Autonomous Systems Lab

Prof. Roland Siegwart

Bachelor-Thesis & Semester-Thesis

Design and Evaluation of a

Fin-Based Underwater
Propulsion System
Spring Semester 2009

Supervised by:
Prof. Dr. R. Y. Siegwart
Dr. C. Pradalier
W. Fischer

Authors:
Benjamin Peter
Roman Ratnaweera

Abstract
In search of underwater locomotion methods as alternatives to propellers, systems
relying on the propagation of waves along a fin have already been designed and
evaluated by several scientists. Considerable effort has been undertaken to optimise their efficiency both by fluiddynamic analysis and experiments on physical
prototypes. One drawback of the systems hitherto has been their electro-mechanical
complexity in that they required many actuators and refined control strategies to
generate the wanted fin undulation. Our approach has been to translate the result of these optimisations into a simpler, purely mechanical model relying on the
principle of camshafts to achieve a similar undulatory fin motion. The goal was to
evaluate whether this type of propulsion system is feasible and whether it will be a
viable alternative to propellers in Autonomous Underwater Vehicles in the future.
The prototype built during the project, CuttleFin, reached comparable speeds to
other undulating robot solutions. Force measurements also showed that the thrust
produced is in qualitative accordance to a simplified fluid dynamics model. This
makes the camshaft approach a promising option for generating an undulating wave
in a membrane-based fin propulsion system, if one is willing to pay the price of lower
flexibility compared to current dexterously actuated solutions.

ii

Contents
Abstract

1 Introduction
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Concurrent Research at ASL . . . . . . . . . . . . . . . . . . . . . .
1.3 Project Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1
1
2
2

2 State of the Art

2.1 Biological Context . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Existing Underwater Robots . . . . . . . . . . . . . . . . . . . . . . .

3
3
5

3 Basic Concept
3.1 Problem Parameters
3.2 Solution Alternatives
3.3 First Prototype . . .
3.4 Manoeuvrability . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

9
9
9
14
18

4 Detailed Design
4.1 Parameter Decisions .
4.2 Fluid Dynamics Model
4.3 Camsystem . . . . . .
4.4 Frame . . . . . . . . .
4.5 Steering . . . . . . . .
4.6 Cam Coupling . . . .
4.7 Specific Wavelength .
4.8 Power Unit . . . . . .
4.9 Fin . . . . . . . . . . .
4.10 Catamaran . . . . . .

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

19
19
20
24
25
25
28
28
29
30
32

5 Evaluation
5.1 Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Measurement Procedure . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

35
35
36
38

6 Conclusion
6.1 Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45
45
45
46

iii

iv

Chapter 1

Introduction
Even though robotics will never achieve the same level of elegance and perfection
as nature, insight can be gained by looking at biological systems for inspiration.
The dream of this line of projects is to design a Autonomous Underwater Vehicle
(AUV) for long term data harvesting in environments such as coral reefs. Figure
1.1 sketches a rough idea of how the shell of a turtle could be combined with the
locomotion concept found in cuttlefish (see figure 1.2). The first step of this project
was to evaluate this type of propulsion and analyse if it was applicable to the task.

Figure 1.1: Design idea for propulsion of a turtle-like AUV (M.

Svoboda)

1.1

Motivation

According to scientists such as K.H. Low [18] [28] and M. Sfakiotakis [25], there is a
demand for efficient underwater propulsion systems that exhibit dextrous manipulation. The principle of undulation presents a promising candidate for these demands.
For military applications, noiseless propulsion and an inconspicuous wake would be
additional incentives. Our reasons for seeking alternate methods of propulsion to
propellers were threefold. One, according to our supervisors prior experience with
AUVs, propellers are prone to entanglement in rich marine environment (algae and
wires) and a more robust system in this respect is desirable. Two, propellers are
1

Chapter 1. Introduction

known to harm wildlife due to their noise emission (for example see [20]) and scarring whales. And three was the expectation of exploring something new.

Figure 1.2: Cuttlefish. The waves propagated along its fins

serve as the role model for robotic imitation. (Photography by
J. Messersmith)
Upon closer inspection it soon turned out that the field was not as novel as we
believed. Analysis of the kinematics of undulating fin was conducted by Sir James
Grey as early as the 1930s. With continued theoretical work as in [14], research has
already produced some successful prototypes such as by Professor Low and others
(see section 2.2) with efforts at optimisations in physical prototypes [5]. Building on this work, our approach has been a mechanical implementation relying on
camshafts to generate the propagating wave. This brings about a dramatic reduction of necessary actuators compared to other undulatory robots (see section 2.2).
This simplification comes at the price of limited flexibility of parameter variation.

1.2

Concurrent Research at ASL

At the Autonomous Systems Lab (ASL) of ETH Zurich, there were two concurrent
and independent underwater robot projects in 2009. Naro [6] was a larger scale
student project that spanned for one academic year, involving six students and
corporate funding. The goal of naro was to design and build an energy-efficient fish
robot, which could imitate the natural motion of fish.
The project presented here is on a far smaller scale of two students during one
semester. The exact problem description will be outlined in section 1.3.

1.3

Project Definition

The project definition was to design and evaluate innovative locomotion concepts
for AUVs with a focus on a membrane-based undulating fin as seen in cuttlefish. In
a preceding project, Benjamin Peter conducted a survey on fin-based underwater
robots [24]. Some of this prior knowledge is presented in chapter 2 and the basic
theoretical background is introduced in chapter 3.

Chapter 2

State of the Art

2.1

Biological Context

An excellent review on the swimming modes for aquatic locomotion is [25]. For the
benefit of the reader, we will give a short summary of the relevant parts for this
project in order to introduce the terminology.

Figure 2.1: Swimming modes classification. (Adapted from [25])

The most common way of generating thrust in fish is by propagating a backwardmoving wave through their body and/or caudal fin, called BCF locomotion (see
figure 2.1). Fewer species rely purely on what is called the median and/or pectoral
fins (MPF) locomotion method to generate their forward thrust. Our inspirational
example, the cuttlefish, is one member of this class. However many fish families,
even those relying on BCF for propulsion, use MPF for manoeuvring and stabilisation [26]. An extreme example of maneoverability is the seahorse, studied in [1].
3

Chapter 2. State of the Art

According to Breder and Edgerton, there are several factors that can be varied in
the undulation to achieve this dexterity, among which are: interdistance, length
and flexibility of fin rays and the amplitude, frequency and phase lag along the fin
in time [25] [1]. The variations of these parameters allows the seahorse to precisely
adjust the force components generated by its dorsal and anal fins.

Figure 2.2: (a) FN is the resulting force of a single ray.

(b) shows that the force vectors of several rays along
the fin adds up to FP parallel to the fin axis. (c) is the
perspective view. (Taken from [25])

Figure 2.2 illustrates how the forces generated by the oscillating deflection of a
single ray lead to a forward thrust perpendicular to deflection plane.
Depending on which of the fins (see figure 2.4) of a fish are responsible for thrust
generation, the locomotion form is called amiiform, gymnotiform or rajiform, shown
in figure 2.3.

Figure 2.3: Amiiform, gymnotiform and rajiform fin

undulations (Taken from [18])

2.2. Existing Underwater Robots

Figure 2.4: Fish anatomy nomenclature

2.2

Existing Underwater Robots

Several bio-inspired AUVs have already been built. Despite the fact that all of these
robots aim to provide alternatives to propellers, their propulsion mechanisms differ
substantially.
Some examples of bio-inspired AUVs:
Aqua ray, a manta ray robot. Several fluidic muscles serve as actuators. [8]
Madeleine, a turtle robot. It uses lift effects of its oscillating fins for thrust
production. [11]
AquaPenguin, a new AUV with a very bendable body that can move smoothly
in any direction. [9]
AQUA, a six legged amphibious robot which is able to swim as well as walk
on shore. [3]
University of Essex fish robot. It uses only one motor to actuate its caudal
fin. [16]
Of course, many other underwater robots exist or are still in development. This list
should only provide an idea of the diversity of AUVs.
For our project we focused on finding and analysing other robots that use fin undulation as primary method of propulsion. Some examples of AUVs using this kind
of locomotion are:
Cuttlefish robot from Nanyang Technological University [17]
Knifefish robot from Northwestern University [4] [5]
Note that all robots we encountered had a large number of actuators, most of them
one actuator per ray. This raised the question if it is feasible to implement fin
undulation with only few actuators, preferably one.

2.2.1

Cuttlefish Robot

Low and his team from Nanyang Technological University in Singapore developed
underwater robot inspired by the cuttlefish (figure 2.6) [17]. However, contrary to
the most common approach of angular ray deflection, they implemented a mechanism with parallel ray deflection.

Chapter 2. State of the Art

Figure 2.5: On the top left: madeleine, a turtle robot. On the top
right: AQUA, a six legged amphibious robot. Middle row: Two
robots from the company festo, Penguin robots called AquaPenguins on the left and the manta ray robot aqua ray on the right.
Bottom row: Fish robot developed by the University of Essex [16]
on the left and Nautical Robot (naro) by ETH Zurich [6] on the
right.

Figure 2.6: One fin of the cuttlefish robot

with parallel ray deflection, developed at
the Nanyang Technological University [17]

2.2.2

2.2. Existing Underwater Robots

Knifefish Robot

Epstein, Colgate and MacIver from Northwestern University in Evanston, Illinois,

USA built a knifefish robot (figure 2.8) [4]. It was inspired by the South American
black ghost knifefish (figure 2.7), a peculiar fish relying on fin undulation as a means
of locomotion. It passes waves along its big anal fin, without bending its body. The
Knifefish is a dextrous swimmer, able to move in nearly any direction. This type of
locomotion is similar to the movement observed in cuttlefish, the difference being
the gymnotiform placement of the undulating fin.

Figure 2.7: South American black ghost knifefish

The knifefish robot ribbon fin has eight rays, each driven individually by a servo
motor. It is designed modularly, The rays have a maximum amplitude of 40 . The
fin is 76.2mm wide and 231 mm long.
With this robot, they varied three parameters: amplitude, frequency and specific
wavelength and measured the resulting thrust [5].

Figure 2.8: Knifefish robot from the Northwestern

University [4]

Chapter 2. State of the Art

Chapter 3

Basic Concept
This chapter will introduce the principle behind fin undulation as means for locomotion.

3.1

Problem Parameters

As mentioned in section 2.1, there are several physical and behavioural factors that
can be varied when considering fin undulations. Table 3.1 shows an overview of
those we deemed relevant and which will be considered hereforth.
Table 3.1: Main parameters
Parameter
Frequency
Amplitude
Number of rays
Waveform
Phase delay
Ray distance
Wavelength
Specific wavelength
Fin length
Fin width

Symbol
f

N
(t)

dray

w
L
b

Description
Frequency of the backpropagating wave
Maximum Amplitude of the fin wave
Number of rays
Angle of each ray as a function of time
Phase delay between two adjacent rays
Distance between to adjacent rays
Wavelength of the fin wave
Ratio of wavelength to total fin length
Total length of the fin
Width of the fin

Parameter relations:
Fin length: L = (N 1) dray
Wavelength: = L

1 2
N 1

Specific wavelength: w =

3.2

Solution Alternatives

As can be seen in the project definition (section 1.3), the initial choice of mechanism
was quite free. A lot of design decisions had to be made in a short time. Table 3.2
summarises some of the options we considered.
9

Chapter 3. Basic Concept

10

Figure 3.1: Problem parameter visualisation.

Maximum deflection amplitude , phase delay between rays , distance between rays dray and fin
width b.

3.2.1

Actuator

There are two types of actuators that have successfully been used in the robots discussed in section 2.2. Those are servomotors and pneumatic cylinders. Microrobots
using shape memory alloys [27] [10] or polymer fins as artificial muscle [12] [23] have
also been reported. Not aiming to reproduce any of them, it was decided to use
a motor instead and rely on a cam-system (section 4.3) to induce the propagating
wave.

3.2.2

Number of Actuators

The choice of a cam-system left the question of having one motor per fin or even one
motor per AUV. The concept of cheap design suggests exploiting the environmental
niche to reduce the complexity of the robotic system as far as possible. For example,
a one-degree-of-freedom fish called Wanda developed at the Artificial Intelligence
Lab of the University of Zurich [30] exhibits surprising manoeuvrability in threedimensional space considering the fact that it only possesses one actuator.
With demands on intuitive controllability (see section 3.4) however, it was decided
to design for one actuator per fin.

3.2.3

Steering

Two concepts for steering were tested in our first prototype, please see section 3.3.2
for details.

3.2.4

Deflection Type

Parallel deflection of the rays is used in [18], which proved to perform well in their
special configuration. However, there were two arguments against this option from
the authors point of view. One was the mechanical complexity that was encountered
when trying to design such a system for the camshaft case. As can be seen from the
quoted paper and [19], even in larger projects a great deal of effort is spent to the
design and control of this parallel mechanism. The second reason was the question
of water proofing an AUV that employed a parallel-moving fin. The argument is
that the angularly deflecting fin allows to seal off the interior at the rotation point
more readily.

11

3.2. Solution Alternatives

Table 3.2: Morphological box.

Function
Actuator

Motor

Servo

Number of actuators

One per ray

One per fin

Deflection type

Angular

Parallel

Steering

Rudder

Ray

Joint closure

Springs

Groove

Fin attachment
Camshaft mounting

Sliding
Fixed

Fixed
Exchangeable

3.2.5

Solutions
Artificial
muscle

One
AUV

Shape Memory Alloy

per

Conjugate
Cam

Cross
pling

cou-

Joint Closure

Cam systems translate a simple input motion into a complex output motion. It is
comprised of a follower and a cam (see figure 3.2) and can be classified by several
criteria, among them type of follower, type of follower motion and type of joint
closure [22].
The first two criteria were quite clearly defined by the problem at hand. The lower
friction of a roller follower makes it perferable to flat-faced or mushroom followers
which both slide along the surface of the cam. And for angular deflection of the
fin-ray, the follower motion would preferably be rotational, not translational. The
remaining question was whether to choose force- or form joint closure to keep the
follower in physical contact with the cam at all times. There were four possibilities
(see figure 3.3), two of which were form closure and two force closure.
(a) Spring system (force closure). The most obvious solution, could be implemented
with extension, compression or leaf springs. The reason for not choosing it were
some experimental problems we experienced. One was the fact that due to the
force pushing onto the contact, the friction losses were quite dramatic because
of the soft nature of our cam material. Another problem was to find mechanical
springs on the very small scale we were testing on at short notice. Trials with
elastic bands were not very promising.
(b) Grooved cam (form closure). This would have been an elegant solution. Lacking

Chapter 3. Basic Concept

12

Figure 3.2: Terminology of cam systems, section

3.2.5. (Taken from [22])

the knowledge of how to achieve the mechanical stability without increasing

the complexity significantly, it had to be discarded. Nevertheless, it remains a
possible alternative for future work, possibly also with conjugate cams (dealt
with in [22], but not studied for this project).
(c) Second camdisc (form closure). This idea was not supported by any examples
in literature at our disposal. Due to the expected mechanical complexity, it was
not tested.
(d) Cross coupling (force closure). Once again a solution not supported by literature, but proposed by our supervisor Wolfgang Fischer. The idea is as simple as
cross-coupling two followers that are 180 phase-shifted relative to each other.
We call it force closure here because the coupling over a wire essentially transmits only a force in one direction, unlike form closure which can exercise forces
in a plane. The advantage of this approach was twofold. For one, we do not increase the friction as in (a) by applying additional normal force on the follower.
At the same time, it can be implemented with comparatively small mechanical
effort. This is further discussed in section 4.6.

3.2.6

Fin Attachment

During the undulation, surface area between rays changes as a function of (t). The
required fin shape was computed, but experiments with the first prototype (section
3.3) showed that the tension of the flexible material lead to torques deforming the
rays when the fin was firmly attached. This torque would have had to be compensated by mechanical adjustments at every ray, without any obvious advantage over
letting the fin just slide through. The torques on the outer rays were then small
enough that no extra bearing was necessary to counteract the material tension. See
section 4.9 for details.

3.2.7

Camshaft Mounting

Mounting adapters on the main shaft instead of the camdiscs directly would have
had the advantage of being able to vary the phase difference between two rays.
The number of steps of the adapter nt determines the minimum phase delay that
can be introduced between two camdiscs (360 /nt ). For the example in figure 3.4,
multiples of = 360 /6 = 60 are possible.

13

3.2. Solution Alternatives

Figure 3.3: Joint closure options (a) to (d) of section 3.2.5

Figure 3.4: Mounting camdiscs

via adapters would allow for
changing the phase delay of two
successive rays.

Chapter 3. Basic Concept

3.3

14

First Prototype

To confirm the thrust production and steering capabilities of the chosen concept, a
first prototype was built (see figure 3.5). The prototype had only six rays with a
phase delay of = 60 . The amplitude was set to = 12.5 .

Figure 3.5: The first prototype. Rear-view of the mechanism attached to the
polystyrene catamaran on the left. Visible on the right are the six actuated rays
used for propulsion. The steering rays and the rudder were attached as a separate
module (see section 3.3.2)

3.3.1

Measuring Thrust

To measure thrust, the prototype was attached to a girder by wires and adjusted so
that the fin was submerged under water. Above the watertank, a ruler was attached
to determine the horizontal deflection of the wires (see figure 3.6).
Using the aforementioned apparatus, the deflections of the wires were measured
for different configurations. This was done by first taking a picture of the initial
position as a reference. The motor was then set to full speed, once in each direction.
For each direction, another picture was taken after the transient phase. From these
pictures, the deflections of the wires were derived. The forward thrust was then
calculated using the equation 3.1 under the assumption that the wire length is
significantly longer than the occurring horizontal deflections:
Fy = mg

y
;
Lwire

(3.1)

Where Fy is the forward thrust, m is the mass of the prototype, g the gravitational
acceleration, y the horizontal deflection of the wires and Lwire the length of a wire
from the attachment point on the girder to the ruler.
Three different waveforms were analysed: A sine wave, a sawtooth wave approximation and a rectangular wave approximation (see figure 3.7).
Each of these waveforms was tested at two different speeds. This was done by
switching the gear adapter, changing the gear ratio between 1:2 and 2:1. Each
measurement was conducted twice, once with a guide restricting the movement of
the wires and once without. The purpose of the guide was to restrict the movement
of the prototype in a plane parallel to the ruler. However, the guide also introduced
additional, if minor friction.

15

3.3. First Prototype

Figure 3.6: Test apparatus for the first prototype.

Figure 3.7: Different waveforms: A sine wave, a sawtooth wave approximation and
a rectangular wave approximation.

Chapter 3. Basic Concept

16

Figure 3.8: Alternative display of the different waveforms: The deflection of the
n-th ray versus the deflection of the (n+1)-th ray.

The results are compared to the fluid dynamics model (see section 4.2) in figure 3.9.
In most cases seen in figure 3.9, the measured thrust was higher than expected
from the fluid dynamics model. The highest and the lowest thrust performance
were displayed by the rectangular and the sawtooth wave, respectively. What is not
visible from the plots is that the high peak torques that were required to perform
the steep slopes of these two wave forms (see figure 3.7) lead to unsmooth operation
conditions. Six times per rotation (once for each ray), the motor nearly stalled. In
this respect, the sine wave performed significantly better.

3.3.2

Steering

Both types of steering, the rudder steering as well as the integrated fin steering
were tested on the prototype. The modular approach allowed a quick switch of the
steering system. The fin steering was realised with three additional rays, two of
which were held in zero deflection by leaf springs. The last ray could be deflected
at a given constant angle and locked in place by screws. A similar approach was
used for the rudder system.
Since only one fin was built, there was a need for a proper apparatus to test the
steering. This was done by attaching the prototype to a catamaran made out of
polystyrene (see figure 3.5). This way, the fin was facing downward, thus being in
the water.
Both steering concepts worked as expected and performed equally well. In both
cases, the turning radius was about the same when moving backwards as when
moving forward. Note that the integrated fin steering only works for very flexible
fin materials, since the total surface area of the fin changes more drastically because
the last ray is held in place.

17

3.3. First Prototype

Figure 3.9: Thrust measurements of the different waveforms compared to the fluid
dynamics model. The crosses represent tests using a wire guide, the dots tests
without guide. Positive thrust means moving forward, negative thrust moving
backwards. The horizontal line shows the expected thrust of a sine wave, calculated by the fluid dynamics model. This is the fast configuration with gear ratio
1:2, resulting in a frequency of about f = 2Hz.

Chapter 3. Basic Concept

3.4

18

Manoeuvrability

A review on Control of Swimming [2] summarises the results of several studies

indicating that there is a tradeoff between stability and manoeuvrability. While
closed-loop stability is a precondition of any system, fish in nature are often openloop instable, seen by the fact that they often end up floating belly up when dead.
This open-loop instability may be the reason why fish are often seen moving their
fins even when at rest. But it also gives them a higher degree of manoeuvrability.
Choosing whether the centre of buoyancy is above or below the centre of mass is
therefore one criterion that will influence the performance of an assembled AUV.
Figure 3.10 shows the directions of possible motion in the fishs body frame. As

Figure 3.10: Motion nomenclature in the

fishs body frame (Adapted from [2])
explained in section 4.2 two undulating pectoral fins would generate the main thrust
so as to contribute to surge, and only little slip. This means that yaw can be achieved
by simply controlling the left and right pectoral fins independently.
As will be discussed in section 4.5, a steering mechanism was integrated to allow
forces in heave direction for both pectoral fins. This allows pitch control if both are
controlled synchronously. Roll manoeuvres are possible when the steering of the
fins is controlled asynchronously. There are no plans to implement slip, but heave
could be possible by pressure tanks on board the AUV. Advanced studies on the
influence of fin placement and control can be found in [13].

Chapter 4

Detailed Design
This chapter will present the fin that was built during the project, CuttleFin.
Thanks to an inhouse rapid-prototyping 3D-printer, the ideas for design could directly be implemented. The individual elements are highlighted and then the operation of the whole fin discussed. Evaluation of experimental results will be dealt
with in chapter 5.

4.1

Parameter Decisions

The following table summarises the parameter decisions discussed in sections 2.1
and 3.1. They are based on literature study and our experiences with the first
prototype (section 3.3).

Table 4.1: Decisions on the main parameters

Parameter
Frequency

Symbol
f

Value
2Hz

Amplitude

15

Number of rays

12

Waveform

(t)

sine

Phase delay

30

Ray distance

dray

20mm

Wavelength
Specific wavelength
Fin length

w
L

240mm
1.09
220mm

Fin width

95mm

19

Comment
Range around 2Hz according to [5]
Range around 15 discussed in 4.3
See specific wavelength,
section 4.7
Most often found in literature, see section 4.3
See specific wavelength,
section 4.7
Chosen due to overall dimension and waveform
See section 4.7
See section 4.7
Was a specification of the
project
Part of the fin design, see
section 4.9

Chapter 4. Detailed Design

4.2

20

Fluid Dynamics Model

To calculate the propulsive force of a transversally undulating fin, a simplified fluid

dynamics model was implemented. The model is based on a basic equation calculating the drag force of a single propulsive element moving through a fluid [4]:
1
2
(4.1)
F = CS ||v|| uv
2
where F is the force exerted by the fluid on the propulsive element, is the density
of the fluid, C is the drag coefficient, S is the effective area of the propulsive element,
v is the velocity of the propulsive element and uv is a unit vector in the direction
of the velocity.
For a smooth, thin and flat body, the parallel force is negligible. Therefore, we
can assume that the force acting normal to the surface of the propulsive element is
nearly equal to the entire force on the element [4]:
1
(4.2)
F
= F = CS[v u ]2 u
2
where F is the component of the drag force acting perpendicular to the surface of
the propulsive element, v is the velocity relative to the fluid and u is a unit vector
normal to the surface of the element within 90 of the velocity vector.
To simplify the calculation of the propulsive force, we made the following assumptions:
The mass of the rays and the fin are negligible
The thickness of the rays and the fin are negligible
All rays are rigid
The fin is approximated by two triangles between each two rays

Figure 4.1: Fluid dynamics model using triangles as an

approximation of the fin. The triangles in dark blue only
contribute to the inward oriented force (x), whereas the
triangles in light blue also provide propulsive force (y).
The normal vectors to the relevant triangles are shown in
red.

21

4.2. Fluid Dynamics Model

Each of these triangles is regarded as a flat plate moving through a fluid. Note
that only half of the triangles contribute to the forward thrust (light blue in figure
(4.1), referred to as relevant triangles). The other half of the triangles only provide
inward thrust (dark blue in figure (4.1)). Instead of implementing a mesh on the
triangles, we calculated the instantaneous velocity of each triangle providing thrust
using the simplification
h
ib
f in
(4.3)
vn (t) = n (t) + n+1 (t)
2
where vn (t) is the magnitude of the instantaneous velocity of the n-th relevant
triangle, n (t) the rotational speed of the n-th ray, n+1 (t) the rotational speed of
the (n+1)-th ray and bf in the width of the fin.
The force of each nonrelevant triangle is determined similarly:
bf in

vn
(t) = n (t)
2

(4.4)

where vn
(t) is the magnitude of the instantaneous velocity of the n-th nonrelevant
triangle.
The time-averaged total force of one full rotation of the camshaft is given by the
following integral:

Ftotal

1
=
T

ZT
F (t)dt

(4.5)

where Ftotal is time-averaged total force of one complete fin undulation, T is the
period, F (t) is the instantaneous resultant force acting on the fin.
We calculated the time-averaged force vector using equations (4.2-4.5) with the
aforementioned simplifications. The results show that the main force is in the
desired direction along the y-axis. There is also a smaller inward oriented force
along the x-axis. It is about 3% of the propulsive force at our operating point
( = 20 , f = 2Hz, w = 1.09). This is in the same range as the value calculated
in [4], which was 4%. As expected, there is no resulting force in the direction of the
z-axis after one complete fin undulation.
To gain a basic knowledge of how the main parameters affect the propulsive force,
we varied amplitude, frequency and specific wavelength.
As seen in figures (4.2-4.4), according to the fluid dynamics model, there is no optimal amplitude or frequency. The higher the amplitude and frequency respectively,
the higher the propulsive force. There are of course effects not considered in this
model, as well as physical constraints, that prohibit the implementation of a fin
wave with infinite amplitude and frequency.
Interestingly, when varying the frequency and keeping the amplitude constant, there
seems to be an optimal specific wavelength where the propulsive force shows a
peak. For an amplitude of = 20 the optimal specific wavelength is equal to
w =1.125. For an amplitude of = 30 it is w = 1.4875. This finding corresponds
approximately to experiments performed in [5].
However, when varying the frequency, the optimal specific wavelength maximising
thrust increases linearly with the frequency, as seen in figure (4.5). Note that
this relation has not been confirmed by literature and may well be a result of the
simplifications of the implemented model.
Using the described model, the expected forward thrust was Fy = 111mN at our
main operating point ( = 20 , f = 2Hz, w = 1.09) as seen in table 4.2.
According to the optimisations of [5], the optimal specific wavelength for our model
would have been 1.125. With the chosen joint closure (3.2.5), we were not able to

Chapter 4. Detailed Design

22

Figure 4.2: Fluid dynamics model: Varying amplitude and frequency at a constant
specific wavelength w = 1.09.

Figure 4.3: Fluid dynamics model: Varying amplitude and specific wavelength at a
constant frequency f = 2Hz.

23

4.2. Fluid Dynamics Model

Figure 4.4: Fluid dynamics model: Varying frequency and specific wavelength at a
constant amplitude = 20 .

Figure 4.5: Optimal, thrust maximising specific wavelength as a function of the

amplitude. It is independent of the frequency.

Chapter 4. Detailed Design

24

Table 4.2: Thrust calculation for selected parameters.

Fin
length
L
220mm
220mm
220mm

Ray
length
Lray
115mm
115mm
115mm

Number
of rays
N
12
12
12

Amplitude

Frequency

10
20
30

f
1Hz
2Hz
3Hz

Specific
wavelength
w
1.09
1.09
1.09

Thrust
Fy
5.4mN
111mN
2200mN

choose arbitrary values for w as described in section 4.7. But w = 1.09 turned out
to be reasonably close.

4.3

Camsystem

In literature, most often the undulating waves considered are sinewaves. Usually
servo motors are used to control the deflection of a ray. Generating the same kind
of propulsion wave with as few actuators as possible leads us to the old principle of
camdiscs. In textbooks on cam design, the described followers are most often used
for translational displacement functions as found in the control of motor valves. One
useful online reference was [29], which also considered our case of a rotational follower. The described method for calculating the camdisc is by inversion. Inversion
is explained in [29] as follows:
The method termed inversion is commonly used in cam profile design.
For example, in a disk cam with translating follower mechanism, the
follower translates when the cam turns. This means that the relative
motion between them is a combination of a relative turning motion and
a relative translating motion. Without changing this feature of their
relative motion, imagine that the cam remains fixed. Now the follower
performs both the relative turning and translating motions. We have
inverted the mechanism.
Furthermore, imagine that the knife-edge of the follower moves along the
fixed cam profile in the inverted mechanism. In other words, the knife
edge of the follower draws the profile of the cam. Thus, the problem of
designing the cam profile becomes a problem of calculating the trace of
the knife edge of the follower whose motion is the combination of the
relative turning and the relative translating.
Since our camdisc had to be imported into the CAD environment where we relied
on spline interpolation of 16 points, a closed-form analytical solution of our camdisc
function was not necessary. Using MATLAB, the inversion problem for our specific
parameters of the mechanism was plotted as seen in figure 4.6.
After an attempt to implement camdiscs for a 30 amplitude, it was clear that
it was not feasible with the specifications used for the CuttleFin. Due to the
errors introduced by the spline approximation of the calculated shape in the CAD
environment and the tolerance limit of the rapid prototyper, the surface of the
camdisc turned out to be slightly uneven. This resulted in unwanted sliding of the
rays, thus highly increasing the friction. Due to these circumstances, it was decided
to restrict the amplitude to a maximum of 20 . In this configuration, the shape of
the camdisc was more smooth and lacked dents, thus having lower demands on
manufacturing tolerance.

25

4.4. Frame

Figure 4.6: Graphic representation of the camdisc, acquired

through implementing the inversion principle to a simulation in
MATLAB. The inner white area represents the shape of the
camdisc.

4.4

Frame

The frame was primarily designed to house the cam-mechanism for propulsion and
steering. In order to minimise the number of affected parts in case of breakage, the
frame was designed in a modular fashion. Figure 4.8 shows one such module that
can be mounted on the frame in figure 4.7.

4.5

Steering

Incorporating a steering unit directly into the undulating wave allows a more compact design. In this configuration, the last ray is directly actuated by a servo motor,
which holds the ray at a constant angle (see figure 4.9). The resulting force component that could be generated by the this setup was verified with our simplified
fluid-dynamics model.
The constant deflection of the last ray was set to be 45 at a distance of 5cm
from the second last ray. The resulting force of 0.8mN was calculated, which is
about 0.7% of the forward thrust. This was enough for manoeuvring the fin as
was confirmed with our first prototype. On CuttleFin, this steering mechanism
could not be implemented, because of the split design of the rays. This design
allows the fin to slip through the rays, thus a less flexible material could be used.
In the connection to the last, directly actuated ray however, high flexibility is key.
Otherwise, the steering ray might become largely deformed, rendering the steering
mechanism useless.
To still be able to manoeuvre the catamaran, described in section 4.10, a simple
steering rudder was implemented instead. It is actuated by a servo motor with a
maximum deflection of 45 (figure 4.10).

Chapter 4. Detailed Design

Figure 4.7: The bare frame plate. To reduce the impact of breakage, the frame was designed to host all components in the form of
small modules.

Figure 4.8: Assembly of a ray unit consisting of seven printed parts

(of which only five are visible) and two brass four cornered shafts.

26

27

4.5. Steering

Figure 4.9: Steering mechanism integrated in the fin. The right most fin segment
does not contribute to the propulsion. The last ray is kept at a certain angle by a
servo to influence the z-component of the resulting thrust.

Figure 4.10: A rudder steering was chosen for the second prototype to allow for
force generation in z-direction.

Chapter 4. Detailed Design

4.6

28

Cam Coupling

As discussed in section 3.2.5, joint closure could easily have been achieved by springs.
However, these would have introduced two sources for energy losses:
Higher friction losses due to the increased normal force on the contact raycamdisc
Internal energy losses of the spring
The solution via coupling two rays with a 180 phase-shift in theory has no losses.
Furthermore, it is easy to implement mechanically, as can be seen in figure 4.11.
The one decisive disadvantage of this solution is the fact that it limits the wave
function to those with even numbers of rays that come in pairs of 180 phase shift,
as is discussed in the following section 4.7.

Figure 4.11: Pulley and wire system. The rear ends of each linked pair of rays
which are 180 out of phase are marked by circles of the same colour. One wire is
highlighted as an example.

4.7

Specific Wavelength

The decision to use a wire coupling system limited the number of realisable phase
delays and therefore specific wavelengths. All linked pairs of rays have to be 180
out of phase. This means that the phase delay can only be a whole fraction of 180 .
Furthermore, the phase delay has to yield a value that guarantees that each ray has
a complement. Therefore, the minimum value of the phase delay is 360 /N , where
N is the number of rays. In addition, the number of rays obviously has to be even.
The specific wavelength directly depends on the phase delay, as shown in equation
4.6:
w=

1
360

N 1

(4.6)

where w is the specific wavelength, is the phase delay and N is the number of
rays. Using equation 4.6, the restrictions of the phase delay can be transferred
to the specific wavelength. Table 4.3 shows the feasible phase delays and specific
wavelengths for 12 rays.
The decision was to use 12 rays to have the flexibility to change the phase delay
between two adjacent rays from 30 to 180 . At a phase delay of 30 , quite a
smooth sine wave shape can be achieved. At a phase delay of 60 , resulting in
almost two full waves on the fin, the sine wave shape is still recognisable. When
further increasing the phase delay, the wavefrom degenerates more and more to a
triangular shape.

29

4.8. Power Unit

Table 4.3: Feasible phase delays and specific wavelengths for 12 rays.
Phase delay
180
90
60
30

4.8

specific wavelength w
0.18
0.36
0.55
1.09

Power Unit

In this section, the specifications of the motor and the gears are discussed and
justified.

4.8.1

Motor

Determining the optimal motor specifications was difficult due to the lack of an
exact requirement profile. Especially the amount of friction in our system was hard
to estimate. Under the circumstances, we chose a motor with a high power output
considering the size of our prototype. The motor should be able to overcome a
lot of friction, at the risk of being overdimensioned and thus losing efficiency. The
main goal of this project being a feasibility study, optimisations of this kind could
be subject of further work.
The chosen motor was a 6V motor from Maxon with 3W power. Its diameter is
13mm, which was small enough to easily implement it in our prototype.
For the steering, we used a 6V servo motor designed for remote controlled modelling.

4.8.2

Gears

For the torque transmission from the motor to the camshaft, it was decided to
use module 1 spur gears. Both the shaft and the motor, were equipped with a 12
tooth gear. To quickly change the gear ratio, adapters which could be mounted
on the 12 tooth gear were printed, increasing the amount of teeth to 18, 24 or 36
teeth and the diameter accordingly (see figure 4.12). To make up for the additional
diameter of the gears with the adapters, the placement of the motor can be adjusted
perpendicular to the camshaft. This mechanism allows a quick variation of the
torque and rotational speed of the camshaft by simply exchanging the gear adapters
and, if necessary, moving the motor frame.

Figure 4.12: Spur gear with three adapters.

Chapter 4. Detailed Design

4.9

30

Fin

Fish fins (figure 2.4) and undulation forms (figure 2.3) introduced in section 2.1
allow us to delineate the current solution used for evaluation purposes from the
proposed final implementation in an AUV. The gymnotiform application served
only as a testbench for measurements. The design idea from figure 1.1 calls for a
rajiform configuration involving two of our proposed fins. Together with a pressure
tank regulating the depth under water, we expect the system to be manoeuvrable
in at least yaw, pitch, surge and heave directions as described in section 3.4.

Figure 4.13: CuttleFin.

If the rays in figure 4.14 provide the skeleton of the fin, it is the connecting membrane
that makes the combination a fin capable of displacing water. Like other undulatory
robots [28] [5], CuttleFin employs a flexible elastomer for the reasons of flexibility
and density. Natural rubber with a thickness of 0.5mm was chosen. Unlike in the
quoted papers, the demand for elasticity was not very high since our rays allowed
the material to slide through (see figure 4.14). The shape that the fin describes in
space had to be projected onto a plane in order for the material to be cut. The
required projection was calculated numerically by a summation of the distances of
two rays. This is a function of time and of radial distance from the camshaft centre.
Figure 4.15 shows the two shapes acquired by two different approximation methods
used. In both cases, the goal was to best approximate the minimum occurring
surface area of a full undulation. This way, there will always be a slight tension.
The density of the material is 1.01g/cm3 , which is very desirable form the buoyancy
point of view.
Based on observation of fish in nature, the ray is designed in such a way as to be
flexible in the direction vertical to the fin and less so in axial direction. This was
achieved by using thin brass profiles of 1mm 2mm.

31

4.9. Fin

Figure 4.14: Closeup of a single fin ray

Figure 4.15: Fin shape approximations. The light blue

shape consists of the projected triangles from the fluid dynamics model. The dark blue line represents the approximation used to manufacture the fin. It is determined by
calculating the inner and outer arc length of the fin. The
lines were drawn as parabolas with approximately the calculated arc length using parameters that fit best.

Chapter 4. Detailed Design

4.10

32

Catamaran

To conduct tests and measurements with CuttleFin, a polystyrene catamaran was

built, like with the first prototype (section 3.3.2).

4.10.1

Body

The catamaran was cut out from a polystyrene block using a heated wire. The
shape of the catamaran is rather straight-forward. Its front and rear end are just
curved upwards to reduce drag (see figure 4.16). However, the shape is not optimal,
as a stream-lined body could be implemented, which would even further reduce
drag.

Figure 4.16: Closeup of the front end of the catamaran.

The prototype was attached in the middle of the catamaran with the fin facing
downward. This allowed testing the mechanism with only one fin instead of two
and without waterproofing the sensitive components such as motor and servo (see
figure 4.17).

4.10.2

Electronics

To ease the measurement procedure, a remote control was integrated. The components transmitter, receiver, speed controller, nickel metal hydride battery pack
were those available at a RC-modelling store. Multimeters monitoring the current
and voltage of the motor are also visible in figure 4.18.

33

4.10. Catamaran

Figure 4.17: CuttleFin attached to the catamaran.

Figure 4.18: Catamaran in its test environment. Visible in

the photograph are electronics for the remote control and
power measurement.

Chapter 4. Detailed Design

34

Chapter 5

Evaluation
The solution described in chapter 4 is evaluated here, based on measurements in
a water channel and the simplified fluid dynamic model described in section 4.2.
The employed apparatus and the measurement approach is discussed before the
presentation and discussion of the numerical results.

5.1

Apparatus

The water-towing tank (figure 5.1) at the Laboratory of Energy Conversion [7]
served as the testing environment. Its dimensions are 40x1x1m.

Figure 5.1: Water-towing tank at the Laboratory

of Energy Conversion of ETH Zurich
The shape of the catamaran is one aspect that could be improved in a future
35

Chapter 5. Evaluation

36

attempt. Due to last-minute overloading of the catamaran, the waterline ended

up even higher than anticipated. The fluid dynamics of the test vehicle probably
influenced the performance adversely by increasing the drag force on the catamaran,
reducing its speed.
The forward thrust of the fin was measured using a spring balance suspended vertically, while a thin wire redirected the horizontal force around a bearing.
The voltage and the current were measured using standard multimeters mounted on
the catamaran. A calibration issue hindered us from using the prepared electronic
board for the same purpose, which lead to the aforementioned overloading.

5.2

Measurement Procedure

This section shows, which of the parameters from table 4.1 were varied to what
extent, and how the values were recorded.

5.2.1

Parameter Variations

Since a change in the amplitude requires the complete removal of the camshaft,
only three different amplitudes were examined, namely = 10 , = 15 and
= 20 . The maximum value was chosen due to mechanical constraints discussed
in section 4.3. Decreasing the amplitude further than 10 would result in a rapidly
diminishing thrust, which would be hard to measure correctly.
For each value of the amplitude, the frequency was altered. This was done in two
different ways. The gear ratio between the motor and the camshaft was changed
using gear adapters (see section 4.8.2). Three gear ratios were used: 2:1 (higher
torque at the cost of speed), 1:1 (same speed and torque as the motor) and 1:2
(higher speed at the cost of torque). With the throttle on the remote control, the
frequency could be further adjusted. This was done at two settings: 100% and 75%
or 50% throttle for each gear ratio. Note that the throttle values below full throttle
are approximations, as they were set and kept manually on the remote control
stick. This however did not influence the measurements, since the frequency was
determined separately. The described configurations resulted in frequencies in the
range of about f = 1Hz to almost f = 3Hz.
The specific wavelength was kept at w = 1.09 at all times. An overview of the
parameter variations can be found in table 5.1.
Table 5.1: Parameter variations
Parameter
Frequency
Amplitude
Specific wavelength

5.2.2

Symbol
f

Range
1Hz - 3Hz
10 , 15 , 20
1.09

Measured Values

For each configuration, the following procedure was conducted at least twice: First,
the catamaran was driven for 5 meters, taking the time and counting the number of
rotations. This allowed the calculation of the frequency f of the camshaft, as well
as the speed v of the catamaran. To measure the actual electrical power consumption P of the motor, the current I flowing through the motor and the voltage V
applied to the motor were determined using common multimeters put directly onto

37

5.2. Measurement Procedure

the catamaran. Second, the forward thrust Fy was measured by attaching the catamaran to the aforementioned spring balance apparatus. The thrust was then read
from the spring balance displacement visually. Figure 5.2 shows the measurement
sequence for = 20 .

Figure 5.2: Measured and evaluated values for one of three experiments. Here at
= 20 .

5.2.3

Measurement Errors

Table 5.2 summarises the expected error margin for each of the measured variables.
Table 5.2: Error estimation for the measured values
Variable
Distance
Time
Counts
Voltage
Current
Thrust

Symbol
d
t
n
V
I
Fy

Typical Value
5m
30s
50
6V
0.5A
300mN

Error Margin
0.02m
0.1s
1
0.1V
0.1A
10mN

Relative Error
0.4%
0.33%
2%
1.6%
0.05%
3.3%

The derived performance figures inherit the error margins as described in table 5.3.
Table 5.3: Error estimation for the derived values
Variable
Speed
Frequency
Input Power
Output Power
Efficiency

Formula
u = d/t
(n 1)/t
Pin = V I
Pout = u Fy
total = Pout /Pin

Relative Error
0.73%
2.33%
1.65%
4.03%
5.68%

Chapter 5. Evaluation

5.3
5.3.1

38

Results
Remarks

During the measurement process, some problems arose. First, the approximation
of the fin shape turned out to be slightly too big for amplitudes of 15 and 10
degrees. However, it was only a minor mismatch. To the authors knowledge, the
resulting temporary tension loss of the fin didnt have a relevant negative impact
on the measurements.
The second issue occurred at all amplitudes and frequencies when using a gear ratio
of 1:2 at full throttle, where the camshaft rotates faster at the cost of torque. In this
configuration, the gears between the motor and the camshaft slipped occasionally.
In one case, the motor itself began to spin in its seating. Due to these circumstances,
the values acquired in cases of gear slippage were marked accordingly.

5.3.2

Speed

The acquired speeds are displayed in figure 5.3. The differences in performance
with = 10 ,15 and 20 can clearly be seen. The maximum speed achieved was
26cm/s at = 20 and f = 2.3Hz. This corresponds to approximately 1 fin length
per second, which is a comparable result to that of rainbow trout with 1.2 body
lengths per second at slow cruising speed [21]. It can be further noted that higher
speeds would have been possible if no gear slippage had occurred (seen as crosses
in figure 5.3).

Figure 5.3: Speed of CuttleFin. The asterisks represent the calculated values for
the amplitudes = 10 ,15 and 20 . Calculations using data where gear slipping
occurred are displayed as crosses.

39

5.3.3

5.3. Results

Comparison to Fluid Dynamics Model

The obtained values of the forward thrust were compared to the fluid dynamics
model described in section 4.2. The measured thrust exceeded the expected values
calculated by the model. For an amplitude of 10 , the thrust was about 7 times
higher than expected, for 15 about 5 times higher and for 20 about 3 times higher,
see figures 5.4-5.6. This discrepancy can be explained by the drastic simplifications
of the model used. Also, the measurement accuracy is rather limited, due to the
fact that all values were obtained manually.
The fluid dynamics model fits better to the measured data when varying the prefactor CS of the drag equation. For the sake of simplicity, only the drag coefficient
C was varied. However, the surface area S, as well as the chosen mesh for the
calculation and even uncertainties in the density of water may contribute to the
deviation of the model. The model fits best for a drag coefficient of C = 5. It is not
clear if this value is coherent, because it is chosen arbitrarily and it seems rather
high for a drag coefficient.

Figure 5.4: Comparison of the measured forward thrust to the fluid dynamics model.
The red asterisks represent the measured values. Measurements, where gear slipping
occurred are displayed as purple crosses. The blue lines show the calculated points
from the fluid dynamics model for different drag coefficients C of the fin.
The results confirm the qualitative nature of the model quite well. The higher the
frequency and the higher the amplitude, the higher was the forward thrust. This
corresponds directly to the model.
To some extent, the quantitative aspect of the model were also confirmed. The
acquired data lies within an acceptable range of the model considering the simplifications, although being higher than expected in all measured values.

Chapter 5. Evaluation

40

Figure 5.5: Comparison of the measured forward thrust to the fluid dynamics model.
The red asterisks represent the measured values. The blue lines show the calculated
points from the fluid dynamics model for different drag coefficients C of the fin.

5.3.4

Efficiency

The Froude efficiency is defined as [2]:

F =

PE
PP

(5.1)

where F is the Froude efficiency, PE is the useful propulsive power and PP is

the time-averaged power expended by the prototype. F reflects the ability of the
swimming body to impart useful kinetic energy to the water. This does not include
friction losses in the mechanism itself, which have to be considered outside the
Froude efficiency.
For steady state aquatic locomotion, high Reynolds numbers and a constant amplitude, the Froude efficiency only depends on the ratio between the speed of the
backward wave travelling along the fin c and the forward speed u of the whole system. The Reynolds number of CuttleFin is in the vicinity of Re = 105 , which
is high enough to use equation 5.2. For slender fish, the Froude efficiency can be
calculated as follows [15]:
1
u
F =
1+
(5.2)
2
c
where u is the speed of the whole system (catamaran) and c is the speed of the
backward wave on the fin.
Figure 5.7 shows the Froude efficiencies of the conducted measurements.
A Froude efficiency of about 65% to 80% compares to other results, for example the
rainbow trout (F = 74%) [21]. According to [2], the range for the Froude number
of fish is from 90% for a carangiform swimmer to 16% for drag-based labriform
locomotion.

41

5.3. Results

Figure 5.6: Comparison of the measured forward thrust to the fluid dynamics model.
The red asterisks represent the measured values. Measurements, where gear slipping
occurred are displayed as purple crosses. The blue lines show the calculated points
from the fluid dynamics model for different drag coefficients C of the fin.

Chapter 5. Evaluation

42

Figure 5.7: Froude efficiency of CuttleFin. The asterisks represent the calculated
values for the amplitudes = 10 ,15 and 20 . Calculations using data where gear
slipping occurred are displayed as crosses.

The overall efficiency is evaluated using the equation 5.3.

total =

u Fy
I V

(5.3)

where total is the overall efficiency coefficient, u is the velocity of the catamaran,
Fy is the forward thrust, I is the electrical current flowing through the motor and
V the voltage applied to the motor.
The maximum overall efficiency is about 2.6% at 20 amplitude at several frequencies. This value seems extremely low for a propulsion system. While energy
efficiency was not the primary goal of this thesis, the sources for this rather poor
result would have to be analysed in a follow-up project.
There are three main causes for losses:
Friction:
The friction between camdisc and follower could be reduced if the camdisc was
made from another material. Furthermore, the tension of the wire-coupling
has an effect on said friction.
Motor and gear:
The motor-gear combination should be chosen to be operating at the highest
efficiency of its operating region. In our preliminary experiments, this was
not the case.
Catamaran:
The catamaran should be more streamlined.

43

5.3. Results

Figure 5.8: Overall efficiency of CuttleFin. The asterisks represent the calculated
values for the amplitudes = 10 ,15 and 20 . Calculations using data where gear
slipping occurred are displayed as crosses.
Considering only the second point of the above list, we can account for part of
the losses from the motor input power to output power that is transmitted via the
camshaft (see table 5.4).
Table 5.4: Error estimation for the derived values
Description
Maxon Motor
Maxon Planetary Gear
Spur Gears

Symbol
motor
planetary
spur

Max Efficiency
67%
60%
90%

Chapter 5. Evaluation

44

Chapter 6

Conclusion
Based on the performance of CuttleFin, we conclude with an assessment of our
work and propose directions for future work.

6.1

Assessment

The goal was to design and evaluate a locomotion concept for AUVs (section 1.3).
The design and construction of the system required so much time that the evaluation was not as rigorous as would have been desirable. The results were highly
satisfactory at first glance, but a more indepth study of the performance would be
recommendable if it is to be implemented in an AUV. Especially the question of
manoeuvrability, which was not addressed in controlled experiments.
Nevertheless, CuttleFin achieves comparable results in terms of speed and thrust
to other undulatory systems with a minimum of necessary actuators. The mechanical realisation by means of a camshaft has on one hand the advantage that the
propulsion system can be designed simple and robust. On the other hand, compared
to other proposed robots, the solution is far less flexible in terms of the variety of
generated waves.
It is the authors belief that this type of locomotion is applicable to AUVs, at least
at the scale of CuttleFin. Scalability in both directions, smaller and larger, would
have to be analysed separately. Whenever robustness of operation in marine flora
and non-invasive coexistence with marine fauna is more important than speed or
efficiency, propulsion through fin undulation is a promising alternative to the widely
used propellers.

6.2

Future Work

CuttleFin offers a wide and exciting spectrum of possible topics for follow-up
projects with the end goal of designing an AUV using camshaft-generated fin undulation as means of propulsion. Some suggestions are:
Rebuild CuttleFin with better materials such as laser-cut camdiscs and
using exclusively non-corrosive metals to perform a more thorough and demanding evaluation of the performance
Analyse the efficiency and compare it quantitatively to (a) other undulationrobots and (b) to other methods of propulsion such as propellers
Investigate scalability of the fin
45

Chapter 6. Conclusion

46

Design an AUV hull while incorporating and water-sealing CuttleFin

Design, implement and test control strategies involving two fins and thus
analyse manoeuvrability.

6.3

Acknowledgements

The authors would like to thank their professor Prof. Dr. Roland Y. Siegwart and
their supervisor Dr. Cedric Pradalier for initiating this exciting project and for
their support and guidance.
We owe a lot of credit for the mechanical design to the experience and skill of our
second supervisor, Wolfgang Fischer.
Technical support with the electronics was provided by Thomas Baumgartner and
Dr. Samir Bouabdallah, and we are very grateful for their time and loaned equipment.
Rapid prototyping and other assembly work was conducted in the mechanical workshop with the assistance of Markus B
uhler and Dario Fenner.
Dr. Sarah Barber of the Laboratory for Energy Conversion was kind enough to let
us use the water-towing tank. Further motivation and support was provided by the
naro team, especially Cedric Siegenthaler, Jakob Buchheim, Thomas Wuhrmann,
Fabian G
unther and Milian Seiler.
We are thankful for Petros Papadopoulos and Elisa Charalampidou, who helped us
with the measurements in the water-towing tank.
We thank Visvanath Ratnaweera and Dominik Peter for cross-reading the first draft.

Bibliography
[1] Breder, C., and Edgerton, H. An analysis of the locomotion of the seahorse, hippocampus, by means of high speed cinematography*. Annals of
the New York Academy of Sciences 43, 4 An Analysis of the Locomotion of
the Seahorse, Hippocampus, by Means of High Speed Cinematography (1942),
145172.
[2] Colgate, J., and Lynch, K. Mechanics and control of swimming: a review.
IEEE Journal of Oceanic Engineering 29, 3 (2004), 660673.
[3] Dudek, G., Giguere, P., Prahacs, C., Saunderson, S., Sattar, J.,
Torres-Mendez, L.-A., Jenkin, M., German, A., Hogue, A., Ripsman, A., Zacher, J., Milios, E., Liu, H., , Zhang, P., Buehler, M.,
and Georgiades, C. AQUA: An amphibian autonomous underwater robot.
Copmuter (2007), 4653. Published by IEEE Computer Society.
[4] Epstein, M., Colgate, J., and MacIver, M. A biologically inspired
robotic ribbon fin. In Proceedings of IEEE/RSJ International Conference on
Intelligent Robots and Systems (2005), Workshop on Morphology, Control, and
Passive Dynamics.
[5] Epstein, M., Colgate, J., and MacIver, M. Generating Thrust with a
Biologically-inspired Robotic Ribbon Fin. Northwestern University, 2006.
[6] ETHZ. Nautical robot. http://www.naro.ethz.ch/. [Online; accessed May
2009].
[7] ETHZ.
Water-towing tank.
http://www.lec.ethz.ch/capabilities/
facilities/water_towing_tank. [Online; accessed May 2009].
[8] Fischer, M. Aqua ray, water-hydraulic manta ray with flapping-wing drive.
Project developed by Festo AG, 2007.
[9] Fischer, M. Aquapenguin, a biomechatronic overall concept. Project developed by Festo AG, 2009.
[10] Fukuda, T., Hosokai, H., and Kikuchi, I. Distributed type of actuators
by shape memory alloy and its application to underwater mobile robotic mechanism. In 1990 IEEE International Conference on Robotics and Automation,
1990. Proceedings. (1990), pp. 13161321.
[11] Kemp, M., Hobson, B., and John H. Long, J. Madeleine: an AUV
propelled by flexible fins.
[12] Kim, B., Kim, D., Jung, J., and Park, J. A biomimetic undulatory tadpole
robot using ionic polymer-metal composite actuators. Smart materials and
structures 14, 6 (2005), 1579.
47

Bibliography

48

[13] Lauder, G., and Drucker, E. Morphology and experimental hydrodynamics of fish fin control surfaces. IEEE Journal of Oceanic Engineering 29, 3
(2004), 556571.
[14] Lighthill, J., and Blake, R. Biofluiddynamics of balistiform and gymnotiform locomotion. Part 1. Biological background, and analysis by elongatedbody theory. Journal of Fluid Mechanics Digital Archive 212 (2006), 183207.
[15] Lighthill, M. J. Note on the swimming of slender fish. Journal of Fluid
Mechanics Digital Archive 9, 02 (1960), 305317.
[16] Liu, J., Dukes, I., and Hu, H. Novel mechatronics design for a robotic
fish. In Intelligent Robots and Systems, 2005.(IROS 2005). 2005 IEEE/RSJ
International Conference on (2005), pp. 807812.
[17] Low, K. Mechatronics and buoyancy implementation of robotic fish swimming
with modular fin mechanisms. Proceedings of the Institution of Mechanical
Engineers, Part I: Journal of Systems and Control Engineering 221, 3 (2007),
295309.
[18] Low, K. Modelling and parametric study of modular undulating fin rays for
fish robots. Mechanism and Machine Theory 44, 3 (2009), 615632.
[19] Low, K., Seet, G., and Zhou, C. Biomimetic Design and Workspace Study
of Compact and Modular Undulating Fin Body Segments. In Mechatronics and
Automation, 2007. ICMA 2007. International Conference on (2007), pp. 129
134.
[20] Madsen, P. How ships traffic noise affects whales in a shipping channel.
http://www.pbs.org/odyssey/odyssey/20030506_log_transcript.
html, 2003. [Online; accessed April 2009].
[21] Nauen, J., and Lauder, G. Quantification of the wake of rainbow trout
(Oncorhynchus mykiss) using three-dimensional stereoscopic digital particle
image velocimetry. Journal of Experimental Biology 205, 21 (2002), 32713279.
[22] Norton, R. Cam design and manufacturing handbook. Industrial Press Inc.,
2002.
[23] Paquette, J., and Kim, K. Ionomeric electroactive polymer artificial muscle
for naval applications. IEEE Journal of oceanic Engineering 29, 3 (2004), 729
737.
[24] Peter, B. Review of fin-based locomotion for underwater robots.
[25] Sfakiotakis, M., Lane, D., and Davies, J. Review of fish swimming
modes for aquatic locomotion. IEEE Journal of Oceanic Engineering 24, 2
(1999), 237252.
[26] Videler, J. Fish swimming. Chapman & Hall, 1993.
[27] Wang, Z., Hang, G., Li, J., Wang, Y., and Xiao, K. A micro-robot
fish with embedded SMA wire actuated flexible biomimetic fin. Sensors &
Actuators: A. Physical 144, 2 (2008), 354360.
[28] Willy, A., and Low, K. Development and initial experiment of modular
undulating fin for untethered biorobotic AUVs. In 2005 IEEE International
Conference on Robotics and Biomimetics (ROBIO) (2005), pp. 4550.

49

Bibliography

[29] Zhang, Y., Finger, S., and Behrens, S. Introduction to mechanisms.

http://www.cs.cmu.edu/~rapidproto/mechanisms/chpt6.html. [Online;
accessed May 2009].
[30] Ziegler, M. Robotic fish wanda. http://www.ifi.uzh.ch/ailab/robots/
robots-a.html. [Online; accessed May 2009].