces. To prevent it from imploding when diving, a separate bladder is
added, seen in Fig. 3.2.3b as a balloon. The idea being to inflate the bladder before the
drone submerges, and as the bladder is softer than the box, it will shrink as the pressure
increases, equalizing the pressure inside the box with the outside. With the addition
of a bladder comes one more buoyant part, which needs to be positioned correctly to
ensure proper orientation at the surface. The size of the bladder is determined by the

11
CHAPTER 3. DESIGN AND PROTOTYPING

(a) Side view of center section.

(b) Eletronics box.

Figure 3.2.3: UAV internals.

amount of volume left in the food­box, thus to minimize the size of the bladder any
remaining space should be filled.

3.2.5 Electrical Components

The eletronics of the plane are placed within the waterproof box as outlined in Fig.
3.2.3b. The flight computer used is a Pixhawk 2.4.1 running Arduplane 4.0.9. For
telemetry two RFD 868 modems are used, and radio­control is done using the FRSKY
r9 system, with a r9sx 868 Mhz. While it is possible to run everything over the RFD
system, two different systems were used for redundancy and to minimize latency. The
power system used is two Turnigy plush 40 A ESC’s which also have built in BEC to
power the Pixhawk. The motors are Turnigy Aerodrive 4240­740 Kv with 15x5.5 inch
propellers. Using two 4s 2200 mAh batteries in parallel this gives a maximal thrust of
6.6 kg, and a Thrust/Weight ratio of 2.

12
CHAPTER 3. DESIGN AND PROTOTYPING

The box used for housing the components transfers heat poorly, and to avoid
overheating the ESC’s are therefore placed outside of the box and waterproofed
separately. The waterproofing is done by coating them in several layers of vinylester.
Another reason to place the ESC’s away from the box is to minimize the amount of
interference with the flight computer. The motors require no protection against water
provided that care is taken so that bushings and similar do not rust.

The GPS was placed in a 3d printed box filled with dielectric gel to protect it from water,
this box was then mounted with screws in the nose as can be seen in Fig. 3.2.3a. While
the initial design called for the radio to also be placed within this box, there was not
enough space and it was instead placed in the electronics box. The concept seen in
Fig. 3.2.3a also called for the servos to be placed inside the food box, using rubber
membranes to actuate the servo movement out to the control surfaces. This proved
too difficult to realize when putting the box together. Both due to lack of space in the
box, but also in making the mechanical joints for the servo actuation. In particular
because the box needed to be easily removable, these connections would also need to
be movable. Instead waterproof servos were used and placed outside the box. The
servos used were HS­5086WP, with an IP67 rating.

13
Chapter 4

Testing
There were many tests done in water to test and tune the buoyancy and water draining
of the drone, as well as the placement of CG and CB. Getting the equilibrium correct
with the tail pointing straight down required much tweaking of the amount of foam
and placement of the electronics box, and the addition of a weight at the tail. To make
takeoff testing easier, it was also sought after to ensure it would position itself correctly,
in a reasonable amount of time, after being placed flat on the surface

4.1 Flight test

Several flight tests were done from the water surface, but no fixed­wing flights were
attempted due to lack of time and space. Many of the attempted starts had problems
due to the drone not being balanced vertically in the water, causing it to tip over. Fig.
4.1.1 shows a failed egress where it starts to take off but tips over slightly and then
proceeds to fall over. As there is very little flow over the tail it has poor pitch control,
which is why it cannot stabilize itself.

(a) Starts to tilt during egress. (b) Drone tilted.

Figure 4.1.1: Failed egress.

14
CHAPTER 4. TESTING

(a) Resting at surface. (b) Step one, motors on.

(c) Wing clearing the water. (d) Water draining.

(e) Hovering.

Figure 4.1.2: Successful egress.

Fig 4.1.2 shows a successful egress, with the full operation taking less than two seconds.
Fig. 4.1.2a shows the UAV’s equilibrium state at the surface. This is where it rests after
floating up, or after being placed flat on the surface, such as after a landing. As can
be noted it is not perfectly balanced in yaw, this due to the bladder being on one side
and different amounts of foam in the wings. In Fig. 4.1.2b step two is shown, which is
throttling up to ≈15 %. This gives enough control for the drone to level out in yaw, and
pulls it up far enough so that the leading edge is at the surface. At this point the foam
is still helping with lifting, as there is foam up to the leading edge.

The next step is to give full throttle, seen in Fig. 4.1.2c which causes the drone to rise
from the surface, and water can be seen flowing out from the wingtips. When the wing
clears the surface it almost comes to a halt, and then continues rising again. While hard

15
CHAPTER 4. TESTING

to see from the figures this can be seen in this test video1 . This appears to happen just
as the box is clear of the surface, and there is no more buoyant parts in the water. This
is also the point in where it differs if it takes off or not, as in Fig. 4.1.1a where it tips
over instead of lifting. Fig. 4.1.2d shows the next step, which is after the entire wing has
cleared the water. There is water flowing out from the center through the hatch, which
is located on the other side in the figure. There is little resistance as the main hull is out
of the water, and the final step of the egress maneuver goes quickly. In fig. 4.1.2e the
final step is shown, with the drone hovering above the surface. Any remaining water
can be seen draining out from the tail­boom. At this point the drone has full control in
all directions, and can either land or throttle up to attempt fixed­wing transition.

4.2 Test videos

Some of the takeoff tests can be seen in the videos below:

• Belly fall and problem with yaw stabilization.

https://www.youtube.com/watch?v=3kB9iG7Qrq0

• Problem with floating state and flight computer.

https://www.youtube.com/watch?v=6TfR_SuLxEE

• Unable to stabilize after clearing water.

https://www.youtube.com/watch?v=wIv5s8gmK0o

• Yaw problems.
https://www.youtube.com/watch?v=wG4eWuWK_O4

• Clean egress.
https://www.youtube.com/watch?v=0Dq2lyXXXBE

• First takeoff test.

https://www.youtube.com/watch?v=rZQBddk5JF0

• Last and best egress.

https://www.youtube.com/watch?v=TPMXwTy6Kpk

1
https://youtu.be/TPMXwTy6Kpk

16
Chapter 5

Discussion

Several water tests were done to tweak the drone before it was able to take off
somewhat reliably. The first takeoff attempt was successful1 , however this after being
placed vertically by hand. The reason it crashed was that the elevator servo came
loose, resulting in loss of pitch control. During the course of the following tests and
development there was about a 1/10 success ratio in terms of successful takeoffs. There
were no fixed­wing flights attempted due to lack of flying­space near the university.
Transporting the drone to a place with enough space to fly is time consuming, and as
many attempted takeoffs resulted in crashes with tweaks needing to be done, this was
postponed. The idea was to refine it enough so that it was able to egress without any
problems before going to a better testing place.

5.1 Development
The first takeoff looks effortless compared to later ones because the drone is floating
higher above the surface, with parts of the wing always visible as seen in Fig. 5.1.1. This
was not a stable state as it would tip over and lay flat on the surface. The following water
tests were done to tune the buoyancy and ensure the plane would float in the correct
position. The amount of foam needed for this was initially overestimated, which led
to the CB being aft of the CG. One reason for this overestimation was that the volume
of the shell and other components were underestimated, and in turn their displaced
volume.
1
https://youtu.be/rZQBddk5JF0

17
CHAPTER 5. DISCUSSION

5.1.1 Buoyancy

The box was placed as far forward as possible to push the CB forward, eventually
limited by the shape of the fuselage. In addition to this a weight of 30 g was added
at the tail. Another solution to this, rather than adding a weight, could be to use
materials with different density, and create a lighter and less stiff tail with a metal
spar for rigidity. Designing the two tails so that the CG of the plane would be the same
for both, once submerged the lesser stiff tail would have a larger weight, neglecting the
need for an external weight.

Takeoff attempts also showed the importance of the drone being placed in a perfectly
vertical position at the surface. Any offset in either direction led to the drone tipping
over as it started to rise, due to the drone having very little control authority in pitch
before the tail cleared the water, which caused several crashes. One potential solution
would be to use the ailerons as elevons during takeoff, however as the position of the
ailerons is close to the CG the lever arm would be very short. When placed flat on the
surface, the fuselage acts as an air cushion as the hatch is placed on the bottom. To
ensure it would sink down to the correct position in a timely manner, extra air inlets
needed to be drilled.

Figure 5.1.1: Early water test.

Once floating properly it was able to egress without much trouble, and hover with

18
CHAPTER 5. DISCUSSION

full control. The water draining worked great as can be seen in Fig. 4.1.2d, with
the whole egress procedure taking less than 2 seconds. Takeoffs were mostly done
in calm conditions with little to no waves, and it is not clear how good the egress
maneuver would be in conditions with larger waves, as it is very sensitive to pitch
disturbances while taking off. There were no depth tests done which makes it difficult
to evaluate how good the bladder worked, the deepest submerging was ≈ 1 m which
worked without any trouble.

5.1.2 Water problems

As all of the tests were done in salt­water, issues arose due to corrosion. The drone
was rinsed with fresh water after each test but this didn’t prove to be enough. The
motors were sprayed with WD40 which worked well, but the connections between
cables gave issues due to rust. This led to instabilities and some connection needing
to be swapped, and while it did not occur during any tests outside, it happened during
an indoor hover test causing power loss. With a more refined solution the box could
be mounted permanently, removing the need for many detachable connectors and the
cables could instead be connected permanently.

Another issue is water creep along cables, which most likely led to at least two
components breaking. While this happened after the drone spent a significant amount
of time in water, it is still something that has to be solved for future development. Using
shrink tube together with hot glue proved to be a solution for some connections where
water entered the cables, but it did not work on all types of connections.

The servos used are only rated for IP67, meaning they are waterproof enough to be
submerged down to 1 m depth, but not further. While it is sufficient to test the takeoff
capabilities as done here, it is not good enough to go down to the desired depth of 10
m.

While the foodbox for electronics was time consuming to fully waterproof with many
connections, once done it was surprisingly reliable. Enduring much vibration during
transport and several crashes during testing, it never broke or leaked. Overall the
biggest damage to the drone occurred from crashing on some rocks, resulting in the
wing cracking. More durable than expected, although a crash from level flight has a
high chance of breaking it.

19
CHAPTER 5. DISCUSSION

An alternative to the foodbox was tested, which was to mount the components on a
plate and then cover them with a soft membrane, this to remove all buoyant parts
except the foam. This proved difficult to waterproof and was very unreliable from an
operational standpoint. Doing any form of maintenance had a high chance of breaking
the seal, and resealing it properly was difficult, in particular around cables.

5.1.3 Electronics

There were initial problems with radio communications as a 2.4 GHz system was used
in the belief that this was enough because the antenna would always be above the
water. However with waves and the drone not always being perfectly straight at the
surface, this proved to be troublesome as the signal frequently dropped. The eventual
equilibrium position which had the nose barely above the surface led to signal problems
even in calm conditions. This system was eventually swapped to a 900 MHz system,
similar to the telemetry, which proved to be more capable. There were also many
issues with the ESC’s as they frequently entered calibration mode when taking off in
stabilization mode. While this can usually be solved through programming, the ones
from YEP caused many issues.

During several takeoff attempts the drone fell over to the side rather than rising, this
can be seen in some of the videos in section 4.2. With two motors, the drone should
be able to correct these types of movements. One reason why this happened might
be that when giving full throttle the thrust vectoring is limited, however even at 90%
throttle this was a problem. It might also be due to EKF problems coming from the
drone having been spun around much in water, which could be solvable by restarting
they flight computer between takeoff attempts. Another concern is that if the plane
pitches far enough, the flight computer thinks of this as a transition to level flight and
tries to correct for that instead of being vertical. Lastly some of the attempted take offs
may have been successful if done with more space, as they had to be aborted to avoid
hitting the shoreline.

As the drone did not fulfill the requirements in terms of flying, it was never tested
together with LoLo as specified. While not presented here, a solution for the mounting
and release mechanism was designed and built, although never tested in water. It can
be seen in Appendix A.

20
CHAPTER 5. DISCUSSION

5.2 Design flaws

CAD design showed that it was feasible to fit all of the servos inside the box (even ESC’s
initially), but this proved difficult to realize. The first problem being the previously
mentioned mechanical part about connecting the servos to the control surfaces through
the box. While there are solutions for getting the actuation through the wall of the box,
such as rubber membranes, making it work while the box is removable proved too
difficult. The second problem was that neither cables, nor the connectors for cables
through the box were included in the CAD models. Different versions of connectors
were used, both home­made using screws, and pre­existing connectors with o­rings.
Both proved waterproof, but they took up significant space on the inside of the box
which would have been needed for the servos. Likewise the amount of cables takes up
a significant amount of space, not leaving much room for potential servos.

Despite these difficulties, it is certainly doable to have the servos inside the box,
but great care should be taken to include cables and connectors when designing.
Using a regular food­box also came with material issues, as the type of plastic did
not allow for any gluing. While the idea of a simple plastic box is good, a better
alternative than a food­box should be used which could allow for simpler connections
and solutions.

The GPS was initially designed to be in its own waterproof compartment without any
gel, and it would be connected to the food box through a pipe so that the pressure
inside it would be equalized. As the food box needed to be removed from the drone
often, this connection between pipe and box needed to be easily detachable. Mounting
such a connection to the box was troublesome for several reasons, and the gel was
used as a simpler solution. Similarly to the waterproofing of ESC’s, this comes with
additional weight which makes it less than ideal for flight. For future development a
better solution should be looked at, which does not come with a weight penalty.

One proposed solution for the heating problem of the ESC’s when placed in the box is to
modify the box by replacing part of the wall with a heatsink. This idea was investigated
but as it was not possible to glue anything to the box, and the space inside already being
limited, it was not used. It is certainly a possible solution, but it requires more planning
and refining of the outline of the components and cables inside the box.

21
CHAPTER 5. DISCUSSION

5.3 Thesis remarks

One reason why the thesis did not manage to build a drone capable of completing the
mission requirements is that there was too much time spent on things which seemed
important at the moment, but in the end turned out to be of little importance to the end
goal. There was time spent on looking at several different airfoils and shapes, which
when looking at the mission goal is of very little use. Most things can fly at this scale, see
flying guitar 2 , and much of the gain of an optimal airfoil is lost due to manufacturing
errors. Finding the best suitable airfoil is important for making the best possible plane,
for a proof of concept it is of little use. Similarly looking at the mission requirements,
there is nothing about flight performance.

During the design and concept phase, many of the concepts for the fuselage originated
from airfoil shapes. While this is aerodynamically efficient and looks good, it is of little
benefit from an operational standpoint. Fitting and accessing a square box inside a
teardrop shape is very difficult, and comes with a lot of wasted space (trapped water).
A better approach would have been to design something very specific for the purpose
of fitting a container, and afterwards trying to ensure proper aerodynamics.
During the course of the thesis, often taking a step back and looking at the bigger
picture would have been beneficial. Putting less emphasis on trying to make each part
as good as possible, and instead focusing on the end goal and what is required for that.

2
https://www.youtube.com/watch?v=npOW99-Cgj0

22
Chapter 6

Conclusions

While the full scope of the project was not completed, parts of it were successful. A
prototype of the drone was designed, built and tested to some (limited) extent. Fixed­
wing flight was never attempted however vertical takeoff from a (brief) submerged
position was successful. The mission specific solutions were mostly successful in the
regard that they were tested. The buoyancy worked well, it returned to a stable vertical
position regardless of how it was placed on or below the surface. While the bladder
was never tested at its design depth, it seemed promising to the extent that it was used.
Using a lighter, non structural box appears to be a valid alternative to the heavier
options. Although time consuming to fully waterproof, once done it kept dry during
multiple crashes and rough handling. The mechanism used for draining the fuselage
of water worked well, a light and passive solution which can be applied elsewhere.
The design of the drone proved to have good control authority, and was capable of
stabilizing itself after egress.

During the course of the thesis there was time spent on details which seemed important
in the moment, but which for the mission requirements was of little use. When
testing and buoyancy tuning it became clear that spending much time on the design, in
particular aerodynamic design, is of little use for a mission like this, and rather it comes
down to testing and adjustments as needed. This time spent on details, together with it
taking a long time to achieve semi reliable takeoffs, led to there not being enough time
to attempt fixed­wing flight.

23
CHAPTER 6. CONCLUSIONS

6.1 Future work

The UAV design seems promising for the task which it is designed for, although it needs
some tweaking and improvements. The water­creep along cables together with some of
the connectors between cables rusting are two big problems, which needs to be solved
before the vehicle can spend a long period of time in water. The time the vehicle spent in
water was enough for components such as screws and push rod connectors to corrode,
and it would be beneficial to replace them with stainless ones.

As previously mentioned the servos are not capable of being submerged 10 m, and
needs to be placed in a waterproof compartment with joints for the actuation. Overall
the interior of the waterproof box should be properly designed and the components
mounted permanently, eliminating the need for many connectors and cables to save
space and weight. However before any larger modifications are made it may be good
to test fixed­wing transition and level flight, to see that it works with the CG position
that is required for correct floating state.

A better solution for the ESC’s and GPS should also be looked at, which is more robust
and lightweight. Placing the GPS in a gel appeared to work well for what was tested
here, but it is still prone to break due to water­creep. The placement of the bladder
also needs more investigation as it shifts the CB once inflated. One idea is to split it
into two smaller ones, and position them in the nose on either side of the GPS box.

Outside of the mechanical problems there is still work to be done to perform

autonomous missions by incorporating the take off procedure into an autonomous
start. Some take­off attempts were done at full throttle, before the vehicle was
stabilized in yaw, which resulted in it tipping over to the side. An autonomous start
should thus have this included, rather than just full throttle. The GPS reception at the
surface also needs to be tested after a cold start. With every test now the vehicle was
started on land and thereafter placed in water, giving plenty of time for GPS connection
before sinking down. As its floating state is very close to the surface, this may cause
issues with GPS reception. Finally the riding capabilities and release mechanism from
LoLo needs to be tested. Both to see the impact it has on LoLo in terms of forces, but
also how the deployment works when moving.

24
Bibliography

[1] AeroVironment. Blackwing. 2021. URL: https : / / www . avinc . com / tms /
blackwing.

[2] Caruccio, Danielle, Rush, Meaghan, Smith, Peter, Carroll, James, Warwick, Peter,
Smith, Eric, Fischer, Caleb, Motylinski, Kevin, Vasconcelos, Lucas F., Costa,
Paulo, and Santos, Dioser F. “Design, Fabrication, and Testing of the Fixed­Wing
Air and Underwater Drone”. In: 17th AIAA Aviation Technology, Integration,
and Operations Conference. DOI: 10.2514/6.2017-4447. eprint: https://arc.
aiaa.org/doi/pdf/10.2514/6.2017-4447. URL: https://arc.aiaa.org/doi/
abs/10.2514/6.2017-4447.

[3] Deutsch, Clemens, Moratelli, Lázaro, Thuné, Sebastian, Kuttenkeuler, Jakob,

and Söderling, Filip. “Design of an AUV Research Platform for Demonstration
of Novel Technologies”. In: 2018 IEEE/OES Autonomous Underwater Vehicle
Workshop (AUV). 2018, pp. 1–8. DOI: 10.1109/AUV.2018.8729729.

[4] Kungsfiskare, Maribot. VTOL drone flying in Idre, Sweden ­ Maribot

Kungsfiskare ­ KTH Maritime Robotics Laboratory. YouTube. 2020. URL:
https://www.youtube.com/watch?v=Inyo0yzCD9w.

[5] L.Waldau. “Development of an aquativ UAV capable of vertical takeoff from

water”. MA thesis. KTH, 2019.

[6] Stewart, William, Weisler, Warren, MacLeod, Marc, Powers, Thomas, Defreitas,
Aaron, Gritter, Richard, Anderson, Mark, Peters, Kara, Gopalarathnam, Ashok,
and Bryant, Matthew. “Design and demonstration of a seabird­inspired fixed­
wing hybrid UAV­UUV system”. In: Bioinspiration & Biomimetics 13.5 (Aug.
2018), p. 056013. DOI: 10.1088/1748-3190/aad48b. URL: https://doi.org/
10.1088/1748-3190/aad48b.

25
BIBLIOGRAPHY

[7] Weisler, Warren, Stewart, William, Anderson, Mark B., Peters, Kara J.,
Gopalarathnam, Ashok, and Bryant, Matthew. “Testing and Characterization
of a Fixed Wing Cross­Domain Unmanned Vehicle Operating in Aerial and
Underwater Environments”. In: IEEE Journal of Oceanic Engineering 43.4
(2018), pp. 969–982. DOI: 10.1109/JOE.2017.2742798.

[8] Z.Ramic. “Rapid prototyping with fiber composites ­ Manufacturing of an

amphibious UAV”. MA thesis. KTH, 2021.

26
Appendix ­ Contents

A Appendix A 28

27
Appendix A

Appendix A

(a) Drone attached to LoLo. (b) Docking mechanism.

Figure A.0.1: LoLo attachment mechanism.

Fig.A.0.1b shows a cradle which was designed to mount the drone to LoLo while
minimizing its impact. The plane would rest in the cradle, attached with a wire from
its belly to the claw seen in the figure. The cradle allows for no movement in roll or
yaw, but free pitching. The idea being that the plane auto­trims itself as LoLo moves,
minimizing the vertical and horizontal forces. The cradle is mounted after the hull on
LoLo is attached, requiring no extra holes or modifications.

28
APPENDIX A. APPENDIX A

(a) Concept blended wing. (b) Concept standard blended wing.

(c) Concept forward sweep. (d) Concept speed racer.

Figure A.0.2: Early Petrel concepts.

29
www.kth.se