IEEE ROBOTICS AND AUTOMATION LETTERS. PREPRINT VERSION. ACCEPTED JANUARY, 2017 1
Efficient Aerial-Aquatic Locomotion with a Single
Propulsion System
Yu Herng Tan
1
, Rob Siddall
1
and Mirko Kovac
1
Abstract—Aerial-Aquatic locomotion would allow a broad
array of tasks in robot enabled environmental monitoring or
disaster management. One of the most significant challenges
of aerial-aquatic locomotion in mobile robots is finding a
propulsion system that is capable of working effectively in both
fluids, and transitioning between them. The large differences
in the density and viscosity of air compared to water means
that a single direct propulsion system without adaptability will
be inefficient in at least one medium. This paper examines
multimodal propeller propulsion using computational tools
validated against experimental data. Based on this analysis we
present a novel gearbox enabling an aerial propulsion system
to operate efficiently underwater. This is achieved with minimal
complexity using a single fixed pitch propeller system, which can
change gear underwater by reversing the drive motor, but with
the gearing arranged to leave the propeller direction unchanged.
This system is then integrated into a small robot, and flights in
air and locomotion underwater are demonstrated.
Index Terms—Mechanism Design, Micro/Nano Robots,
Biologically-Inspired Robots, Aerial Systems: Mechanics and
Control, Field Robots
I. INTRODUCTION
L
OCOMOTION in unstructured terrain is a significant
challenge to miniature robots, often requiring operation
in water, air and on the ground. In particular, aerial-aquatic
robots face major challenges, and must accommodate the
increased structural loads, fluid inertia and drag encountered
underwater, without compromising the weight and lifting area
requirements of flight.
Addressing these challenges would allow unique robot
operation in a wide variety of environments, such as ti-
depools, wetlands or canal systems, enabling autonomous
monitoring of contaminants and ecosystem health. To achieve
this, we are developing a novel robot, called the Aquatic
Micro Air Vehicle (AquaMAV) [1] which is capable of aerial
and aquatic locomotion. The AquaMAV will be able to dive
directly into the water at high speeds to achieve initial depth,
subsequently retaking flight using a high powered burst of
thrust (figure 1), as demonstrated in [2]. But while able to
escape water, this robot had no means of propelling itself
underwater. To add aquatic locomotion it is attractive to use
Manuscript received: September 10 2016; Revised December 7 2016;
Accepted January 19 2017.
This paper was recommended for publication by Editor Jonathan Roberts
upon evaluation of the Associate Editor and Reviewers’ comments.
1
Department of Aeronautics, Imperial College London,
London, SW7 2AZ, UK. Correspondence should be directed to
r.siddall@imperial.ac.uk. A video attachment to this
paper is available online: tinyurl.com/AerialAquaticPropulsion
Digital Object Identifier (DOI): see top of this page.
Fig. 1: Concept sketch of an Aquatic Micro Aerial Vehicle
diving into the water, gathering data and retaking flight.
152 mm Propeller Diameter
280 mm Wingspan
Elevons
150 mm Chord
Multimodal
Epicyclic Gearbox
Drive motor
Battery
Radio Receiver
with Control Servos
Total Mass:
50 grams
Fig. 2: Miniature 50 gram aircraft which uses a dual mode
gearbox to achieve energetically efficient aerial-aquatic loco-
motion with a single propeller.
the same propulsion system as is used for flight, as this
reduces the weight and complexity of the system. However,
the increase in load a propeller in a denser fluid results in
much slower rotation speeds, and means a significant loss of
motor efficiency.
In this paper, we will examine the efficiency of propeller
operation in air and water, using a blade element code
(QPROP [3]) to investigate a variety of propeller geometries
commonly used in Micro Aerial Vehicle (MAV) propulsion.
These theoretical predictions are then validated against un-
2 IEEE ROBOTICS AND AUTOMATION LETTERS. PREPRINT VERSION. ACCEPTED JANUARY, 2017
derwater thrust tests, and the model is used to investigate an
ideal aerial-aquatic transmission. Based on the results of these
investigations, we propose a bimodal epicyclic transmission
to allow efficient propeller operation in air and water. This
gearbox is then fabricated and efficient propulsion of a
miniature aircraft (figure 2) in air and water is demonstrated.
By using the direction of the motor to control the setting
of the gearbox, the proposed design provides a simple and
effective solution to achieving efficient aerial-aquatic locomo-
tion. Compared to previous designs for multimodal operation
in air and water, this system significantly improves the overall
performance of the propulsion system in the two media, with
minimal mechanical complexity and no additional actuators.
II. E
XISTING AERIAL-AQUAT I C ROBOTS
Although the development of fully functional aerial-
aquatic robots is still in its early stages, several prototypes
have demonstrated multimodal abilities to varying degrees
[4], [5]. A summary of developments presented in [6] clas-
sifies the types of aerial-aquatic vehicles according to the
degree of autonomy and operation functions. There exists
an abundance of vehicles with mission profiles that include
contact with water, such as seaplanes with surface take-off
capabilities or submarine-launched UAVs. However, neither
of these classes of vehicles are capable of self-propulsion un-
derwater. Seaplane-type vehicles are too buoyant to submerge
themselves and submarine-launched UAVs are typically as-
sisted by discardable support systems underwater, before they
transition to the air.
Producing an aerial-aquatic system which is capable of
active propulsion in both media has proved challenging, with
few successful prototypes presented in recent years. Multiple
studies have investigated the possibility of aerial-aquatic
locomotion with a single propulsor, a preferred approach
since weight has a major impact on aerial performance. [7]
and [8] investigated the use of flapping foils for dual mode
propulsion, using changes in morphology and kinematics to
ensure efficient thrust production in different media. Both [7]
and [8] demonstrated the efficacy of this strategy in tunnel
tests. On a much smaller platform, [9] demonstrated aerial-
aquatic locomotion with an insect scale vehicle. This system
used flapping wings for both aerial and aquatic propulsion,
with changes in stroke rate resulting in dynamically similar
flows in both media. However, the robot was powered exter-
nally so energetic efficiency was not of immediate concern.
Recently, quadrotor platforms with hybrid aerial-aquatic
capabilities have been demonstrated, using aerial propellers
for locomotion underwater [10], [11]. [10] uses buoyancy
control to transition to a submersible mode, while [11]
demonstrates direct air-water transition using a dual-propeller
system. Both robots use the same motor-propeller combina-
tion in air and water. Although both demonstrated the ability
to move underwater, the use of aerial propulsion system in
off-design conditions results in the system operating at very
low efficiencies.
0 2000 4000 6000 8000 10000 12000 14000
Rotation rate (rpm)
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Torque (Nm)
Torque-Speed Characteristics in Air and Water
152 x 76 mm propeller in water
152 x 76 mm propeller in air
40 W Brushless Motor
Max motor
efficiency
Max motor
Max motor
Power out
Power out
Operating point in water
Thrust = 4.8 N
Operating point in air
Thrust = 1.4 N
Fig. 3: Torque speed curves for a 152 mm diameter, 76
mm pitch propeller in air and water and a 40 W example
brushless motor. We have marked the rotation rate at which
the motor operates at maximum efficiency (η
M
, equation
2) and produces maximum output power (Qω) with dashed
lines. The torque-speed curve of the propeller shifts steeply to
the left in water, increasing equilibrium torque and decreasing
operating rpm. In air, the propeller operates at near maximum
motor efficiency, but in water the motor is at only 6% of its
peak efficiency speed.
III. COMPUTATIONAL INVESTIGATIONS
The differences between electric propeller operation in air
and water were first investigated by analysing motor-propeller
combinations using QPROP, examining the effect of propeller
geometry, motor parameters and operating conditions on
the thrust production and efficiency of the overall system.
QPROP computes the steady-state behaviour of propeller
systems using an enhanced blade-element and vortex met-
hod built on the method of Larrabee [12]. It computes an
accurate propeller aerodynamic model by accounting for the
propeller’s self-induction.
The motor-propeller system is defined by three files con-
taining operating conditions, motor properties, and propeller
geometry: The motor is described by a linear model (defined
by the zero-load current, internal resistance and characte-
ristic rpm/V) while the propeller is defined geometrically
using a series of chord and twist values along the blade,
and aerodynamically by section lift and drag coefficients.
The simulation’s fidelity is then principally limited by the
accuracy of propeller geometry and aerodynamic coefficients.
The propellers used in this study are commercially availa-
ble components, commonly used in small scale aerial robots.
We refer to these props as, for example ‘152x76mm’ referring
to a propeller with a 152mm diameter and a 76mm blade
pitch. The propellers tested in section IV were modelled using
a 3D laser scanner (FARO Edge ScanArm, 25µm accuracy)
to generate surface point clouds (see section IV, figure 5B).
These were then parsed into QPROP input files using Matlab.
TAN et al.: EFFICIENT AERIAL-AQUATIC LOCOMOTION 3
Velocity (m/s)
012345
Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Effect of Throttle on Underwater Efficiency
Propeller Efficiency,
Motor Efficiency,
Locus of maximum
Velocity (m/s)
0 5 10 15 20 25 30
Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Underwater Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Total Efficiency,
T
Velocity (m/s)
0 5 10 15 20 25 30
Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Aerial Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Total Efficiency,
T
Velocity (m/s)
012345
Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Effect of Throttle on Underwater Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Locus of maximum
T
points
Velocity (m/s)
0 5 10 15 20 25 30
Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Underwater Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Total Efficiency,
T
Velocity (m/s)
0 5 10 15 20 25 30
0
1
Aerial Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Total Efficiency,
T
Velocity (m/s)
012345
Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Effect of Throttle on Underwater Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Locus of maximum
T
points
Velocity (m/s)
0 5 10 15 20 25 30
Efficiency
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Underwater Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Total Efficiency,
T
Velocity (m/s)
0 5 10 15 20 25 30
Aerial Efficiency
Propeller Efficiency,
P
Motor Efficiency,
M
Total Efficiency,
T
8 V
Motor Voltage = 2 V
A B C
Fig. 4: Efficiencies curves of the same motor-propeller system in air (A) and in water (B) showing that the properties of
the medium greatly affect the matching of the system. Efficiency curves shown are for the coupled motor-propeller, such
that the forces produced by the propeller effect the operating speed of the motor and vice versa. Beyond a certain forward
speed η
P
< 0 and the propeller is creating drag. (C) Shows the effect of throttling the motor on underwater efficiency
by increasing the applied voltage from 2V to 8V. The locus of maximum total efficiency points is marked, showing that
throttling has little effect on total efficiency for a mismatched system. Efficiency curves at lower voltages are shown with
increased transparency.
A cylinder is fitted to the central hub propeller hub to locate
the blade’s rotation axis, and cross sections are taken of the
blade scan at 30 equally spaced sections, from which chord
length, pitch angle and airfoil sections can be extracted for
input into QPROP.
Our objective in this investigation is to maximise the
efficiency of a propeller propulsion system, achieved by ma-
tching propeller and motor efficiencies. Propeller efficiency
is defined by the ratio of propulsive power out (thrust, T
multiplied by forward speed, v) to shaft power in (shaft
torque, Q multiplied by angular speed, ω),
η
P
= T v/Qω (1)
and motor efficiency is the ratio of shaft power to electrical
power:
η
M
= Qω/V I (2)
where V and I are the input voltage and current. Total system
efficiency is then:
η
T
= η
P
η
M
= T v/V I = P
o
/P
i
(3)
where P
i
and P
o
denote input and output power. Efficiency
is therefore zero when static and v = 0, and is strongly a
function of forward velocity, which determines the relative
motion of propeller blades to the surrounding fluid.
However, to highlight the problems of operating aerial
propellers underwater, we have first analysed the torque
requirements of a static 152x76mm propeller in air and water,
driven by a 10 gram brushless motor currently used for aerial
propulsion in an AquaMAV prototype [13]. The motor has a
peak output power of 40 W, and an unloaded speed of 2000
rpm/V. Figure 3 shows propeller torque against rotation speed
in water and in air, and the torque-speed characteristic of the
case study motor. The motor characteristic is given by a first
order model of the motor:
Q =
(V −
ω
K
v
)
1
R
− I
0
1
K
v
(4)
where the relation between the shaft torque, Q and rotational
speed, ω, is given by voltage, internal resistance (R), no-
load current (I
0
) and rpm/V (K
v
), the latter three being
characteristic values of a given motor. The rotation speed
at which a motor produces maximum output power (Qω)
and maximum efficiency (η
m
, equation 2) can be calculated
analytically, and are marked on figure 3. When the motor
output is connected to a propeller, the operating point of the
motor-propeller system is at the intersection of the motor and
propeller torque characteristics (figure 3). Here it can be seen
that the increased fluid density in water shifts the propeller
characteristic backward significantly, forcing the system to
operate at a much lower rpm. Here, reduction in speed means
that the two propeller flows are dynamically similar, and
blade tip Reynolds numbers,
Re = ρωD
2
/µ (5)
where ρ is the fluid density, µ viscosity and D propeller
diameter, are 1.08×10
6
and 1.20×10
6
in air and water
respectively. However, the thrust produced in water is 4.8 N,
over three times the thrust produced in air (1.4 N). So while
the motor can produce significantly more force underwater,
it must do so at 10% of its maximum power output speed,
and 6% of its maximum efficiency speed.
The matching problem of aerial aquatic operation can
be seen more clearly when forward velocity is taken into
account. In figure 4A and 4B the efficiencies of the same
4 IEEE ROBOTICS AND AUTOMATION LETTERS. PREPRINT VERSION. ACCEPTED JANUARY, 2017
50
0
-50
20
10
20
-80
40
60
80
-60
-40
-20
0
0
Z
Y
X
0 100 200 300 400 500
0
0.5
1
1.5
2
2.5
3
152 x 76 mm Propeller Underwater
Experimental results (2-blade)
QPROP (2-blade)
Experimental results (3-blade)
QPROP (3-blade)
0 100 200 300 400 500
0
0.5
1
1.5
2
2.5
3
127 x 76 mm Propeller Underwater
Experimental results (2-blade)
QPROP (2-blade)
Experimental results (3-blade)
QPROP (3-blade)
0 100 200 300 400 500
0
0.5
1
1.5
2
2.5
3
178 x 89 mm Propeller Underwater
Experimental results (2-blade)
QPROP (2-blade)
Experimental results (3-blade)
QPROP (3-blade)
D FE
Thrust (N)
Angular Velocity (rpm)Angular Velocity (rpm) Angular Velocity (rpm)
Thrust (N)
Thrust (N)
A CB
178 x 89 mm
152 x 76 mm
127 x 76 mm
Scanned 178 x 89 mm Propeller
Fig. 5: Thrust tests: A: 3-blade variants of propellers tested. B: Point cloud from 3D scan of propeller geometry. C:
Experimental set up, showing motor and electronics with case removed, and full assembly mounted to a waterproof force
sensor. D-F: Experiments results for six different aerial propellers used underwater, compared with QPROP predictions, with
error bars shown on measured data. Two and three blade variants of three different propeller sizes were tested. Propellers
matched well to simulation.
motor-propeller combination are plotted based on QPROP
output data over a range of forward velocities in air and
water. Beyond a certain speed, the propeller cannot rotate
quickly enough, the propeller creates drag rather than positive
thrust, and η
P
< 0. It can be seen that while maximal
propeller efficiency is not greatly reduced in water (η
P
=
69% compared to 75% in air), the motor speed is much
lower, forcing the system to operate at a very low efficiency
(η
T
= 5% compared to 51% in air). This highlights the key
problem in using an aerial propulsion system underwater;
it is difficult to achieve good motor-propeller matching in
both media. Conversely, if the motor used has a torque
characteristic appropriate to the higher resistance in water,
the low maximum rpm will result in negligible thrust in air.
The curves shown in figure 4A and 4B show the motor
operating at full power. When operating at lower power,
the efficiency of the system is expected to increase. This
is because the equilibrium torque of the system increases
with the voltage applied, increasing the mismatch between
the motor design torque and required torque. In figure 4C,
the effect of changing the motor supply voltage on efficiency
is shown. The motor is simulated operating from 2V to
8V, and the maximum total efficiency at each voltage is
calculated. The locus of these points is plotted in figure 4C.
Propeller and motor efficiency curves are also shown for
each simulated voltage, plotted with decreasing transparency
as motor voltage is increased. The results show a small but
insignificant increase in total efficiency as voltage is reduced
(from 4.2% at 2V to 4.6% at 8V). This shows that although
the vehicle will not necessarily always operate at full throttle
when travelling, reducing motor power has little effect on
total system efficiency.
IV. EXPERIMENTAL VERIFICATION
In order to verify the computational predictions made using
QPROP, several propellers were tested underwater (examples
shown in figure 5A) and torque-speed curves were recorded,
using encoders (Pololu 12CPR Magnetic Encoder) and a
6-axis force balance (ATI Gamma IP68). Because of the
dynamic similarity of the flows in air and water (section III),
testing in water was deemed sufficient to validate prediction
in both media. In addition, testing in water has advantages
over testing in air, as the slower rotation speeds and larger
forces in water can be measured more accurately with sen-
sors.
During testing, propellers were driven by a 10 W brushed
gearmotor with a 50:1 gearbox (Pololu 50:1 HPCB6V),
which could provide the torque necessary to drive the pro-
pellers underwater, without drawing damaging current loads
(QPROP simulations showed a large current draw beyond
the limits of safe operation when the brushless outrunner
described in section III was used underwater).
The encoder mounted to the back of the motor was used
to measure rotational speeds of the motor shaft, with an
accuracy of 600 pulses per revolution of the output shaft. The
motor and encoder were contained in a sealed streamlined
housing, mounted to the force balance via a 30cm aluminium
strut. A propeller drive shaft was passed through the casing
TAN et al.: EFFICIENT AERIAL-AQUATIC LOCOMOTION 5
using a sealed bearing and connected to the motor output via
an Oldham coupling (figure 5C).
The motor power was controlled using pulse-width modu-
lation (PWM), with all control and data acquisition performed
in the LabView environment. Each propeller was driven from
10-100% PWM duty cycle, increasing in 10% increments.
For every measurement thrust and rpm data were recorded
for 10 seconds and averaged.
A. Results and Analysis
The results of three propeller tests are shown in figures
5D-F, with thrust curves shown for tests with 2 and 3 blade
variants. The results show a close agreement with QPROP
predictions across the three propellers presented.
Three-blade propellers were found to generate more thrust
than a two-blade propeller with identical blade dimensions as
expected, with an additional blade increasing the resistance
load on the propeller as well as the lift produced by the
blades. This increases the torque load at equilibrium for the
motor-propeller system, resulting in higher current draw and
thrust compared to two-blade propellers. As the motor used
has a relatively high design torque, the system was generally
well-matched across the range of propellers tested, allowing
a large range of speeds to be tested until the motor reached
its maximum output torque.
The computational predictions were found to be a close
fit to the experimental data and show that the simulations
provide an accurate representation of the actual system.
A possible source of the minor discrepancy between the
theoretical estimate and the measured data is the effect of
significantly higher hydrodynamic forces acting on the blade
underwater causing slight deformation near the tips. This
flexibility was not accounted for in the QPROP simulations.
Nevertheless, the experimental results were found to match
the simulations closely, indicating that any effect caused by
this deformation was not significant. Overall efficiency of the
system cannot be concluded from the experiments as only
static tests were conducted, meaning that η
P
=0. However,
other investigations in moving flowfields have also shown
QPROP to be quite accurate at prediction of propeller flows
at similar Reynolds numbers [14], [15], [16].
This confirms the results from section III that using an
aerial propulsion system directly in off-design conditions will
result in highly inefficient propulsion in water, in addition to
damagingly high current draws for motors not designed to
sustain high torque loads. In seeking a compromise between
aerial and aquatic performance, the torque requirements to
achieve desirable aquatic propulsion would result in a large
reduction of thrust in air. As the operating rpm and torque
is an equilibrium point based on the motor-propeller com-
bination, a significant improvement in performance can be
achieved by using a more flexible system that is capable of
altering one of these variables to suit the operating medium.
V. AERIAL-AQUATIC LOCOMOTION
From the propellers investigated, it is clear that using
the same motor-propeller combination for multimodal loco-
motion will be highly inefficient in at least one medium.
This problem could be addressed to some extent by variable
pitch propellers, but the mechanical complexity makes this
challenging to implement for small scale vehicles. A more
straightforward alternative is the use of two separate propul-
sion systems, optimised independently. However, the vehicle
would have to carry the weight of an unused system at any
point in its mission, and the inactive propulsor may also incur
drag penalties.
Because the problem is not that aerial propellers are
necessarily inefficient underwater, but that motors are poorly
matched, the use of a variable transmission to ensure good
matching is sensible. However, this again may be difficult
to implement at the small scale, and requires a system
for changing gear between media. Rather than employ an
actuated gearbox, which incurs a weight and complexity
penalty, we propose that reversing motor direction is a simple
and lightweight means of controlling a two-speed gearbox for
an aerial-aquatic robot.
To investigate the efficacy of a transmission system, we
have used QPROP to compute an ideal transmission for an
aerial propulsion system operating underwater. In table I,
we list several propellers and compute the gear reduction
which maximises efficiency underwater, if the 40W brushless
motor described in section III is used as a drive. Across
the range of propellers simulated, it can be seen that using
the optimal gearing gives an order of magnitude increase
in efficiency underwater. This comparison once again shows
the significance of motor-propeller matching in efficiency.
Specifically, large gains can be made by adjusting the torque
characteristics of the motor, while varying the diameter and
number of blades within the design range contributes much
less to performance. Under optimal gearing, the underwater
efficiency is capable of obtaining a similar range as the aerial
efficiency. This means that the system is able to produce
significantly more thrust underwater, whilst also drawing a
smaller amount of power. The power required underwater is
also of a similar magnitude to that required in air, showing
that underwater locomotion will not place any strain on the
electronics of the vehicle.
A. Gearbox Mechanism
The simplest way to control the gearbox is to use the drive
shaft to automatically engage the gearbox when operating
in one direction (water mode) and disengage the gearbox
in the other direction (air mode or direct drive), avoiding
the complexity of a mechanical gear change, and the need
for additional actuators. However, in both modes the output
needs to be spinning in the same direction as the propeller is
unchanged. In order to achieve this, the gearbox must reverse
the direction of the output when engaged, which is done
using a planetary gearset in fixed-carrier mode (figure 6A-
D). A second epicyclic stage in fixed-ring mode is added
after the first to achieve the full reduction required, while
preserving the original direction change. Both the output and
input driveshaft to the gearbox are connected to the propeller
using separate sprag clutches, which permit rotation in only
