26.03.2013 Views

Development of a Flapping Wing Mechanism - Student Projects

Development of a Flapping Wing Mechanism - Student Projects

Development of a Flapping Wing Mechanism - Student Projects

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Oliver Breitenstein<br />

<strong>Development</strong> <strong>of</strong> a <strong>Flapping</strong><br />

<strong>Wing</strong> <strong>Mechanism</strong><br />

Semester Project<br />

Autonomous Systems Lab (ASL)<br />

Swiss Federal Institute <strong>of</strong> Technology (ETH) Zurich<br />

Supervision<br />

Dr. Samir Bouabdallah, Stefan Leutenegger<br />

and<br />

Pr<strong>of</strong>. Dr. Roland Siegwart<br />

Spring Semester 2009


Contents<br />

Abstract iii<br />

Acknowledgements iv<br />

1 Introduction 1<br />

2 Review 3<br />

2.1 Aerodynamics <strong>of</strong> flapping wings . . . . . . . . . . . . . . . . . . . . . 3<br />

2.1.1 Wagner Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . 3<br />

2.1.2 Leading edge vortex . . . . . . . . . . . . . . . . . . . . . . . 3<br />

2.1.3 Clap and fling mechanism . . . . . . . . . . . . . . . . . . . . 4<br />

2.1.4 Rotational lift . . . . . . . . . . . . . . . . . . . . . . . . . . . 5<br />

2.1.5 <strong>Wing</strong>-wake interactions . . . . . . . . . . . . . . . . . . . . . 6<br />

2.1.6 Lift force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6<br />

2.2 <strong>Flapping</strong> wings in nature . . . . . . . . . . . . . . . . . . . . . . . . 7<br />

2.2.1 Insects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7<br />

2.2.2 Hummingbirds . . . . . . . . . . . . . . . . . . . . . . . . . . 10<br />

2.2.3 Bats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13<br />

2.2.4 Birds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16<br />

2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17<br />

3 Concepts 21<br />

3.1 General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 21<br />

3.1.1 Objective characteristics . . . . . . . . . . . . . . . . . . . . . 21<br />

3.1.2 Flight control . . . . . . . . . . . . . . . . . . . . . . . . . . . 22<br />

3.1.3 Actuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22<br />

3.2 Concepts for wing flapping . . . . . . . . . . . . . . . . . . . . . . . 23<br />

3.2.1 Concept A . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23<br />

3.2.2 Concept B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24<br />

3.2.3 Concept C . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25<br />

3.2.4 Concept D . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />

3.3 Concepts for wing pitching . . . . . . . . . . . . . . . . . . . . . . . 29<br />

3.3.1 Active pitching . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />

3.3.2 Passive pitching . . . . . . . . . . . . . . . . . . . . . . . . . 32<br />

4 Evaluation 35<br />

4.1 Evaluation <strong>of</strong> concepts . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />

4.1.1 Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />

4.1.2 <strong>Flapping</strong> concepts . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

4.1.3 Pitching concepts . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

4.2 Expected weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

4.3 Expected power consumption . . . . . . . . . . . . . . . . . . . . . . 38<br />

i


5 CAD Design 39<br />

5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39<br />

5.2 Transmission <strong>of</strong> motor torque . . . . . . . . . . . . . . . . . . . . . . 40<br />

5.3 Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41<br />

6 Conclusion 43<br />

A Motor datasheet 45<br />

ii


Abstract<br />

This project aims at the development <strong>of</strong> a bio-mimetic propulsion mechanism for<br />

a <strong>Flapping</strong> <strong>Wing</strong> Micro Aerial Vehicle, without considering the aerodynamics <strong>of</strong><br />

the wings in the design. This artificial bird will be the size <strong>of</strong> approximately 10-<br />

20cm. Therefor the aerodynamic phenomena in flapping flight are studied and<br />

summarized. It covers the leading-edge vortex (LEV), the clap-and-fling effect,<br />

rotational lift and wing wake interactions. This is followed by a review <strong>of</strong> natural<br />

flappers. The aerodynamic and kinematic pattern <strong>of</strong> hummingbirds, bats, insects<br />

and small birds are summarized. Based on this review several different concepts <strong>of</strong><br />

mechanisms for flapping wings are generated, which are seperated for the flapping<br />

motion and the pitching motion. Using a qualitative evaluation, the quality <strong>of</strong><br />

the concepts are determined according to different criteria such as weight, size,<br />

robustness, mechanical complexity, expected power consumption and accuracy. The<br />

best concept is used as basis for a 3D CAD design <strong>of</strong> the mechanism, which should<br />

mainly reproduce the desired kinematics. During the design process the focus is<br />

set more on getting a robust and simple mechanism, which could be used as a test<br />

bench for further investigations and measurements. Concluding, the mechanism is<br />

manufactured and assembled to prove the feasibility.<br />

iii


Acknowledgements<br />

I’d like to thank Dr. Samir Bouabdallah and Stefan Leutenegger for their good<br />

guidance and the useful inputs they contributed. Specially during the last part, the<br />

CAD-Design, when time was short, their experience was very supportive. Also I’d<br />

like to thank Dr. Bret Tobalske from the University <strong>of</strong> Montana and Maria Jose<br />

<strong>of</strong> Berkeley, giving me deeper informations about the hummingbird flight, which<br />

helped me alot understanding the crucial parts <strong>of</strong> it for developing a mimicking<br />

flapping device.<br />

iv


List <strong>of</strong> Figures<br />

1.1 Schematic drawing <strong>of</strong> DelFly I taken from www.delfly.nl . . . . . . . 1<br />

1.2 <strong>Flapping</strong> wing mechanism <strong>of</strong> ROBUR taken from IROS 2007 www.flyingrobots.org 2<br />

2.1 Leading edge vortex on the wing[19] . . . . . . . . . . . . . . . . . . 4<br />

2.2 Evolution <strong>of</strong> a leading edge vortex in (A) two dimensions and (B)<br />

three dimensions during linear translation starting from rest [19] . . 4<br />

2.3 Schematic representation <strong>of</strong> the clap (A-C) and fling (D-F) [19] . . . 5<br />

2.4 Three phases <strong>of</strong> the wing rotation [7] . . . . . . . . . . . . . . . . . . 5<br />

2.5 <strong>Wing</strong>-wake interaction during stroke reversal [19] . . . . . . . . . . . 6<br />

2.6 Flight forces for the drosophila during hovering [21] . . . . . . . . . . 7<br />

2.7 General pattern for the wing motion <strong>of</strong> Drosophila Melanogaster [16] 8<br />

2.8 Kinematics <strong>of</strong> Drosophila Melanogaster [9] . . . . . . . . . . . . . . . 8<br />

2.9 Force production in two cycles [16] . . . . . . . . . . . . . . . . . . . 9<br />

2.10 <strong>Wing</strong> motion relative to the body flying at velocities <strong>of</strong> 0 − 12ms −1 [2] 10<br />

2.11 Angles describing bird-centered wing and body kinematics in rufous<br />

hummingbirds [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11<br />

2.12 Variation <strong>of</strong> chord angle relative to body-plane during wingbeats at<br />

velocities <strong>of</strong> 0 − 12ms −1 [2] . . . . . . . . . . . . . . . . . . . . . . . 12<br />

2.13 Wake structures in frontal and side plane [17] . . . . . . . . . . . . . 13<br />

2.14 Flow field vorticity at end <strong>of</strong> upstroke, (a) frontal view at shoulderplane,<br />

(b) side view at midwing-plane [17] . . . . . . . . . . . . . . . 13<br />

2.15 Anatomical structure <strong>of</strong> the bat wing [5] . . . . . . . . . . . . . . . . 14<br />

2.16 Sequences <strong>of</strong> images from below and in front <strong>of</strong> bat during on cycle<br />

starting at beginning <strong>of</strong> the downstroke [6] . . . . . . . . . . . . . . . 14<br />

2.17 Example trajectories <strong>of</strong> the different wing regions for 3m/s (left) and<br />

9m/s (right) [5] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14<br />

2.18 Example <strong>of</strong> wing tip motion [6] . . . . . . . . . . . . . . . . . . . . . 15<br />

2.19 Velocity and vorticity fields around a bat wing in slow forward flight<br />

(1 m/s) at the time instance when the wing is in horizontal position<br />

during the downstroke [10] . . . . . . . . . . . . . . . . . . . . . . . . 15<br />

2.20 <strong>Wing</strong>span ratio as a function <strong>of</strong> flight velocity compared among bird<br />

species [2] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16<br />

2.21 Representative wing kinematics in a zebra finch engaged in flapbounding<br />

flight at 2m/s (A) and 12m/s (B) [4] . . . . . . . . . . . . 17<br />

3.1 Schematic drawing <strong>of</strong> concept A1 . . . . . . . . . . . . . . . . . . . . 23<br />

3.2 Sketch for kinematics <strong>of</strong> general structure for the flapping motion . . 24<br />

3.3 Trajectories <strong>of</strong> centered joint for one cycle for different ratios L/r . . 24<br />

3.4 Schematic drawing <strong>of</strong> concept A2 . . . . . . . . . . . . . . . . . . . . 25<br />

3.5 Schematic drawing <strong>of</strong> concept B1 (left) and B2 (right) . . . . . . . . 25<br />

3.6 Schematic drawing <strong>of</strong> concept C . . . . . . . . . . . . . . . . . . . . . 26<br />

3.7 Calculation <strong>of</strong> the bending line . . . . . . . . . . . . . . . . . . . . . 26<br />

3.8 Sketch for calculation <strong>of</strong> the dynamics . . . . . . . . . . . . . . . . . 27<br />

v


3.9 Results <strong>of</strong> the force on the link and the needed torque during one<br />

flapping cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28<br />

3.10 Schematic drawing <strong>of</strong> concept D . . . . . . . . . . . . . . . . . . . . 29<br />

3.11 Actively adapting pitch angle using the trailing edge . . . . . . . . . 30<br />

3.12 Geometric sketch for calculations <strong>of</strong> the trailing edge motion . . . . 30<br />

3.13 Left: Trajectories <strong>of</strong> leading edge, traling edge and chord angle,<br />

Right: Shifted graph for the leading edge motion for comparison<br />

<strong>of</strong> the harmoinc behaviour <strong>of</strong> the trailing edge’s motion . . . . . . . 31<br />

3.14 Actively adapting pitch angle using the leading edge . . . . . . . . . 31<br />

3.15 Simulated chord angle for horizontal actuation <strong>of</strong> wing rod . . . . . . 32<br />

3.16 Sketch <strong>of</strong> general principle for passive pitching at the hinge . . . . . 33<br />

3.17 Passive pitching done at the wings . . . . . . . . . . . . . . . . . . . 33<br />

4.1 Structure <strong>of</strong> the wing . . . . . . . . . . . . . . . . . . . . . . . . . . . 38<br />

5.1 Overview <strong>of</strong> resulting mechanism . . . . . . . . . . . . . . . . . . . . 40<br />

5.2 Connection <strong>of</strong> motor to rotating link . . . . . . . . . . . . . . . . . . 40<br />

5.3 Design for guiding the center joint . . . . . . . . . . . . . . . . . . . 41<br />

5.4 Assembly <strong>of</strong> wing joint . . . . . . . . . . . . . . . . . . . . . . . . . . 42<br />

5.5 Structure <strong>of</strong> the wing attachments . . . . . . . . . . . . . . . . . . . 42<br />

vi


Chapter 1<br />

Introduction<br />

Over the past twenty-five years interest in small unmanned aerial vehicles has greatly<br />

increased. Specially for reconnaissance and surveillance missions these vehicles are<br />

<strong>of</strong> great use. Most <strong>of</strong> them, which are used today incorporate traditional methods<br />

for lift and thrust, a propeller for thrust and fixed wings with an appropriate pr<strong>of</strong>ile<br />

to gain enough lift. Also rotary drive systems as can be seen by helicopters are used<br />

by some. However natural flying creatures are still superiour in terms <strong>of</strong> manoeuverability,<br />

lightweight and endurance.<br />

This fact motivates to find a MAV, which mimics the flapping motion <strong>of</strong> small birds,<br />

bats or insect, to have the same advantages. Also the improving technology, for instance<br />

lightweight and robust materials and better batteries, make this task more<br />

feasible and therefor the field <strong>of</strong> research <strong>of</strong> <strong>Flapping</strong> <strong>Wing</strong> MAVs has increased<br />

remarkably over the past years.<br />

Several <strong>Flapping</strong> <strong>Wing</strong> MAVs are already developed. The most successful is the<br />

DelFly, which has been realised by a group <strong>of</strong> undergraduate students at TU Delft<br />

in the Netherlands. Many other vehicles built so far use two wings, as it can be<br />

observed in nature at birds. DelFly is more a copy <strong>of</strong> the dragonfly, it uses two pairs<br />

<strong>of</strong> wings (see figure 1.1). The wings flap in counterphase and almost touch each<br />

other when they come together, for which reason it is assumed that it makes use <strong>of</strong><br />

the clap-and-fling effect (Chapter 2.1.3). However the question, why this concept<br />

works so well is still open, the investigations and measurements to reveal the secret<br />

have just started.<br />

Figure 1.1: Schematic drawing <strong>of</strong> DelFly I taken from www.delfly.nl<br />

1


Chapter 1. Introduction 2<br />

Other UAVs are yet less successful compared to the DelFly. However some other<br />

promising projects are still ongoing. For instance ROBUR from the University <strong>of</strong><br />

Paris, France. It has bigger dimensions, comparable to them <strong>of</strong> a seagull, and uses<br />

a more heavy and complex mechanism (figure 1.2), but again can perform much<br />

more wing motions. It can independently control the pitch angle <strong>of</strong> the wing and<br />

the flapping speed.<br />

Figure 1.2: <strong>Flapping</strong> wing mechanism <strong>of</strong> ROBUR taken from IROS 2007<br />

www.flyingrobots.org<br />

Another project from the University <strong>of</strong> California, Berkely, which investigates smaller<br />

dimensions is the robotic insect <strong>of</strong> Robert Wood [24]. It is at-scale <strong>of</strong> insects and<br />

has a fascinating lift production for this small scale. However longer flights are not<br />

possible, as the lift is indeed enough to let the Robotic insect fly, but also to carry<br />

a battery and control modules is due to limitations <strong>of</strong> the actual technology not yet<br />

possible.<br />

In the following chapters, it is presented how a flapping wing mechanism for an<br />

artificial bird with a size <strong>of</strong> approximately 20cm is developped. This project is the<br />

first step into this direction. Therefor in chapter 2 a detailed literature review is<br />

done to see which natural flapping flyer is most suitable for mimicking. Also briefly<br />

the general aerodynamic phenomena <strong>of</strong> flapping wings are summarised, which have<br />

to be considered and could give useful inputs. The result <strong>of</strong> this investigation is<br />

then used as a starting point to generate different concepts (chapter 3), without<br />

going too deep into the design and only theoretical calculations are done to check<br />

the feasibility. In chapter 4 the concept are compared with each other and the best<br />

is chosen to design in 3D CAD, which is briefly described in chapter 5.


Chapter 2<br />

Review<br />

2.1 Aerodynamics <strong>of</strong> flapping wings<br />

Compared to fixed wing flight, flapping the wings induce in general different aerodynamic<br />

phenomena. Most <strong>of</strong> the airflow is turbulent and due to permanently<br />

changed wing position and orientation, more the unsteady aerodynamics have to<br />

be considered. Because these informations could be <strong>of</strong> use for the development <strong>of</strong> a<br />

flapping wing mechanism, in this section shortly the main aerodynamics phenomena<br />

<strong>of</strong> flapping wings are described, using [11] as main input.<br />

2.1.1 Wagner Effect<br />

When a wing with a high angle <strong>of</strong> attack starts suddenly to move, the airflow vortices<br />

do not immediately get their steadystate value. The circulation slowly approaches<br />

to it. This delay results <strong>of</strong> a combination <strong>of</strong> two phenomena [19]. Firstly the fluid<br />

is not perfect, meaning it has a viscous behaviour on the stagnation point and so<br />

it takes some time to establish the Kutta condition. Also during the process the<br />

vorticity is generated and again shed at the trailing edge, while this shed vorticity<br />

forms a starting vortex. The velocity field near the wing, which is induced by the<br />

shed vorticity at the trailing edge counteracts to the bounding <strong>of</strong> the vortex to the<br />

wing. Only when the starting vortex has moved enough far away <strong>of</strong> the trailing<br />

edge, the moved wing gets its maximum circulation. This slow developement <strong>of</strong><br />

circulation was first proposed by Wagner in 1925 and so is called as the Wagner<br />

effect.<br />

Unlike the other unsteady mechanisms described below, this effect is not as strong.<br />

Specially at Reynolds numbers, which are typically for small birds or insects it<br />

can be neglected for flapping wings. However for more detailed studies <strong>of</strong> the<br />

aerodynamics, it is still considered.<br />

2.1.2 Leading edge vortex<br />

One <strong>of</strong> the most important effects for flapping wing flight is the leading edge vortex<br />

(LEV), which is created at high angles <strong>of</strong> attack. Operating the wing at a high angle<br />

<strong>of</strong> attack leads for a steady flow regime to flow separation and stall. However in<br />

unsteady flow, the created vortex at the leading edge, stays attached to the wing for<br />

a great part <strong>of</strong> the downstroke. This attached vortex induces a velocity downwards<br />

and so increases the lift force as shown in figure 2.1. Only when the vorticity <strong>of</strong> the<br />

leading edge vortex gets too large, the flow is not reattached before the trailing edge<br />

any more and a trailing edge vortex is formed, where the wing is in state similar<br />

to stall, which results in a sudden drop <strong>of</strong> lift. This described behaviour, for a<br />

3


Chapter 2. Review 4<br />

Figure 2.1: Leading edge vortex on the wing[19]<br />

Thick black lines indicate the downwash due to the generated vortex system<br />

two-dimensional wing motion, is called dynamic or delayed stall. The evolution <strong>of</strong><br />

the leading edge vortex for a translating wing starting from rest is shown in figure<br />

2.2. For the three-dimensional case as shown in figure 2.2, the leading edge vortex<br />

Figure 2.2: Evolution <strong>of</strong> a leading edge vortex in (A) two dimensions and (B) three<br />

dimensions during linear translation starting from rest [19]<br />

is more stable and no trailing edge vortex forms. Several different studies try to<br />

explain the stability <strong>of</strong> the formed leading edge vortex [12] [3], which is only present<br />

for the three-dimensional case. But still newer studies show, that the LEV has long<br />

been underestimated and is far more complex than assumed so far [13].<br />

2.1.3 Clap and fling mechanism<br />

Another phenomenon is the clap and fling mechanism showed in figure 2.3. Here<br />

the wings come together at the end <strong>of</strong> each upstroke to perform a so called ’clap’.<br />

After the clap the trailing edges <strong>of</strong> the wings stay connected, while the leading edges<br />

are increasing their distance to each other, which is called as ’fling’. So an opening<br />

angle is created. When the wings then start their downstroke, air is sucked into this<br />

funnel-like geometry, which induces a bound vortex at the leading edge <strong>of</strong> each <strong>of</strong><br />

the wings, and each created vortex acts as a starting vortex for the other wing. As<br />

described by Weis-Fogh [22] this annihilation allows the circulation to be builded up<br />

more rapidly, because the Wagner effect (see section 2.1.1) is suppressed. Another


5 2.1. Aerodynamics <strong>of</strong> flapping wings<br />

Figure 2.3: Schematic representation <strong>of</strong> the clap (A-C) and fling (D-F) [19]<br />

Black lines show trajectory <strong>of</strong> the airflow, dark blue arrows represent the by the airflow induced<br />

velocity, light blue arrow shows the net force on the airfoil.<br />

advantage <strong>of</strong> the clap is that the created vortices during upstroke are vanishing<br />

during the clap, they cancel each other out as they are oriented in opposite direction.<br />

Many insects make use <strong>of</strong> the fling to create a rotational airflow circulation, while<br />

the clap is not performed by all insects. According to Ellington [8] the clap is<br />

avoided by most <strong>of</strong> the insects because the permanent clapping can damage the<br />

wings and more a ’almost’ clap is performed. Also for birds similar observations<br />

were made, for instance during the take<strong>of</strong>f <strong>of</strong> pigeons [14]. Although no full clap and<br />

fling is performed, the wings almost touch each other at the back and it is assumed,<br />

that in this way similar air circulations are produced, which give additional lift.<br />

2.1.4 Rotational lift<br />

Near the end <strong>of</strong> every stroke mainly insects but also some small birds (e.g. hummingbirds)<br />

are rotating their wings, which allows to maintain a positive angle <strong>of</strong><br />

attack during the whole wingbeat cylce. The three different phases are shown in<br />

Figure 2.4: Three phases <strong>of</strong> the wing rotation [7]


Chapter 2. Review 6<br />

figure 2.4. The angles <strong>of</strong> attack during downstroke and upstroke are αd and αu respectively.<br />

ω indicates the angular velocity. This rotation at the end <strong>of</strong> each stroke,<br />

also gives additional lift. According to Dickinson the generated lift force strongly<br />

depends on the the position <strong>of</strong> the rotation axis. For instance rotations about the<br />

trailing edge show a better lift generation compared to rotations about the leading<br />

edge for instance. Also the timing <strong>of</strong> the rotation has an effect on the produced lift,<br />

which is analysed for instance in [7].<br />

2.1.5 <strong>Wing</strong>-wake interactions<br />

The back and forth motion <strong>of</strong> the wings used by insects make the wings interact with<br />

the shed vorticity <strong>of</strong> the prior strokes, which acts positively on the lift generation.<br />

Figure 2.5 shows the principle <strong>of</strong> wake capture. At the end <strong>of</strong> the translational<br />

Figure 2.5: <strong>Wing</strong>-wake interaction during stroke reversal [19]<br />

U∞ indicates the freestream velocity, dark blue arrows show the induced velocity field, light blue<br />

arrows presents the aerodynamic force<br />

phase (A), the wing starts the rotation (B), which causes the vortices at the edges<br />

to shed <strong>of</strong>f the wing (C). This induces a strong velocity field, which pushes against<br />

the wing (D) and so increases the lift force at the beginning <strong>of</strong> the next halfstroke<br />

(E). In the following translational phase, again a LEV is created (F). This wingwake<br />

interaction also allows to let the pitch motion <strong>of</strong> the wing done passively, as<br />

this additional lift, at the beginning <strong>of</strong> the stroke, rotates the wing to the desired<br />

orientation, to maintain a positive angle <strong>of</strong> attack.<br />

2.1.6 Lift force<br />

The lift force is produced by the four unsteady effects described above. The most<br />

important one is the LEV, because it is the only one, which is responsible for lift<br />

during the flapping, the translational phases <strong>of</strong> the strokes. The other three effects<br />

enhance the lift production mainly during the rotational phases. In figure 2.6 an<br />

example <strong>of</strong> the generated lift force is shown. During hovering, the horizontal component<br />

<strong>of</strong> the red arrows in figure 2.6 cancel out during one whole stroke cycle. The<br />

vertical component equals the body weight.<br />

Of course the wings play also an important role for the produced lift force. Like for<br />

fixed wing aircraft, the wing pr<strong>of</strong>ile determines lift and drag coefficients (CL and<br />

CD). However for flapping flight, these also differ. For a steady state flow regime,<br />

like for fixed wing aircraft, these two coefficients can be deteremined independently


7 2.2. <strong>Flapping</strong> wings in nature<br />

Figure 2.6: Flight forces for the drosophila during hovering [21]<br />

The red arrows indicate the net forces during down- and upstroke.<br />

from each other. For an unsteady flow field they can not be seperated anymore [11].<br />

<strong>Flapping</strong> flight is usually performed with a high angle <strong>of</strong> attack, to get the above<br />

described LEV, which induces a force normal to the wing surface. Hence the resultant<br />

force is drag- and lift force in one hand. Therefor to have a similar description<br />

as for fixed wing flight Dickinson [20] defined a circulatory coefficient, which can be<br />

merged out <strong>of</strong> the usual drag and lift coefficient.<br />

<br />

CT = C2 D + C2 L<br />

(2.1)<br />

However, this coefficient also has to be determined experimentally. Also for a given<br />

wing pr<strong>of</strong>ile, and known lift- and drag coefficients for a steady flow regime, there<br />

is no way around to obtain the circulatory coefficient, but to make experimental<br />

measurements, because the unsteady effects <strong>of</strong> the flapping flight give different<br />

results.<br />

2.2 <strong>Flapping</strong> wings in nature<br />

For developing a flapping wing mechanism different flying animals are studied. As<br />

the future MAV should have the ability to hover, mainly animals with hovering capabilites<br />

are examined. Also the dimensions should approximately match the MAV,<br />

that for a first approach the feasibility can be taken for granted. Therefor in this<br />

section the wing motions <strong>of</strong> hummingbirds, bats and smaller birds are summarized.<br />

Although insects are much smaller and will not serve as main input for the flapping<br />

wing mechanism, some ideas may be extracted and hence roughly the kinematics<br />

and aerodynamics are summarized by the example <strong>of</strong> the Drosophila fruit fly.<br />

2.2.1 Insects<br />

The stroke shape in flying insects are varying remarkably. The wing tip makes<br />

depending on the insect, different motions. Oval, figure-eight or pear-shaped trajectories<br />

[16], or combinations <strong>of</strong> those patterns are done. Also some insects may<br />

change the stroke trajectory for strong manoeuvers. And for increasing forward<br />

flight again other wing motions occur. Because this is a wide range, the most simple<br />

wing motion during hovering done by the drosophila fruit fly is investigated. For


Chapter 2. Review 8<br />

Figure 2.7: General pattern for the wing motion <strong>of</strong> Drosophila Melanogaster [16]<br />

further informations on the behaviour <strong>of</strong> other wing motions performed by other<br />

insects see for instance [16] [9].<br />

<strong>Wing</strong> motion<br />

In the most common form <strong>of</strong> hovering in insects the wings move along an approximately<br />

horizontal stroke plane with approximately equal and relatively high angles<br />

<strong>of</strong> attack during the downstroke and upstroke. This is done by fast rotating the<br />

wing at the end <strong>of</strong> each half stroke. The general pattern can be seen in figure 2.7.<br />

The whole stroke cycle can be described by a sinusoidal motion or a triangular motion<br />

depending on the insect. For the Drosophila Melanogaster the stroke trajectory<br />

is more a triangular motion with an amplitude <strong>of</strong> 130-160 degrees and a flapping<br />

frequency <strong>of</strong> 250 Hz. The stroke plane angle with respect to the horizontal is about<br />

10 degrees, while the body angle is tilted about 60 degrees. These values measured<br />

by [9] are presented in figure 2.8. Another important aspect is the ratio <strong>of</strong> the<br />

duration <strong>of</strong> the downstroke compared to the upstroke, which is approximately 0.8<br />

and shows that usually the downstroke is performed faster than the upstroke.<br />

Figure 2.8: Kinematics <strong>of</strong> Drosophila Melanogaster [9]<br />

(A) <strong>Wing</strong> tip trajectory in degrees, (B) <strong>Wing</strong> tip path drawn with respect to the body, which is<br />

represented by an arrow


9 2.2. <strong>Flapping</strong> wings in nature<br />

Aerodynamics<br />

Insects are able to hover by using a range <strong>of</strong> possible unsteady high-lift mechanisms,<br />

including rotational circulation, clap-and-fling and wake capture (see section 2.1).<br />

However, arguably the most important mechanism is the leading-edge vortex, which<br />

may generate up to 66% <strong>of</strong> the total lift in insect flight [23]. Consequently the high<br />

angle <strong>of</strong> attack to create the LEV is crucial for generating enough lift force.<br />

The almost symmetric stroke pattern, meaning that upstroke and downstroke are<br />

very similar as described above, 50% <strong>of</strong> the resulting lift force comes out <strong>of</strong> the<br />

downstroke, respectively out <strong>of</strong> the upstroke. Also the wing material is flexible, such<br />

that a camber occurs, which additionally gives lift force. Due to no morphological<br />

constraints this camber is inverted during the upstroke, which allows to maintain<br />

almost the same aerodynamic forces acting on the wing as during the downstroke.<br />

In figure 2.9 exemplary the generated forces are shown, which are taken out <strong>of</strong><br />

measurements made with a flapping device having the similar stroke trajectory as<br />

the Drosophila melanogaster [16].<br />

Figure 2.9: Force production in two cycles [16]<br />

(A) Vertical force acting on the wing (black line), (B) Translational angular wing motion (black),<br />

the wing’s angle <strong>of</strong> attack (blue) and heaving motion (green)<br />

It can be seen that the lift force is generated during up and downstroke. Also the<br />

clap and fling plays a role at the transition <strong>of</strong> the down- and upstroke and generates<br />

extra lift force. It is important to notice that the generated lift force highly depends<br />

on the stroke trajectory [16]. One can assume that if a wing is moved with the<br />

same angle <strong>of</strong> attack and rotational velocity, the same lift force should occur. But<br />

different stroke trajectories change the airflow pattern, the created vortices and so<br />

the generated lift force. Therefore insects have in general different stroke pattern,<br />

which are more or less effective, but at least enough to let them fly.


Chapter 2. Review 10<br />

2.2.2 Hummingbirds<br />

Most studies <strong>of</strong> hummingbirds are based on the rufous hummingbirds, because <strong>of</strong><br />

their practical properties for experimental measurements in the wind tunnel. They<br />

can be trained and thrive well in captivity. Although they have a body mass <strong>of</strong><br />

3-4g and a wing span <strong>of</strong> 110mm and so would be too small and lightweight for a<br />

prototype MAV, their wing motion still can be mimicked because the biggest existing<br />

hummingbird, the Giant Hummingbird (Patagonia Gigas) has according to<br />

biologists in general patterns the same kinematics, weights about 20g and has a<br />

wingspan <strong>of</strong> 280mm.<br />

Figure 2.10: <strong>Wing</strong> motion relative to the body flying at velocities <strong>of</strong> 0 − 12ms −1 [2]<br />

(A) Dorsal view with bird silhouette at mid-downstroke. (B) Lateral view with bird silhouette at<br />

start <strong>of</strong> downstroke.<br />

<strong>Wing</strong> Motion<br />

The rufous hummingbird flaps their wings with a frequency <strong>of</strong> 40-45 Hz. For bigger<br />

species the flapping frequency decreases. For instance the giant hummingbird has a<br />

flapping frequency <strong>of</strong> about 10-15 Hz. The main characteristics <strong>of</strong> the wing motion<br />

for several different forward flight speeds (0m/s−12m/s) can be seen in figure 2.10.<br />

Black circles indicate position <strong>of</strong> wingtips, white circles indicate position <strong>of</strong> wrists<br />

which is approximately in the middle <strong>of</strong> the wing.


11 2.2. <strong>Flapping</strong> wings in nature<br />

During upstroke <strong>of</strong> slow flight (0m/s and 2m/s), the tips and wrists trace in reverse<br />

nearly the same paths that were exhibited during downstroke. The lateral view reveals<br />

the wingtip describing an upwardly concave path, where the tips also follow<br />

a slight horizontal figure-8 pattern. In figure 2.12 the flapping motion is shown<br />

more detailed. The wrist elevation indicates the position <strong>of</strong> the wrist relative to the<br />

mid-frontal plane, which is described by the bird’s torso, and the chord angle describes<br />

the pitching <strong>of</strong> the wing with respect to this body plane. It can be seen that<br />

the flapping motion is sinusoidal, where the downstroke is performed insignificantly<br />

faster than the upstroke. Also the chord angle follows a sinusoidal trajectory, with<br />

a phase shift and an <strong>of</strong>fset compared to the wrist elevation.<br />

Figure 2.11: Angles describing bird-centered wing and body kinematics in rufous<br />

hummingbirds [2]<br />

β is the body angle w.r.t. horizontal, γh and γb is the stroke plane angle relative to horizontal<br />

respectively to body angle<br />

During upstroke almost no wing folding is present. According to [2], [17] the<br />

wingspan ratio upstroke:downstroke is about 0.98 for slow flight and decreases to<br />

0.90 for faster flying speeds up to 12m/s, where most <strong>of</strong> the flexing is done at the<br />

outer parts <strong>of</strong> the wing, between the wrist and the tip. As for slow speeds this<br />

ratio stays more or less constant, the wings can be taken as kinematically ’rigid’<br />

compared to other avian species.<br />

As can be seen in figure 2.11 for transition from hovering to a forward flight speed<br />

<strong>of</strong> 2m/s the stroke plane angle with respect to the body γb can be assumed to be<br />

constant. In general mainly the body angle β is tilted to achieve a forward velocity<br />

for slow flight speeds. For higher speeds <strong>of</strong> course more parameters are varying<br />

significantly. For instance it can be seen in figure 2.12 that the maximal chord angle<br />

reduces significantly for increasing forward flight speed and generally the stroke<br />

amplitude increases.<br />

Aerodynamics<br />

Although the aerodynamic characteristics <strong>of</strong> the hummingbirds wingbeat are very<br />

complex, several studies reveal some useful information. The main flow pattern<br />

can be described as shown in figure 2.13. During the flapping motion trailing-tip<br />

vortices are created. These vortices induce starting and stopping vortices <strong>of</strong> the<br />

downstroke. The resultant air circulation origined <strong>of</strong> these vortices are the main<br />

effects, besides the usual aerodynamic phenomena <strong>of</strong> flapping wings (Section 2.1),<br />

which are adequate to support the weight <strong>of</strong> the hummingbirds. A more detailed<br />

illustration can be seen in figure 2.14.


Chapter 2. Review 12<br />

Figure 2.12: Variation <strong>of</strong> chord angle relative to body-plane during wingbeats at<br />

velocities <strong>of</strong> 0 − 12ms −1 [2]<br />

<strong>Wing</strong>beat duration is expressed as a precentage <strong>of</strong> entire wingbeat. Broken line indicates wrist<br />

elevation relative to body-plane. Shaded area represents downstroke. Values are means ±s.d<br />

In the frontal view, the tip vortices <strong>of</strong> the downstroke (D) and the upstroke (U)<br />

are indicated. In the side view, between the stopping vortex <strong>of</strong> the downstroke (D)<br />

and the starting vortex <strong>of</strong> the upstroke (U) is a pocket <strong>of</strong> vorticity LEVD created<br />

at the leading edge <strong>of</strong> the wing during the rapid wing pronation at the beginning<br />

<strong>of</strong> the preceding downstroke, and carried through the downstroke to be shed during<br />

the supination at the beginning <strong>of</strong> the upstroke. The resultant airflow downwards<br />

gives the needed lift force for hovering.<br />

More studies on the airflow revealed that a force asymmetry between upstroke and<br />

downstroke is present. Hummingbirds produce 75% <strong>of</strong> their weight support during


13 2.2. <strong>Flapping</strong> wings in nature<br />

Figure 2.13: Wake structures in frontal and side plane [17]<br />

the downstroke and only 25% during the upstroke [17], although the kinematics <strong>of</strong><br />

the wing motion is symmetric, as for insects. It is assumed that this asymmetry<br />

is present due to slight difference <strong>of</strong> the angular velocity during downstroke and<br />

upstroke, a missing leading edge vortex during upstroke and several musculoskeletal<br />

and planform material properties, which do not allow the hummingbird’s wing to<br />

behave equally efficient as the insect’s wing. For instance during the downstroke<br />

the wing is slightly cambered, while during the upstroke the wing is not capable to<br />

invert the camber, which gives a significant loss <strong>of</strong> the produced lift force.<br />

Figure 2.14: Flow field vorticity at end <strong>of</strong> upstroke, (a) frontal view at shoulderplane,<br />

(b) side view at midwing-plane [17]<br />

2.2.3 Bats<br />

There are many bats species living on earth, which differ in size, weight and some<br />

other anatomical aspects [15]. But as the wing motion was observed to be similar<br />

for most <strong>of</strong> the species [18], mainly the studies about the lesser short-nosed fruit<br />

bat Cynopterus Brachyotis are considered, which give a sufficient insight to the<br />

aerodynamics and kinematics aspects <strong>of</strong> the bat flight.<br />

<strong>Wing</strong> motion<br />

As can be seen in figure 2.15 the bat wings possess more than two dozen joints, which<br />

can be controlled independently [5] and has bones that deform adaptively during<br />

the motions <strong>of</strong> the wingbeat cycle. Of course this anatomical structure is crucial for<br />

the motion <strong>of</strong> the wing and so very complex trajectories are fullfilled. As can be seen<br />

in figure 2.16 the general motion is characterized by a cambered wing during the<br />

downstroke, and a folding <strong>of</strong> the wing during the upstroke. To simplify the upstroke<br />

it could be described as additional delays for joints approaching the thorax with


Chapter 2. Review 14<br />

Figure 2.15: Anatomical structure <strong>of</strong> the bat wing [5]<br />

Figure 2.16: Sequences <strong>of</strong> images from below and in front <strong>of</strong> bat during on cycle<br />

starting at beginning <strong>of</strong> the downstroke [6]<br />

respect to the wing tip. So the motion <strong>of</strong> the next inner joint, the finger joint,<br />

compared to the wingtip is delayed, while the wrist then again is delayed compared<br />

to the finger joint and so on [5]. During downstroke the wing is approximately<br />

stretched, with a almost synchronous movement <strong>of</strong> all joints but also with increasing<br />

delays for the inner wing parts as can be seen on the right diagram <strong>of</strong> figure 2.17. The<br />

shoulder is the most proximal point <strong>of</strong> the wing. The wrist is the next distal joint,<br />

followed by the MCP III and the tip <strong>of</strong> the third digit as the furthest measurement<br />

point <strong>of</strong> the wing. Hence the kinematics are not simple. Even if only the wing<br />

Figure 2.17: Example trajectories <strong>of</strong> the different wing regions for 3m/s (left) and<br />

9m/s (right) [5]<br />

Zero represents the vertical position <strong>of</strong> the animal’s center <strong>of</strong> mass. Radius in red (lower arm),<br />

Humerus in dark blue (upper arm), MCP in light blue (knuckle), shoulder in black<br />

tip position is observed, it can be seen in figure 2.18 that the trajectory can not<br />

be realised by a simple mechanical mechanism. Also for increasing flight speed for<br />

example the wingtip elevation increases significantly, and the shoulder follows an<br />

entirely other trajectory compared to slow flight or hovering.<br />

According to [5] also changes in the length <strong>of</strong> the different bones and the membrane<br />

in the wing occur, which is again a reason for the above presented complex wing<br />

motion.


15 2.2. <strong>Flapping</strong> wings in nature<br />

Figure 2.18: Example <strong>of</strong> wing tip motion [6]<br />

Circles indicate the wing tip position for one whole cycle; The cross indicates the center <strong>of</strong> mass<br />

<strong>of</strong> the bat’s thorax.<br />

Aerodynamics<br />

Lift mechanisms in bat flight origined <strong>of</strong> unsteady effects are not studied very detailed<br />

yet. Regardless some measurements <strong>of</strong> the airflow using digital particle image<br />

velocimetry were documented. According to [10] the wing camber during downstroke<br />

is about 18% <strong>of</strong> the wing chord and the average angle <strong>of</strong> attack, where the<br />

wing is operating is about 50 ◦ . It is important to notice that if a fixed wing operates<br />

at such values, it would stall and lose lift, which already presumes that the bat<br />

flight is very complex and not very simply comparable with other flying animals.<br />

The main contribution to the lift force was found to be given by the LEV [10], which<br />

is shown in the following more detailed. Figure 2.19 show that the flow separates<br />

Figure 2.19: Velocity and vorticity fields around a bat wing in slow forward flight<br />

(1 m/s) at the time instance when the wing is in horizontal position during the<br />

downstroke [10]<br />

at the leading edge, generating an area <strong>of</strong> high negative vorticity. Behind this area<br />

the airflow reattaches, which results in an attached and laminar flow at the trailing


Chapter 2. Review 16<br />

edge. The vorticity is stronger near the wingtip (C) and deacreases toward the wing<br />

root (A).<br />

At the trailing edge, mainly distally on the wing, an area with negative vorticity is<br />

found, which results <strong>of</strong> a strong rotational movement before the end <strong>of</strong> the downstroke,<br />

which also enhances lift generation (see section 2.1.4). During the upstroke<br />

the vortex, which generates much <strong>of</strong> the lift in flapping-wing flight, is not documented<br />

well. It does not appear to origin in the wingtips as it is the case for the<br />

downstroke. According to biologists the starting point for the vortex seems to be<br />

somewhere in the middle <strong>of</strong> the wing, which again shows, that the complex wing<br />

structure, with the many joints is crucial for the whole bat flight.<br />

2.2.4 Birds<br />

For this section mainly the smaller birds are considered. Bigger birds are using more<br />

aerodynamic effects as for fixed wing flight, for instance gliding. Small birds need<br />

to generate the lift force by flapping the wings. However there are many different<br />

types <strong>of</strong> birds, which also have different kinematics.<br />

In general the wing can be tentatively separated into two parts, the outer wing and<br />

the inner wing. The inner wing acts like an aircraft wing, it is the lift developing<br />

part <strong>of</strong> the wing. When a bird flaps its wing it is the inner wing that moves the<br />

smallest distance, thus the lift it generates is due, to a large extent, on the airstream<br />

produced by forward momentum. The inner wing is also the most cambered part<br />

<strong>of</strong> the wing and this is made possible by the extensive bones and connective tissue<br />

that can hold this shape better than feathers. This means that it can generate more<br />

lift per surface area than the outer wing, it also means that it will stall more easily.<br />

The outer wing is the powerplant <strong>of</strong> the wing, it produces lift, but more crucially<br />

Figure 2.20: <strong>Wing</strong>span ratio as a function <strong>of</strong> flight velocity compared among bird<br />

species [2]<br />

it produces forward momentum. It is less cambered than the inner wing and more<br />

flexible and it is this flexibility that leads to the momentum. As the wing is flapped<br />

downward the outer wing tends to twist slightly forwards, this is due to a number<br />

<strong>of</strong> reasons, one being that air passing under the wing tends to well up toward the


17 2.3. Summary<br />

tip and as it does so it forces its way out under the back <strong>of</strong> the wingtip, tilting the<br />

wing forward.<br />

During the upstroke the feeders at the outer wing are spread to reduce the drag.<br />

Also the wing is folded for most <strong>of</strong> the species significantly (see for instance figure<br />

2.20).<br />

Unfortunately, there is not much literature dealing explicitly with the aerodynamics<br />

<strong>of</strong> small birds. Also note, that compared to hummingbirds no so detailed informations<br />

about the kinematics could be found for flight during hovering, because <strong>of</strong><br />

their less practical properties for experimental tests. Nonetheless briefly the flapping<br />

parameters are given exemplarily for the zebra finch, which belongs to the<br />

same family as the siskins and is a good representation for most <strong>of</strong> the small birds.<br />

Kinematics <strong>of</strong> flapping flight in the zebra finch<br />

Zebra finches have a body mass <strong>of</strong> about 13g with a wingspan <strong>of</strong> 170mm. They<br />

flap their wings with about 24Hz and a stroke amplitude <strong>of</strong> 135 ◦ , which decreases<br />

significantly for increasing the flight velocity [4]. As for hummingbirds the body<br />

angle is tilted for increased forward flight speed. For hovering the body angle with<br />

respect to the horizontal is about 50 ◦ , which decreases down to 15 ◦ for a flight<br />

velocity <strong>of</strong> 12m/s. The angle <strong>of</strong> incidence for the wing is for hovering about 75 ◦<br />

and decreases for a flight velocity <strong>of</strong> 12m/s to 15 ◦ . However the chord angle stays<br />

approximately constant for all flying velocities at about 20 ◦ .<br />

Another important aspect, is that the finch not regularly flaps it’s wings. Depending<br />

on the flight velocity the wing is bounded after several stroke cycles for some time<br />

instances. For higher velocities almost 50% <strong>of</strong> the time, the small birds hold their<br />

wings close to the body, do not flap them and can save so some energy. This<br />

behaviour can be seen in figure 2.21, where the wingtip elevation and the wingspan<br />

are shown for a flight speed <strong>of</strong> 2m/s (A) and 12m/s (B). As no lift is generated<br />

Figure 2.21: Representative wing kinematics in a zebra finch engaged in flapbounding<br />

flight at 2m/s (A) and 12m/s (B) [4]<br />

with flapping or gliding during the bounded time span, the aerodynamic properties<br />

<strong>of</strong> the body come to be crucial.<br />

2.3 Summary<br />

In the following table the characterization <strong>of</strong> the kinematics <strong>of</strong> the different investigated<br />

flying animals (Insects-Drosophila fruit fly, rufous Hummingbirds, Shortnosed<br />

Bats Cynopterus brachyotis) are summarized for hovering flight. Note that<br />

morphological data and results out <strong>of</strong> biological experiments are taken either as average<br />

values or most suited values. Specially for insects like the Drosophila fruit fly,


Chapter 2. Review 18<br />

the accuracy <strong>of</strong> measurements is limited, because <strong>of</strong> their small size, and deduced<br />

informations <strong>of</strong> experiments done with accurate models which represents good results<br />

for the insect flights are presented.<br />

Insects Hummingbirds Bats Siskin<br />

Weight [g]


19 2.3. Summary<br />

Siskin/Finch<br />

Advantages -very maneuverable<br />

-hovering and forward flight possible<br />

Drawbacks -wing folding is significant during upstroke<br />

-no constant flapping frequency for increasing forward flight<br />

speed<br />

-stroke amplitude reduces significantly for increasing forward<br />

flight<br />

Small birds also have an acceptable hover ability. But compared to hummingbirds<br />

it is decreased. According to biologists, the hummingbird’s should can do more<br />

different motions, for which reason crucial changes in the kinematics <strong>of</strong> the flapping<br />

occur. Also the wrist, the joint at the approximate midpoint <strong>of</strong> the wing, is more<br />

essential. Specially during upstroke the wing is folded significantly, which would<br />

be very difficult to implement in a MAV. Also for transition from hovering to forward<br />

flight many different flapping parameters are changing significantly, whereas<br />

no resonable simplification for a flapping device can be estimated. Hence, the small<br />

birds, are not taken into account for further investigations.<br />

Bat<br />

Advantages -very maneuverable<br />

-hovering and forward flight possible<br />

-low flapping frequency compared to animal’s size<br />

-can generate greater lift for less energy due to stretchy membrane<br />

Drawbacks -very complex wing structure, more than two dozen independently<br />

controlled joints<br />

-highly articulated motion and complex kinematics<br />

-deforming bones<br />

The study <strong>of</strong> the bat flight also exclude the bat wing motion as a main input for<br />

developing a MAV. A simple mechanical flapping mechanism could not be realised,<br />

because the wing motion is far too complex, with more than a dozen independently<br />

controllled joints, which would let the MAV be too heavy. Also no simplifications<br />

could be found, which would allow to make a simplified kinematic model and still<br />

follows the wing’s trajectory in a similar way as the natural bat.<br />

However some ideas could be filtered out <strong>of</strong> the bat-flight as for example, to let the<br />

outer wing parts follow a delayed trajectory with respect to the inner wing parts,<br />

which roughly describes the bat-flight. Also attaching the wing to the tail could be<br />

a reasonable idea.


Chapter 2. Review 20<br />

Hummingbird<br />

Advantages -very maneuverable<br />

-hovering and forward flight possible<br />

-almost no wing folding during upstroke<br />

-at first sight a simplified mimic wing motion is achievable with<br />

a mechanical mechanism<br />

-flapping frequency stays constant for every flight speed<br />

Drawbacks -twisting phenomena along the wing axis is present like in other<br />

birds<br />

-kinematic parameters variation more complex for increasing<br />

forward flight<br />

-pitching is done actively<br />

As can be seen the hummingbird seems to be a reasonable choice for mimicking.<br />

Specially for transition from hovering to slow forward flight, very few kinematic<br />

parameters are changing, which simplifies the later control challenges. For more<br />

increasing the forward flight speed <strong>of</strong> course more parameters are varied, but for<br />

the first approach this can be neglected. Compared to small birds almost no wing<br />

folding is present, hence the wings can be assumed to designed without a joint, as<br />

it is the case for insects. Although the pitching <strong>of</strong> the wing is done actively by the<br />

hummingbird, this can be still achieved to copy. As shown above, the pitch angle<br />

also follows a more or less harmonic pattern. The twisting phenomena along the<br />

wing axis, also does not represent a big obstacle, as this can be solved by using<br />

flexible wings, which adapts itself to the aerodynamic loads.<br />

Therefor, to generate first concepts for the flapping wing mechanism, mainly the<br />

hummingbird motion is considered, which could be extended with ideas described<br />

for the bat-flight or simplified by some kinematic aspects <strong>of</strong> the insect.


Chapter 3<br />

Concepts<br />

3.1 General Considerations<br />

3.1.1 Objective characteristics<br />

As described above, hummingbirds are chosen to mimic. Therefor the dimensions<br />

<strong>of</strong> the Giant Hummingbird (Patagonia gigas) are taken as a starting point for the<br />

design. According to biologists the kinematics <strong>of</strong> the Giant Hummingbird are similar<br />

to the above described pattern <strong>of</strong> the rufous hummingbird and so can be also<br />

considered as the motion which the flapping mechanism has to fulfill.<br />

Dimensions<br />

The following table summarizes the dimensions <strong>of</strong> the Giant Hummingbird which<br />

are used.<br />

Weight 25g<br />

<strong>Wing</strong>span 280mm<br />

Aspect Ratio 6.73<br />

<strong>Wing</strong>chord ≈ 40mm<br />

Body width ≈ 50mm<br />

<strong>Wing</strong> length ≈ 115mm<br />

<strong>Flapping</strong> motion<br />

The general characteristics <strong>of</strong> the flapping motion are presented in the table below.<br />

<strong>Flapping</strong> frequency 15 Hz<br />

Stroke Amplitude ≈ 110 ◦<br />

Body angle during hovering 50 ◦<br />

Stroke plane angle during hovering 60 ◦<br />

flapping pattern sinusoidal<br />

chord angle trajectory sinusoidal<br />

max/min chord angle 100 ◦ / -35 ◦<br />

21


Chapter 3. Concepts 22<br />

3.1.2 Flight control<br />

<strong>Flapping</strong> flight is rather complex when control aspects are considered. For birds and<br />

insects several parameters <strong>of</strong> the flapping motion are changed to perform different<br />

maneuvers. For some control tasks several different ways can lead to the desired<br />

result. For instance for changing the forward flight velocity the pitch angle <strong>of</strong> the<br />

whole flying animal is changed. Therefor either the mean flapping angle is changed,<br />

the angle <strong>of</strong> attack is altered and/or the stroke amplitude is varied. The rolling<br />

angle can be controlled by increasing the flapping amplitude and/or the angle <strong>of</strong><br />

attack <strong>of</strong> the outer wing. For more complicated maneuvers many <strong>of</strong> the flapping<br />

parameters are changed simultaneously. Of course a flapping mechanism, which can<br />

be controlled in such a way would be much too complex and therefor too heavy for<br />

a MAV. Of course a more deep study is needed for a good flight control, but this<br />

can only be done, when the flapping mechanism is finished and implemented in a<br />

MAV. But as a first approach it is adequate to consider only the simplest control<br />

aspects.<br />

According to biologists, studying the hummingbird’s wing motion, a simple way to<br />

change the flight velocity is to tilt the body angle. As a first approach this can<br />

be done by shifting the center <strong>of</strong> gravity <strong>of</strong> the MAV forward or backward and/or<br />

using servos at the tail <strong>of</strong> the MAV. Changing other parameters <strong>of</strong> the flapping<br />

motion and taking this into account for developing a flapping mechanism would be<br />

to complicated at this early stage <strong>of</strong> the project.<br />

To change the flight direction also a simple solution is needed. In general birds<br />

change several parameters, for instance the stroke amplitude and the angle <strong>of</strong> attack,<br />

<strong>of</strong> each wing seperately. This again would be to complex, for which reason it is<br />

considered to change the orientation <strong>of</strong> the tail to deviate the air flow as a first<br />

assumption. Of course this has also to be investigated more deeply, when a flapping<br />

prototype is present.<br />

Therefor the development <strong>of</strong> a first flapping mechanism can be done independently<br />

<strong>of</strong> these control aspects. More precisely the flapping device needs only one actuator,<br />

which has to generate the correct motion to produce enough lift force. The control<br />

issues can be solved by using servos which change the orientation <strong>of</strong> the tail.<br />

3.1.3 Actuator<br />

To have a reasonable design for a MAV as less actuators as possible should be used<br />

to reduce the power consumption and the mass. Also in general the mechanical<br />

complexity then reduces, less joints and links are needed to transfer the forces <strong>of</strong><br />

the actuators to the wings and so is more lightweight.<br />

A brief investigation <strong>of</strong> the available actuators showed, that no reasonable linear<br />

actuator can be used. Either they are too big and too heavy or can not bring<br />

up the force needed for the flapping motion or the linear displacement needed for<br />

the stroke amplitude. As the future MAV is considered to be <strong>of</strong> a size similar to<br />

the Giant Hummingbird piezoactuators can also be excluded due to the too small<br />

generated forces. DC-Motors can fulfill these first constraints. Some, specially<br />

brushless DC-motors, could be found which have a reasonable torque, an acceptable<br />

power consumption and still a weight which is small enough to integrate in a MAV.<br />

Therefor the flapping mechanism will be designed using a rotary drive system.


23 3.2. Concepts for wing flapping<br />

3.2 Concepts for wing flapping<br />

3.2.1 Concept A<br />

To have a sinusodial flapping motion as it is present for the hummingbird (see<br />

figure 2.12), the main structure <strong>of</strong> the flapping mechanism can be approximated<br />

with a circular motion, which is generated with a rotational actuator and where the<br />

movement in the direction <strong>of</strong> one main axis is transmitted to the wings according<br />

to figure 3.1. However in such a way, the sinusodial motion <strong>of</strong> the wings only can be<br />

approximated. The resulting trajectory <strong>of</strong> the wing tip depends on the up and down<br />

Figure 3.1: Schematic drawing <strong>of</strong> concept A1<br />

movement <strong>of</strong> the centered guided joint, which again depends on the parameters L<br />

and r (see figure 3.2). Only for L going to infinity a perfect sinusodial motion with<br />

amplitude r can be achieved. For a good approximation therefor L has to be chosen<br />

much larger than r. The kinematic relationship is given with equations 3.1 and 3.2<br />

and is shown in figure 3.3 for various ratios L/r 1 .<br />

sin α =<br />

r · cos δ<br />

L<br />

(3.1)<br />

y = r · sin δ + L · cos α (3.2)<br />

Already a ratio higher than 2:1 for L:r can be considered as an approximation<br />

which is good enough to achieve an acceptable sinusodial motion. This can be either<br />

done by increasing L or decreasing r. It is important to point out, that for decreasing<br />

r, which affects the amplitude <strong>of</strong> the sinusodial movement <strong>of</strong> the centered joint, also<br />

the distance b has to be adapted according to equation 3.3 to get the desired stroke<br />

amplitude <strong>of</strong> βmax = 55 ◦ .<br />

b =<br />

r<br />

tan (βmax)<br />

(3.3)<br />

To reduce this dependency <strong>of</strong> b to the amplitude r, an additional horizontal link<br />

can be inserted according to figure 3.4. Instead one joint, the whole link is moved<br />

up and down and is connected over two joints to the wings to transmit this motion.<br />

Therefor the length <strong>of</strong> this link can be adjusted and assures more liberty for the<br />

later dimensioning <strong>of</strong> the different link lengths. However an additionally joint is<br />

needed, which <strong>of</strong> course reduces the efficiency.<br />

1 Generated with matlab kinematic circular.m


Chapter 3. Concepts 24<br />

Figure 3.2: Sketch for kinematics <strong>of</strong> general structure for the flapping motion<br />

Figure 3.3: Trajectories <strong>of</strong> centered joint for one cycle for different ratios L/r<br />

3.2.2 Concept B<br />

This concept is based on the same general structure as concept A, as the given<br />

sinusodial flapping motion does not let much margin for big variations. Therefor the<br />

general kinematic pattern is the same as described above in section 3.2.1. Anyway<br />

the structure presented in figure 3.5 can also be a good solution. The actuator’s<br />

torque is transmitted via two gears to the associated wing. The advantage <strong>of</strong> this<br />

concept as a starting point for the further designing is, that each wing can be<br />

treated somehow independently <strong>of</strong> each other in terms <strong>of</strong> the flapping motion and<br />

leaves therefor more space for further ideas to control each wing independently.<br />

However the additional gears increase the friction and the complexity and so also<br />

the efficiency and the weight respectively.<br />

To increase the robustness <strong>of</strong> the design, the flapping can be actuated according<br />

to the right side <strong>of</strong> figure 3.5. Instead <strong>of</strong> just actuating the wings, a more solid


25 3.2. Concepts for wing flapping<br />

Figure 3.4: Schematic drawing <strong>of</strong> concept A2<br />

Figure 3.5: Schematic drawing <strong>of</strong> concept B1 (left) and B2 (right)<br />

tube, where the wings can be inserted in, is moved. This tube can be attached via<br />

a rotational joint to the main structure where also the motor is attached at and<br />

gives so more stability to the flapping device.<br />

3.2.3 Concept C<br />

To reduce the number <strong>of</strong> the needed joints the actuation can be done by using a<br />

flexible part according to figure 3.6. The bending <strong>of</strong> the rod at the middle induces<br />

a motion at the wings. For the flexible part a material can be used which has a<br />

good flexibility and still has a enough high stability as carbon or titanium.<br />

For a brief inspection <strong>of</strong> the feasibility <strong>of</strong> this concept the theory <strong>of</strong> mechanics for<br />

calculating the bending line <strong>of</strong> a rod is used. The bending line can be calculated<br />

according to equation 3.4,<br />

d 2 w(x)<br />

dt 2<br />

= −My(x)<br />

EIy<br />

(3.4)<br />

whereas E is the modulus <strong>of</strong> elasticity <strong>of</strong> the used flexible material. The bending<br />

torque in the y-direction My and the moment <strong>of</strong> inertia in the y-direction <strong>of</strong> the<br />

rod Iy is calculated as follows<br />

<br />

My(x) =<br />

Iy = dh3<br />

12<br />

F x<br />

2<br />

F x<br />

2<br />

for 0 < x < b<br />

− F · (x − b) for b < x < 2b<br />

(3.5)<br />

(3.6)


Chapter 3. Concepts 26<br />

Figure 3.6: Schematic drawing <strong>of</strong> concept C<br />

The rod’s dimensions are specified by it’s width d and height h as shown in figure 3.7.<br />

Using the boundary conditions 3.7 and integrating equation 3.4 gives the maximal<br />

deflection at the midpoint between the two wing holdings needed to get the desired<br />

stroke amplitude <strong>of</strong> β=55 ◦ (equation 3.8 2 ).<br />

dw(0)<br />

dt = tan 55◦ , w(0) = 0 , dw(b)<br />

dt = 0 (3.7)<br />

F =<br />

w(b) =<br />

4EIy tan 55◦<br />

b2 −F b3<br />

12EIy<br />

+ b tan 55◦<br />

<br />

2b tan 55◦<br />

=⇒ w(b) =<br />

3<br />

Figure 3.7: Calculation <strong>of</strong> the bending line<br />

(3.8)<br />

Because the deflection <strong>of</strong> the rod needs a certain force to attain the desired<br />

stroke amplitude, it has to be checked if an actuator can be found, which generates<br />

enough force. Therefor the whole mechanism is modelled in a simplified way as<br />

presented in figure 3.8. It is important to notice that also the following calculations<br />

are just a rough approximation to check for the fundamental feasibility <strong>of</strong> this<br />

concept and if it has to be investigated more deeply. Also the forces acting on the<br />

wings and the wings itself are not included yet, as these forces can not be calculated<br />

exact enough and so just would blur the results.<br />

The behaviour <strong>of</strong> the bending rod can be modelled in a simple way as a spring with<br />

a point mass ms, which represents the mass <strong>of</strong> the link with length L. The spring<br />

constant c and the corresponding force Fc generated by the compressed or stretched<br />

2 matlab file bending line flex.m


27 3.2. Concepts for wing flapping<br />

spring for this arrangement is defined as<br />

c = 48EIy<br />

(2b) 3<br />

(3.9)<br />

Fc = c(y − y0) (3.10)<br />

whereas y0 is the length <strong>of</strong> the unloaded spring and is set as a first instance for<br />

δ = 0.<br />

Figure 3.8: Sketch for calculation <strong>of</strong> the dynamics<br />

Fs represents the force acting on the link and M the generated torque by the<br />

actuator. The maximal value for y, which is calculated in equation ?? is equal to<br />

the amplitude <strong>of</strong> the sinusodial motion. As L ≫ r the radius can be approximated<br />

as r ≈ w(b). Using the laws <strong>of</strong> conservation <strong>of</strong> the momentum for the link and the<br />

angular momentum for the rotating disc the following equations <strong>of</strong> motion can be<br />

derived:<br />

ms · d2 y<br />

dt 2 = Fs cos α − cy + cL cos α0 (3.11)<br />

θ · d2 δ<br />

dt 2 = M − rFs cos δ − α (3.12)<br />

for δ = 0 and θ is the inertia matrix <strong>of</strong> the rotating disc. Using<br />

equations 3.1 and assuming a constant angular speed ˙ δ the equations for ˙α, ¨α are<br />

where sin α0 = r<br />

L<br />

˙α = − ˙ δ · sin δ · Z −0.5 · r<br />

L<br />

¨α = − ˙ δ 2 · cos δ · Z −0.5 · r<br />

L − ˙ δ · sin δ · r<br />

L · −r2 ˙ δ sin 2δZ −1.5<br />

2L2 (3.13)<br />

(3.14)


Chapter 3. Concepts 28<br />

with Z = 1−r2 cos δ 2<br />

L2 .<br />

Derivating equation 3.2 with respect to time an expression for ¨y is obtained<br />

¨y = −r ˙ δ 2 − L¨α · sin α − L ˙α 2 · cos α (3.15)<br />

Using equations 3.15, 3.13 and 3.14 into equation 3.11 the force on the link can<br />

be calculated during one cycle (equation 3.16). Inserting it into equation 3.12 the<br />

needed torque for the bending is obtained.<br />

Fs = ms¨y − cy − cL cos α0<br />

cos α<br />

(3.16)<br />

For numerical values a carbon rod with dimensions <strong>of</strong> 0.1mm x 2mm and a<br />

modulus <strong>of</strong> elasticity <strong>of</strong> 110 ′ 000 N<br />

mm2 is used. The parameter b is chosen according<br />

to the estimated value <strong>of</strong> the body width <strong>of</strong> the hummingbird, which corresponds<br />

to the distance between the two wing mountings as mentioned in chapter 3.1. The<br />

results are presented in figure 3.9, which show the torque M needed during one cycle3 with the maximal value <strong>of</strong> slightly under 4mNm. The positive torque indicates that<br />

the actuator has to push the link, while negative torques represents the situations<br />

when the actuator has the break the motion <strong>of</strong> the link due to the reaction <strong>of</strong> the<br />

spring-like behaviour <strong>of</strong> the bended rod.<br />

Figure 3.9: Results <strong>of</strong> the force on the link and the needed torque during one<br />

flapping cycle<br />

It can be seen that with such dimensions for the carbon rod, an applicable<br />

actuator could be found, which is enough lightweight and still can bring up enough<br />

torque 4 . However if the thickness <strong>of</strong> the rod is increased to 0.2mm, already a much<br />

higher torque is needed and the size and weight <strong>of</strong> the actuator would grow too<br />

much. Another disadvantage is that the distance between both wing holdings can<br />

not be reduced much more, then again a higher force is needed to bend the flexible<br />

rod. Also the wings are not considered yet, which again increases the torque which<br />

has to be generated by the motor.<br />

Taking these aspects into account a working flapping device using this concept will<br />

not be guaranteed, for too heavy wings no real flapping motion could be produced,<br />

only the flexible part would bend withouth generating the desired motion for the<br />

wings.<br />

3 generated with dynamics flap rot const.m<br />

4 see for instance: www.faulhaber.com


29 3.3. Concepts for wing pitching<br />

3.2.4 Concept D<br />

All the above ideas induce a linear motion between the centered joint and the wing<br />

holding, because <strong>of</strong> the relatively high stroke amplitude. To get rid <strong>of</strong>f the linear<br />

motion a structure as shown in figure 3.10 could be used. The wing is attached<br />

similar as in concept B2 to a connector, which is attached to the main structure<br />

and can rotate about one axis, allowing to flap the wing in one plane. Between the<br />

connector and the actuation point three joints are arranged so that the link in the<br />

middle does not just move up and down, but rather adopts its orientation that the<br />

joint most proximal to the wing is routed on a circular trajectory and therefor does<br />

not induce a linear motion into the direction <strong>of</strong> the connector. The kinematics are<br />

Figure 3.10: Schematic drawing <strong>of</strong> concept D<br />

similar to those described in section 3.2.1. The only difference is that the parameters<br />

have changed places. Here y is determined, it is actuated in a sinusodial way, and<br />

the angle δ, which above described the state <strong>of</strong> the cycle is now the flapping angle<br />

(see figure 3.10). Note that δ with the additionally inserted joint not makes a<br />

whole cycle. By adapting the correct link lengths and the actuating amplitude,<br />

the maximum opening angle <strong>of</strong> 55 ◦ can be obtained. The general equations for the<br />

kinematics can therefor easily be taken out <strong>of</strong> equations 3.1 and 3.2. Hence also no<br />

exact sinusodial flapping motion is present, but using the same convention as above<br />

(L ≫ r) a good approximation can be found.<br />

However this would be a nice solution, this concept needs the most joints <strong>of</strong> the<br />

described concepts above. Also for implementing this concept later on for a real<br />

MAV could be difficult because the links and the joints have to be guided and<br />

supported to increase the stability <strong>of</strong> this arrangement.<br />

3.3 Concepts for wing pitching<br />

3.3.1 Active pitching<br />

As showed in chapter 2.2.2 the hummingbird controls the pitch angle (chord angle)<br />

<strong>of</strong> the wing approximately in a harmonic sinusodial motion for one flapping cycle.<br />

Therefor to mimic the wing motion an obvious solution would include also to control<br />

the pitch angle actively with the same actuator. As showed in chapter 3.2 an<br />

approximated sinusodial flapping trajectory could be produced. By optimizing the<br />

<strong>of</strong>fset between the sine wave <strong>of</strong> the flapping and the chord angle, setting up the<br />

desired pitch angle for every state <strong>of</strong> the whole cycle with the same rotary actuator<br />

should be possible. The remaining question, at which point to actuate the wing to<br />

set up the pitch angle has to be investigated.


Chapter 3. Concepts 30<br />

Actuating the wing’s trailing edge<br />

One idea could be to attach the trailing edge <strong>of</strong> the wing to the main body, where it<br />

is connected to the actuator and is moved according to the flapping cycle to achieve<br />

the desired pitch angle as showed in figure 3.11. As a first approximation roughly<br />

Figure 3.11: Actively adapting pitch angle using the trailing edge<br />

the values <strong>of</strong> the hummingbird’s flapping trajectory as can be seen in figure 2.12 are<br />

taken. Note that the studies base on the smaller rufous hummingbirds and not the<br />

giant hummingbird, for which reason the dimensions <strong>of</strong> the rufous hummingbirds<br />

[2] are taken to check the feasibility. The wrist elevation is taken as a cosine wave,<br />

Figure 3.12: Geometric sketch for calculations <strong>of</strong> the trailing edge motion<br />

while the trajectory <strong>of</strong> the chord angle has an <strong>of</strong>fset <strong>of</strong> about 190 degrees to it.<br />

According to the geometric drawing (figure 3.12) the following equation relates the<br />

position <strong>of</strong> the trailing edge <strong>of</strong> the wing to the position <strong>of</strong> the leading edge, which<br />

is described by the wrist elevation.<br />

yT = yL − c sin(α) (3.17)<br />

The resulting trajectories are shown in figure 3.13 5 . The left graph shows the<br />

actual motion <strong>of</strong> the leading edge, the trailing edge and the original chord angle.<br />

On the right side the cosine wave <strong>of</strong> the leading edge is shifted over the wave <strong>of</strong> the<br />

trailing edge to clarify the result. There it can be seen that the trailing edge is not<br />

moved totally harmonic. One half <strong>of</strong> the cycle is performed a little faster than the<br />

5 matlab file active pitch.m


31 3.3. Concepts for wing pitching<br />

Figure 3.13: Left: Trajectories <strong>of</strong> leading edge, traling edge and chord angle, Right:<br />

Shifted graph for the leading edge motion for comparison <strong>of</strong> the harmoinc behaviour<br />

<strong>of</strong> the trailing edge’s motion<br />

other. This anomaly increases for bigger wings and hence if the dimensions <strong>of</strong> the<br />

giant hummingbird are used the inaccuracy increases also. If this way <strong>of</strong> setting up<br />

the pitch angle is used, it has to be considered, that the wing has to be made <strong>of</strong> a<br />

stretchy material. If the actuation is not done absolutely exact, which is the case if<br />

only one actuator is used, the chord length will vary and induce stress on the wing.<br />

However a big advantage would be that the mechanism for flapping and pitching<br />

could be done in a seperate manner, if the motor lies between both edges <strong>of</strong> the<br />

wing.<br />

Actuating the wing’s leading edge<br />

As the chord angle describes a sinusodial trajectory as the flapping motion a more<br />

acurate way to control the pitch angle would be to actuate directly the leading edge<br />

rod <strong>of</strong> the wing as shown in figure 3.14. The pitch control can be done by just a<br />

Figure 3.14: Actively adapting pitch angle using the leading edge<br />

horizontal movement, with the same harmonic behaviour as the flapping motion,<br />

actuated directly on the wing rod <strong>of</strong> the leading edge at a distance t <strong>of</strong> the center


Chapter 3. Concepts 32<br />

<strong>of</strong> the rod. Of course the mechanical structure is not as easy as presented in figure<br />

3.14. As the amplitude for the chord angle is more than 100 degrees, the pin on the<br />

rod drifts away from the actuator tool. This needs a complex pitching mechanism<br />

to have the friction reduced. A simple but not so clean way to solve this, would<br />

be to add a spiral spring which pushes the pin constantly onto the actuator tool.<br />

However a more difficult problem would be, that the same rod is actuated for the<br />

flapping motion and also moves up and down. Hence this mechanism has to be as<br />

near as possible to the body <strong>of</strong> the MAV.<br />

Despite these design challenges, the advantage <strong>of</strong> this concept can be showed by<br />

calculating the kinematics. The needed amplitude s <strong>of</strong> the horizontal movement is<br />

expressed by equation 3.18.<br />

s = t · tan ɛ (3.18)<br />

By putting this amplitude into the same sinusodial motion as the flapping and<br />

adjusting the <strong>of</strong>fset, the simulated pitch angle can be obtained according to figure<br />

3.15 6 . It can be seen, that also here small errors occur. But as the sine wave for the<br />

chord angle only is an approximation and the real chord angle as showed in figure<br />

2.12 even more equals the simulated angle, this method seems to be more accurate<br />

than actuating the trailing edge.<br />

Figure 3.15: Simulated chord angle for horizontal actuation <strong>of</strong> wing rod<br />

3.3.2 Passive pitching<br />

A more simple way to adjust the pitch angle can be done by let the wing passively<br />

adapt it’s chord angle. This is possible as inertial and aerodynamic loads tend to<br />

decrease the angle <strong>of</strong> attack. However this would not be as accurate as controlling<br />

the angle actively. But as can be seen in [2] the angle <strong>of</strong> attack for the hummingbird<br />

flight stays more or less constant during one flapping cycle. Of course at the beginning<br />

<strong>of</strong> the upstroke and the downstroke the angle <strong>of</strong> attack will intuitively have<br />

the biggest error according to the desired value. But several other MAV already<br />

use this way for adjusting the pitch angle <strong>of</strong> the wing with success, for which reason<br />

6 matlab file active pitch2.m


33 3.3. Concepts for wing pitching<br />

this approach can be considered as suitable.<br />

Therefor two main ideas came up. The first one can be used if the wing rod is<br />

inserted into a tube, which is attached to the main structure and is actuated for<br />

flapping (see Concept B2 in chapter 3.2). The main connection between the wing<br />

rod and the tube is done with spiral springs as shown in the sketch in figure 3.16.<br />

If the flapping is performed, the wing tends to decrease the angle <strong>of</strong> attack and<br />

starts to rotate. As the spring induces a counter-torque to this rotation, a constant<br />

angle <strong>of</strong> attack could be obtained. Unfortunately there is no way for calculating a<br />

good enough approximation for the forces and torques acting on the wing. Hence<br />

the strength <strong>of</strong> the springs has to be identified experimentally with a test bench <strong>of</strong><br />

the flapping device and completely designed wings. Another approach is to include<br />

Figure 3.16: Sketch <strong>of</strong> general principle for passive pitching at the hinge<br />

the pitching mechanism already into the wings. The general idea is adopted <strong>of</strong><br />

Robert Wood’s Robotic Insect [24]. As shown in figure 3.17 the fibre <strong>of</strong> the wings<br />

could be designed using a sandwich-like structure. The middle part is made <strong>of</strong> a<br />

flexible material and is surrounded by a more stiff material. Right beneath the<br />

wing rod <strong>of</strong> the leading edge, which is actuated for flapping, the stiff material is<br />

removed, and lets the wing surface rotate around the leading edge. By investigation<br />

the maximal pitch angle can be adjusted by removing more or less <strong>of</strong> the stiff<br />

material. If the angle increases, a touching <strong>of</strong> the outher parts occur and blocks a<br />

further bending <strong>of</strong> the flexible material. Also the flexible material can be modelled<br />

Figure 3.17: Passive pitching done at the wings<br />

as a spring, wherefor like for the first approach, the correct adjustments have to be<br />

done experimentally. However this idea saves weight and reduces the complexity<br />

<strong>of</strong> the flapping mechanism. Another advantage is, that this approach for pitching<br />

the wing only depends on the wings and not on the mechanism which generates the<br />

flapping motion. In contrast the first approach, using springs, already presumes<br />

some construction elements for the attachement <strong>of</strong> the wing to the main body.


Chapter 3. Concepts 34


Chapter 4<br />

Evaluation<br />

4.1 Evaluation <strong>of</strong> concepts<br />

After generating different ideas for the flapping motion, these concepts will be evaluated<br />

according to several criteria described below, to chose the most suitable concept<br />

as a starting point for the CAD design. As not all exact dimensions and parameters<br />

for the above described concepts are present yet, this evaluation is done in a more<br />

qualitative way. Nonetheless this selection is important as it affects the whole future<br />

design process.<br />

4.1.1 Criteria<br />

The concepts are evaluated according to following critera:<br />

• Weight<br />

• Size<br />

• Robustness<br />

• Mechanical complexity<br />

• Expected power consumption<br />

• Accuracy<br />

Note that some criteria are overlapping and depend on each other. For instance a<br />

mechanism which is more complex and needs more joints, also in general consumes<br />

more power. The most important aspects for a MAV are the weight, as it has to<br />

be minimized or should left space for further weight reduction, and the power consumption,<br />

which also needs to be as low as possible. It would not make sense to<br />

integrate a flapping mechanism in a MAV, which consumes too much power and<br />

needs a bigger battery, which again increases the weight. However the expected<br />

power consumption can not yet be determined accurate enough, to get significant<br />

differences between the various concepts for the flapping motion, whereas mainly<br />

the number <strong>of</strong> joints and the generated friction is taken as an indicator for the<br />

performance in terms <strong>of</strong> power-saving.<br />

The general dimensions are already fixed at this stage (see chapter 3.1), for which<br />

reason the size more indicates the feasibility to minaturize the mechanism, but has<br />

less priority as the actual dimensions are considered to be small enough for a prototype<br />

MAV. Also the robustness plays an important role. The mechanism has to<br />

be stable enough to achieve a flapping frequency <strong>of</strong> 15Hz. Many links decrease the<br />

35


Chapter 4. Evaluation 36<br />

robustness too. To stabilize them, they need to be guided somehow on the main<br />

structure <strong>of</strong> the MAV, which would again increase the complexity and induce more<br />

friction. The accuracy is only used for evaluating the concepts <strong>of</strong> the pitch motion,<br />

which differ for each concept. Due to the fact that all concepts for the flapping<br />

motion base on the same general principle, all generate the similar wing beat trajectory<br />

as shown in chapter 3.2.1, for which reason this criterion is not used for the<br />

flapping concepts.<br />

Taking into account the above considerations an evaluation matrix can be generated<br />

with weights indicating the importance <strong>of</strong> each criterion:<br />

Criterion Weight<br />

Weight 6<br />

Size 1<br />

Robustness 4<br />

Mechanical complexity 5<br />

Expected power consumption 5<br />

Accuracy 3<br />

4.1.2 <strong>Flapping</strong> concepts<br />

Using the above matrix the concepts <strong>of</strong> chapter 3.2 are compared to each other and<br />

a rank is assigned (6 is the best and 1 is the worst), considering the facts described<br />

above and in chapter 3.2.<br />

Weight Size Robust. Mech.<br />

Complex.<br />

Concept A1 5 3 3 5 6<br />

Concept A2 4 2 4 4 5<br />

Concept B1 2 6 5 3 2<br />

Concept B2 1 6 6 2 2<br />

Concept C 6 1 2 6 3<br />

Concept D 3 4 1 1 4<br />

Power<br />

Consumpt.<br />

The outcome <strong>of</strong> this comparison is shown in the table below, where the total points<br />

and the rank (1 best, 6 worst) are presented.<br />

Total Points Rank<br />

Concept A1 100 1<br />

Concept A2 87 3<br />

Concept B1 63 4<br />

Concept B2 56 5<br />

Concept C 90 2<br />

Concept D 51 6<br />

As can be seen, concept A1 seems to be the most suitable to be used as a starting<br />

point and is therefor used as a guidance for the CAD design. Concept C also seems<br />

reasonable to be followed. Nonetheless it is neglected, due to the big uncertainty<br />

for the feasibility, how it was described in chapter 3.2.3.<br />

4.1.3 Pitching concepts<br />

This process is repeated for the concepts described in chapter 3.3 and again presented<br />

in tables below.


37 4.2. Expected weight<br />

Weight Size Robust.<br />

Active-Trailing Edge 2 2 2<br />

Active-Leading Edge 1 1 1<br />

Passive-Spring 3 3 3<br />

Passive-<strong>Wing</strong> 4 4 4<br />

Mech.<br />

Complex.<br />

Power<br />

Consumpt.<br />

Active-Trailing Edge 2 2 3<br />

Active-Leading Edge 1 1 4<br />

Passive-Spring 3 3 2<br />

Passive-<strong>Wing</strong> 4 4 2<br />

Total Points Rank<br />

Active-Trailing Edge 51 3<br />

Active-Leading Edge 33 4<br />

Passive-Spring 69 2<br />

Passive-<strong>Wing</strong> 90 1<br />

Accuracy<br />

For the pitching motion <strong>of</strong> the wing, the concepts for passive wing pitching are<br />

superior using this evaluation method. This is certainly the case, because the evaluation<br />

matrix is laid out in such a way, that the whole mechanism stays simple and<br />

is as lightweight as possible. If experimental data would be available, how more lift<br />

force with a more accurate pitching motion can be generated, the weight for the<br />

accuracy can be adapted. For instance, if the more complex and heavier mechanism<br />

for actively controlling the pitch motion <strong>of</strong> the wing, regains more lift force than<br />

the additional weight needed, controlling the wing’s pitch angle actively would be<br />

superior to the concepts for passively controlled wing pitch angles.<br />

However, for this first approach developing a flapping mechanism, passiv wing pitching<br />

is adequate.<br />

4.2 Expected weight<br />

Knowing on which concept to focus on, first speculations on the expected weight<br />

<strong>of</strong> the MAV can be made. Using the dimensions <strong>of</strong> chapter 3.1 the weights <strong>of</strong> the<br />

different parts can be estimated. Note that in the following only a rough approximation<br />

is done, which is based on an internet research <strong>of</strong> several suppliers <strong>of</strong> parts<br />

for model aircrafts 1 .<br />

The wings are assumed to have the structure out <strong>of</strong> carbon, which is covered with<br />

mylar as shown schematically in figure 4.1. Three carbon (1.55g/cm 3 ) rods with<br />

diameter <strong>of</strong> 2mm can be used as support to attach the 0.0005mm thick mylar sheet<br />

(7g/m 2 ) on it. With this design, one simple wing weights about 1.5g.<br />

As mentioned above two servos are needed for control purposes, whereas each one<br />

weights about 1g. Including the weight <strong>of</strong> the RC receiver (1g), the brushless DC<br />

motor (≈ 6g) and the battery (6.5g), the electronic payload measures about 15.5g.<br />

A reasonable approximation for the structure <strong>of</strong> the MAV, which includes the found<br />

concept for the flapping mechanism, would be about 10g.<br />

Therefor the expected weight <strong>of</strong> the MAV can be assumed to be ≈ 30g.<br />

1 for instance www.microbrushless.com


Chapter 4. Evaluation 38<br />

Figure 4.1: Structure <strong>of</strong> the wing<br />

4.3 Expected power consumption<br />

Using the expected weight <strong>of</strong> 30g, the needed mechanical power can be calculated.<br />

According to [1] the mechanical power can be expressed by following equation:<br />

<br />

W<br />

P = W<br />

(4.1)<br />

2ρSE<br />

where W = mg ≈ 0.29N is the total mass in expressed in [N], ρ = 1.29kg/m3 the<br />

air density and SE is the effective operational area or sweeping area <strong>of</strong> both wings.<br />

Usually the sweeping area can be considered to be about 70% <strong>of</strong> the circular disc S<br />

swept by both wings [1]. Rewriting equation 4.1 as follows<br />

<br />

W<br />

S<br />

P = W<br />

2ρ SE<br />

(4.2)<br />

S<br />

and inserting SE<br />

≈ 70%, W<br />

= 31.52 N<br />

we obtain for the power P = 0.393W .<br />

S S m<br />

According to [1] an efficiency coefficient for the hover ability has to be included. As<br />

the wing motion will be similar to the hummingbird’s wing motion, it is reasonable<br />

to take the same value for the coefficient ηH = 60%. Including the mechanical<br />

efficiency <strong>of</strong> the mechanism ηM<br />

according to equation 4.3.<br />

the total needed mechanical power is calculated<br />

Ptot = P<br />

ηHηM<br />

(4.3)<br />

The efficiency coefficient for the mechanism ηM can be calculated by summing up<br />

the efficiency <strong>of</strong> the joints (≈ 80% for each joint) and the efficiency <strong>of</strong> the motor,<br />

which is typically around 60% for the chosen brushless DC motor 2 . Considering the<br />

above chosen concept, the total needed mechanical power is about 1.36 W.<br />

With this estimated value, the range for finding an appropriate motor can be constrained.<br />

However, as the objective <strong>of</strong> this project is not to build the whole robot,<br />

this calculation just strengthens the evidence that such a motor can be found, which<br />

can fulfill the power and the weight requirements. Also the electrical power consumption<br />

can then be calculated, when the best matching motor could be found,<br />

which allows to estimate the time the MAV could fly, without recharging the batteries.<br />

2 see for instance www.wes-technik.de


Chapter 5<br />

CAD Design<br />

The ’winner’ concept <strong>of</strong> chapter 4.1 is now converted into a test bench, using 3D<br />

CAD s<strong>of</strong>tware. As the forces on the wings are still unknown and just can be identified<br />

experimentally in a correct manner on a finished flapping mechanism, no exact<br />

stability calculations can be done for the different parts <strong>of</strong> the flapping device. Also<br />

those forces highly depend on the wing design. Therefor it is considered to make<br />

the flapping mechanism as robust as possible, trying to copy the kinematic <strong>of</strong> the<br />

hummingbird and neglect weight constraints. Hence, the weight reduction has to<br />

be done later on, when also the wings are completely designed and measurements<br />

are done using the designed test bench, but this is not part <strong>of</strong> this project.<br />

5.1 Overview<br />

The resulting mechanism can be seen in figure 5.1. It has a width <strong>of</strong> about 12cm,<br />

a length <strong>of</strong> 14cm and a height <strong>of</strong> about 8cm. All the parts needed for flapping are<br />

attached to 2mm thick aluminium plate, which is screwed on two shorings. These<br />

shorings have an inclination <strong>of</strong> 10 ◦ which corresponds to the stroke plane angle<br />

with respect to the horizontal <strong>of</strong> the hummingbird’s wing motion. A DC-motor<br />

with a planetary gearhead (reduction ratio 3.71:1) is attached below the plate and<br />

transmits the generated torque to a link, which starts to rotate. The datasheets<br />

for the motor and the gearhead can be seen in the appendix. This rotating link<br />

is connected via a pin to another link, which is guided on the other end in such a<br />

way, that this end point is always centered between the two wing joints. Therefor<br />

two ball bearings are needed to allow the transmission link to rotate freely. Due<br />

to the circular motion <strong>of</strong> the rotating link, the transmission link is moved forward<br />

and backward, which produces the same kinematic pattern as calculated in chapter<br />

3.2.1. This movement is transmitted via another pin to two wing attachements,<br />

which are placed on wing joints, playing the role <strong>of</strong> center <strong>of</strong> rotation for the movement<br />

<strong>of</strong> these attachements.<br />

Due to time constraints, all parts were printed, except the ball bearings, the aluminium<br />

plate and the pins, which were made <strong>of</strong> steel. Therefor the dimensions <strong>of</strong><br />

the links had to be chosen big enough to assure the stability. However the links are<br />

supported in the vertical direction only on one point, the connecting point to the<br />

motor, which has to carry both links. Therefor the connection at the ball bearings<br />

have to be very tight, to ensure that the bending moments, caused by gravity, are<br />

as small as possible.<br />

39


Chapter 5. CAD Design 40<br />

Figure 5.1: Overview <strong>of</strong> resulting mechanism<br />

5.2 Transmission <strong>of</strong> motor torque<br />

To transmit the torque from the motor the following assembly is used (see figure<br />

5.2). The rotating link has a fork like structure, where the motor shaft is in-between,<br />

with the notch aligned at the face <strong>of</strong> the link. A small part, which fits between the<br />

two legs <strong>of</strong> the link, is pushed on the motor shaft and attached to the link using<br />

a screw. The dimensions were chosen such that, by inserting the screw, the small<br />

part is pressed against the motor shaft. However this is not absolute necessary, as<br />

already the notch secures the connection. Nonetheless this force fitting again gives<br />

a more secure force transmission.<br />

Figure 5.2: Connection <strong>of</strong> motor to rotating link


41 5.3. Joints<br />

5.3 Joints<br />

Assuring that the transmission link and the wing attachements can follow the considered<br />

motion, the joints are the most important parts. However this part could<br />

not be solved perfectly. Pushing the center joint forward and backward, the distance<br />

between it and the wing joints changes. As the stroke amplitude is wanted<br />

to be 110 ◦ , the variation <strong>of</strong> the distance is a bigger problem and can not be solved<br />

with this general structure. Therefor either the center joint can be optimized, using<br />

ball bearings to reduce friction, and at the wing joints a loose connection is needed,<br />

or the wing joints are optimized, but then again the center joint will cause more<br />

friction.<br />

For increasing the stability, the second approach is better, as the wings then are<br />

guided more accurately, for which reason the center joint is a more loose connection.<br />

Center joint<br />

Hence the connection <strong>of</strong> the transmission link to the wing attachements is done<br />

according to figure 5.3. The pin is guided through big slots in the wing attachements.<br />

During the forward and backward motion <strong>of</strong> the pin, the attachements are pushed<br />

to follow its way. Due to the movement <strong>of</strong> the pin in the slots friction occurs. To<br />

optimize this design the wing attachement has to be made <strong>of</strong> different materials.<br />

The touching area with the pin could be <strong>of</strong> a material which has low friction, while<br />

the rest should be made <strong>of</strong> a stiff and lightweight material. Nonetheless, this design<br />

is still acceptable as a first approach, as the friction still is pretty small.<br />

To have the center joint stay always centered, it has to be guided. This is done by<br />

a slot in the aluminium plate, where the pin is moved in. To decrease the friction a<br />

ball bearing is glued to the pin. Note that the diameter <strong>of</strong> the ball bearing is just<br />

slightly smaller than the width <strong>of</strong> the slot, to assure that the ball bearing does not<br />

brake itself, when from both sides a touching occurs.<br />

Figure 5.3: Design for guiding the center joint


Chapter 5. CAD Design 42<br />

<strong>Wing</strong> joint<br />

As described above, the wingjoint is optimized. Therefor two ballbearings are used.<br />

One inside the structure <strong>of</strong> the wing joint, the other lying above. On the upper ball<br />

bearing the wing attachement is arranged in such a way that only the inner part <strong>of</strong><br />

the ball bearing is touched. The outer part is then connected only to the motionless<br />

structure <strong>of</strong> the wing joint. To attach the wings, holes with a diameter <strong>of</strong> 2mm and<br />

Figure 5.4: Assembly <strong>of</strong> wing joint<br />

a depth <strong>of</strong> 20mm are made into the wing attachements as can be seen in figure 5.5.<br />

To increase the stability, the structure is thickened around the holes.<br />

Figure 5.5: Structure <strong>of</strong> the wing attachments


Chapter 6<br />

Conclusion<br />

The resulting mechanism, designed according to the kinematic requirements <strong>of</strong> the<br />

desired flapping motion, is acceptable. Due to time constraints no exact measurements<br />

with the manufactured mechanism could be done. Also no information <strong>of</strong> the<br />

generated lift force can be deducted, as this is only possible to measure when the<br />

wing designing is done, which includes the adjustment <strong>of</strong> the wing’s flexibility for<br />

the pitching. And at this time, the mechanism consists <strong>of</strong> a mock-up <strong>of</strong> the wings.<br />

However it can already be seen, that the flapping motion follows the kinematic<br />

pattern as described in chapter 3.2.1, which is a good approximation for the hummingbird’s<br />

flapping. Also first tries showed, that the flapping frequency <strong>of</strong> 15Hz is<br />

reachable, although still many improvements can be done.<br />

The biggest improvements can be done for instance at the center joint. There is still<br />

relatively much friction present, which could be reduced. Also the guidance <strong>of</strong> the<br />

center joint leaves much space for enhancement. Nonetheless the strongly varying<br />

distance between the center joint and the wing joint, can not be eliminated with<br />

the chosen concept, for which reason it can also be considered to choose another<br />

approach for the design, which reduces this weakness. Another reasonable concept,<br />

would be to pitch the wing actively using the same actuator as for the flapping (see<br />

chapter 3.3.1), in the case the passive pitching would fail to deliver enough lift force.<br />

But most <strong>of</strong> these investigations can only be done after the next steps, for instance<br />

the wing design, are finished. With appropriate wings, also exact measurements<br />

can be done and the generated lift force can be discovered. Using a high speed<br />

camera, a more exact comparison <strong>of</strong> the resulting motion and the hummingbird’s<br />

wing motion can be done. Further, if these tasks show good results, implementing<br />

this motion into a real MAV would be the next step to the final objective <strong>of</strong> a flying<br />

flapping wing MAV.<br />

Concluding, as this project is the first step in the direction <strong>of</strong> designing a <strong>Flapping</strong><br />

<strong>Wing</strong> MAV, many useful informations and experience could be gathered, which for<br />

sure support further investigations into this direction.<br />

43


Chapter 6. Conclusion 44


Appendix A<br />

Motor datasheet<br />

45


Appendix A. Motor datasheet 46


Appendix A. Motor datasheet 48


Bibliography<br />

[1] Akira Azuma. The biokinetics <strong>of</strong> flying and swimming. Second Edition, 2002.<br />

[2] Bret W. Tobalske; Douglas R. Warrick; Christopher J. Clark; Donald R.<br />

Powers; Tyson L. Hedrick; Gabriel A. Hyder; Andrew A. Biewener. Threedimensional<br />

kinematics <strong>of</strong> hummingbird flight. The Journal <strong>of</strong> Experimental<br />

Biology 210, 2368-2382, 2007.<br />

[3] J. M. Birch and M. H. Dickinson. Spanwise flow and the attachment <strong>of</strong> the<br />

leading-edge vortex on insect wings. Nature, 412, 729-733, 2001.<br />

[4] Kenneth P. Dial Bret W. Tobalske, Wendy L. Peacock. Kinematics <strong>of</strong> flapbounding<br />

flight in the zebra finch over a wide range <strong>of</strong> speeds. The Journal <strong>of</strong><br />

Experimental Biology 202, 1725–1739, 1999.<br />

[5] Sharon M. Swartz; Jose Iriarte-Diaz; Daniel K. Riskin; Arnold Song; Xiaodong<br />

Tian; David J. Willis; Kenneth S. Breuer. <strong>Wing</strong> structure and the aerodynamics<br />

basis <strong>of</strong> flight in bats. In 45th AIAA Aerospace Sciences Meeting and<br />

Exhibit, AIAA Paper 2007-42, Reno, Nevada, 2007.<br />

[6] Xiaodong Tian; Jose Iriarte-Diaz; Kevin Middleton; Ricardo Galvao; Emily<br />

Israeli; Abigail Roemer; Allyce Sullivan; Arnold Song; Sharon Swartz; Kenneth<br />

Breuer. Direct measurements <strong>of</strong> the kinematics and dynamics <strong>of</strong> bat flight.<br />

Bioinspiration & Biomimetics 1 (2006) S10-S18, 2006.<br />

[7] MICHAEL H. DICKINSON. The effects <strong>of</strong> wing rotation on unsteady aerodynamic<br />

performance at low reynolds numbers. The Journal <strong>of</strong> Experimental<br />

Biology 192, 179–206, 1994.<br />

[8] C. P. ELLINGTON. The novel aerodynamics <strong>of</strong> insect flight: Applications to<br />

micro-air vehicles. The Journal <strong>of</strong> Experimental Biology 202, 3439–3448, 1999.<br />

[9] A. Roland Ennos. The kinematics and aerodynamics <strong>of</strong> the free flight <strong>of</strong> some<br />

diptera. The Journal <strong>of</strong> Experimental Biology 142, 49-85, (1989.<br />

[10] et al. F. T. Muijres. Leading-edge vortex improves lift in slow-flying bats.<br />

Science 319, 1250, 2008.<br />

[11] Stefan Gisler. Unsteady aerodynamics and control issues in flapping flight.<br />

Studies on mechatronics, Autonomous Systems Lab (ASL), Swiss Federal Institute<br />

<strong>of</strong> Technology Zurich (ETH), 2008.<br />

[12] H. Liu and K. Kawachi. A numerical study <strong>of</strong> insect flight. Journal <strong>of</strong> Computational<br />

Physics, 146, 124-156, 1998.<br />

[13] Y. Lu and G. X. Shen. Three-dimensional flow structures and evolution <strong>of</strong> the<br />

leading-edge vortices on a flapping wing. Journal <strong>of</strong> Expermental Biology, 211,<br />

1221-1230, 2008.<br />

49


Bibliography 50<br />

[14] Werner Nachtigall. Vogelflugforschung in deutschland. Journal <strong>of</strong> Ornithology.<br />

125, 157-187, 1984.<br />

[15] Ulla M. Norberg. Allometry <strong>of</strong> bat wings and legs and comparison with bird<br />

wings. Philosophical Transactions <strong>of</strong> the Royal Society <strong>of</strong> London. Series B,<br />

Biological Sciences, Vol. 292, No. 1061 (Jun. 10, 1981), pp. 359-398, 1981.<br />

[16] Fritz-Olaf Lehmann; Simon Pick. The aerodynamic benefit <strong>of</strong> wing–wing interaction<br />

depends on stroke trajectory in flapping insect wings. The Journal<br />

<strong>of</strong> Experimental Biology 210, 1362-1377, 2007.<br />

[17] Douglas R. Warrick; Bret W. Tobalske; Donald R. Powers. Aerodynamics <strong>of</strong><br />

the hovering hummingbird. Nature 435 1094-1097, 2005.<br />

[18] U. M. Norberg; J. M. V. Rayner. Ecological morphology and flight in bats<br />

(mammalia: Chiroptera): <strong>Wing</strong> adaptions, flight performance, foraging strategy<br />

and echolocation. Phil. Trans. R. Soc. Lond. B 316, 335-427, 1987.<br />

[19] Sanjay P. Sane. The aerodynamics <strong>of</strong> insect flight. The Journal <strong>of</strong> Experimental<br />

Biology 206, 4191-4208, 2003.<br />

[20] Rosalyn Sayaman Steven N. Fry and Michael H. Dickinson. Unsteady mechanisms<br />

<strong>of</strong> force generation in aquatic and aerial locomotion. Amer. Zool., 36,<br />

537-554, 1996.<br />

[21] Rosalyn Sayaman Steven N. Fry and Michael H. Dickinson. The aerodynamics<br />

<strong>of</strong> hovering flight in drosophila. The Journal <strong>of</strong> Experimental Biology 208,<br />

2303-2318, 2005.<br />

[22] Weis-Fogh T. Quick estimates <strong>of</strong> flight fitness in hovering animals, including<br />

novel mechanisms for lift production. The Journal <strong>of</strong> Experimental Biology 59,<br />

169-230, 1973.<br />

[23] van den Berg; Coen; Ellington; Charles P. The three-dimensional leading-edge<br />

vortex <strong>of</strong> a hovering model hawkmoth. Philosophical Transactions: Biological<br />

Sciences 352, 329-340, (1997.<br />

[24] R. J. Wood. The frst take<strong>of</strong>f <strong>of</strong> a biologically inspired at-scale robotic insect.<br />

IEEE Transcactions on Robotics, 24:341-347, 2008.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!