See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/267683178

Bio-inspired design of flapping-wing micro air

vehicles
ARTICLE AUGUST 2005

CITATIONS

DOWNLOADS

VIEWS

23

61

38

4 AUTHORS, INCLUDING:
Kevin Jones

Max Platzer

Naval Postgraduate School

Naval Postgraduate School

77 PUBLICATIONS 864 CITATIONS

189 PUBLICATIONS 1,495 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Max Platzer

Retrieved on: 10 August 2015

THE AERONAUTICAL JOURNAL

AUGUST 2005

385

Bio-inspired design of flapping-wing

micro air vehicles
K. D. Jones
C. J. Bradshaw
J. Papadopoulos
M. F. Platzer
Department of Mechanical and Astronautical Engineering
Naval Postgraduate School
Monterey, CA
USA

ABSTRACT
In this paper the development and flight testing of flapping-wing
propelled, radio-controlled micro air vehicles are described. The
unconventional vehicles consist of a low aspect ratio fixed-wing
with a trailing pair of higher aspect ratio flapping wings which flap
in counterphase. The symmetric flapping-wing pair provides a
mechanically and aerodynamically balanced platform, increases
efficiency by emulating flight in ground effect, and suppresses stall
over the main wing by entraining flow. The models weigh as little as
11g, with a 23cm span and 18cm length and will fly for about 20
minutes on a rechargeable battery. Stable flight at speeds between 2
and 5ms1 has been demonstrated, and the models are essentially
stall-proof while under power. The static-thrust figure of merit for
the device is 60% higher than propellers with a similar scale and
disk loading.

NOMENCLATURE
c
Cp
Ct
f
k
L/D
U

aerofoil chord length

coefficient of power
coefficient of thrust
frequency in Hz
reduced frequency, 2fc/U
lift over drag ratio
freestream velocity

x
z

streamwise coordinate
cross-stream coordinate
angle-of-attack in degrees
Propulsive efficiency, Ct /Cp
Phase angle in degrees

1.0 INTRODUCTION
The brilliant success of organisms which achieve flight using
flapping-wing propulsion has been an inspiration to humankind for
hundreds if not thousands of years. This has led to the, perhaps
erroneous, belief that flapping-wings are superior to other forms of
propulsion. The justification for this belief is that nature has chosen
flapping-wings through natural selection. Of course this argument is
rather naive, as it ignores the initial conditions and constraints of the
evolutionary process. In truth, one does not find many rotating parts
in nature, and therefore a reasonable argument is that nature did not
select flapping wings over propellers, but rather propellers were
excluded from the process entirely.
However, there are several arguments supporting flapping-wing
propulsion which do hold merit. For example, rotary propellers
generate torque and create helical slipstreams, both of which degrade
vehicle performance and handling, particularly at low speeds.
Additionally, flapping-wings may easily have a larger actuator area,

Paper No. 2950. Manuscript received 11 January 2005, accepted 18 July 2005.

386

THE AERONAUTICAL JOURNAL

AUGUST 2005

Figure 1. NPS flapping-wing MAV.

Figure 2. Frank Kiesers winning X-wing flapper.

and therefore a lower actuator loading, which may lead to increased

efficiencies. On the other hand, flapping-wings have inherent
mechanical losses and increased structural loads. The structural and
inertial requirements likely limit flapping-wing propulsion to smaller
scales, however, at these smaller scales low Reynolds number effects
further complicate the problem, for both conventional and flappingwing propellers.
Many have argued about the relative merits of fixed, flapping or
rotary-wing design, see for example Woods et al(1). While most in
the micro air vehicle (MAV) community have concentrated on fixedwing aircraft, the capabilities of flapping-wing designs in nature,
which can hover, maneuver in tight spaces, and even land on a
ceiling, greatly exceed anything humankind has been able to
produce, and this alone is motivation enough to pursue flappingwing flight.
The creativity, ambition and persistence of countless enthusiasts
have led to a variety of successful flapping-wing flyers over at least
the past 130 years. These designs are often classified as either
biomimetic (a design which mimics nature) or biomorphic or bioinspired (a design which is inspired by nature, but is not a copy of
nature). The configuration developed in this investigation, shown in
Fig. 1, clearly falls into the second category.

building rubberband powered ornithopters for more than a century,

and to date, the greatest endurance has been achieved by something
that did not look at all like a bird, using instead a biplane flapping
mechanism, as shown in Fig. 2. The biplane or X-wing canard
flapper, built in 1985 by Frank Kieser, set a new record for indoor
free flight endurance(3). Variations of this design still hold the
record. While the model does not appear to look like anything in
nature, in fact, the flapping of its wing-pairs is much like a bird
flying in ground effect.
Following a similar path, a collaboration between SRI and the
University of Toronto produced the Mentor, reportedly the first
flapping-wing aircraft to hover under its own power(4). Like Frank
Kiesers design, the Mentor is designed to capitalise on the WeisFogh or clap-and-fling effect, and seems to be designed primarily for
hovering flight a flapping-wing helicopter of sorts, although
forward flight has also been demonstrated. Neither design looks
anything like a bird or insect, but many birds and insects make use
of the Weis-Fogh effect and were an inspiration for these designs.
We have found it instructive to try to discern which aspects of
animal morphology evolved as a result of organic and/or environmental limitations, and which aspects are performance enhancing.
While this is not always simple, one approach we have found useful
is to look at how animal behavior is used to compensate for design
limitations. For example, birds have evolved with a single pair of
wings, so configurations such as the model shown in Fig. 2 are not
possible. However, the fact that birds fly low over the ocean
indicates that birds have indeed discovered the benefits of flapping
in ground effect, a behaviour that makes them aerodynamically
similar to the X-wing configuration. The exploitation of ground
effect is one of the behavioral adaptations which are utilised in this
investigation.
For the present design, much of the development process has been
documented over the years in AIAA papers and other publications. A
fairly detailed summary of our past work can be found in Jones et al(5)
and Jones and Platzer(6), with references to most of our other publications. In this paper the history is briefly summarised, with more details
provided on progress made since the earlier publications, and some
summarising remarks on what the future might bring.

1.1 Biomimetic designs

Flapping-wing model aircraft date back at least to 1874, when
Alphonse Penaud built a rubber-band powered ornithopter. Since
then, with just a few exceptions, almost all successful flapping-wing
aircraft have been biomimetic, at least for the flapping-wings.
Probably the most well known biomimetic flapper in the aerospace
community is AeroVironments Microbat. Their philosophy was to
take a functional rubberband powered design and substitute an
electrical motor drive-train for the rubberband and add a radio for
control(2). They make an interesting note, stating that the specific
energy of the rubberband is comparable to that of the
motor/gearbox/battery/DC-DC-converter assembly. In 2003, a
version of the Microbat using a two-cell Lithium-polymer (Li-poly)
battery, three channel radio, and with a 23cm span and 14g total
weight, reportedly made a flight of about 25 minutes.
1.2 Biomorphic designs
Projects like the Microbat clearly demonstrate that biomimicry can
lead to success, but one is left to ponder whether or not there is a
better way. For example, model airplane enthusiasts have been

2.0 DESIGN PHILOSOPHY

As previously mentioned, much can be learned from observations of
animal flight, but care must be taken to determine which features
have evolved to enhance performance, and which were forced due to
organic or environmental constraints. For example, a bird flaps its

JONES ET AL

BIO-INSPIRED DESIGN OF FLAPPING-WING MICRO AIR VEHICLES

387

Experimental schematic.

Static dye indicates separation

Figure 3. Predicted benefits of ground effect.

Figure 4. Thrust measurements in ground effect.

wings such that the flap-amplitude varies along the span. This does
not appear to be an optimal arrangement, but the bird does not have
an alternative. We would prefer to flap the wing with constant
amplitude to produce thrust from the root section of the wing, and to
provide for a more efficient span loading.
In the early years of our research, significant effort was directed
toward simulations of fairly basic flapping-wing mechanics in an
effort to better understand the complicated unsteady flow
phenomena, and with the hope of developing numerical tools to aid
in a design optimisation methodology. Unsteady panel methods were
developed in conjunction with flow visualisation and laser Doppler
velocimetry (LDV) measurements in wind and water tunnels(7-10),
and several significant conclusions were made.
First, the benefit of flapping in ground effect was quantified, both
numerically and experimentally, as shown in Figs 3 and 4. For these
simulations, an airfoil was plunged sinusoidally at zero angle of
attack, with the specified frequency, and with plunge amplitude
0.4c. The ground-effect simulations were modeled using an image
airfoil below ground, with a mean separation of 14c, corresponding
to a mean distance from the ground of 07c. The experimental

Flapping dye indicates eattachment

Figure 5. Flow reattachment due to flow entrainment.

measurements were performed on a biplane model, where the two

wings flapped in counterphase, essentially duplicating the panelcode model emulating ground effect through symmetry. A 5 to
10% increase in both the thrust coefficient and propulsive efficiency
were predicted. Additionally, the agreement between experiment and
the panel code were quite good, particularly for the low experimental
Reynolds numbers (less than 5 104).
A secondary benefit of the biplane arrangement was that the
model, as a whole, was dynamically balanced, minimising vibrations
of the apparatus. While dynamic balancing was convenient for
testing in a wind-tunnel, it appeared to be even more advantageous
for a flying model. Flapping a single wing in flight would result in
oscillations of the fuselage work done with no gain in performance. However, by flapping two wings in counterphase, the
fuselage would remain steady, and all the work would be used to
move components which directly produce thrust.
It is likely that birds have evolved to compensate for this
imbalance. Their neck is highly articulated, such that their head is
inertially stable for improved vision, and their body and tail most
likely create additional thrust as they oscillate in opposition to the
wing motion. However, it is doubtful that any man-made
ornithopters benefit from fuselage oscillations, as this requires a
level of sophistication beyond our current capabilities.
Another aerodynamic phenomenon we wished to exploit was the
ability to suppress flow separation using the flapping wings.
Dynamic stall delay due to oscillatory pitching is a fairly well known
phenomenon, of particular interest to rotary-wing engineers(11,12).
However, a lesser known application is the use of a flapping wing
downstream of a larger airfoil, using the favorable pressure gradient
ahead of the flapping wing to suppress flow separation on the larger
wing. Water-tunnel experiments demonstrated this phenomenon for
flow over a backward-facing step(8), and for separation control of
flow over several blunt trailing edge airfoils(9,13), as shown in Fig. 5.
In the upper image the general experimental setup is shown, illustrating a large, stationary airfoil with a cylindrical leading edge and a
rather abrupt cusped trailing edge. The airfoil thickens slightly along
the chord to keep the boundary layer thin. Dye is emitted from a
small hole just upstream of the cusped trailing edge region, where
the flow is expected to be attached. Following the stationary wing is
a small wing which may be flapped. The lower images show experimental snapshots. In the left image the trailing wing is stationary,

388

THE AERONAUTICAL JOURNAL

AUGUST 2005

Figure 8. Aeroelastic cambering wings.

Figure 6. 15cm length/span MAV propulsion model.

Figure 9. Static performance of cambering wings.

Figure 7. Performance of the propulsion model.

and the flow completely separates at the onset of the recompression.

In the right image, with the trailing wing flapping, the dye indicates
that the flow has reattached.
It was hoped that a similar success might be realised on a
flapping-wing MAV. This facet is of particular interest for low speed
MAVs, as the flight Reynolds numbers may be quite low, on the
order of 2 104, where the flow is laminar, and separation is likely.
These considerations led us to the configuration shown in Fig. 1,
with a biplane-pair of trailing wings, flapping in counterphase, coupled
with a large fixed wing located just upstream of the flapping wings.

3.0 DESIGN METHODOLOGY

As previously mentioned, the early years of our research were
focused on developing simulation software which might be used in
an optimisation algorithm to design an efficient flapping-wing
MAV. Unfortunately, what our research demonstrated was that the
low Reynolds numbers led to flow physics which could not be
adequately predicted using inexpensive methods such as our panel
code. Simulations with a two-dimensional unsteady Navier-Stokes
solver provided much better agreement with experimental measurements(10,14), but were far too costly to be embedded as part of

an iterative design process. Therefore, in the end we had to resort to

a more hands-on or heuristic approach, using failures and successes
to drive the evolutionary design process.
3.1 Mechanics of thrust generation
One of the first lessons to be learned was simplicity in design. On a
large model complicated mechanisms could be developed to add
several degrees of freedom to a flapping mechanism. However,
working on a MAV-scale, size and weight issues dominated, and the
flapping mechanism had to be reduced to a single active degree of
freedom plunge. Test models, like the one shown in Fig. 6 were
assembled, allowing for the direct measurement of thrust, and it
quickly became apparent that a pitch degree of freedom was
necessary. To keep things mechanically simple, this was done
passively by attaching the flapping wings to the flapping mechanism
with a flexible joint so that they were able to pitch aeroelastically.
Typical performance from the models is shown in Fig. 7.
Inspired by Cylinder et al(15), a second passive degree of freedom
was attempted camber. The model is shown in action in Fig. 8,
with the flash-frozen wings shown at mid-stroke as they pulled away
from each other. An elastic angle of attack of about 30 is visible, as
well as the fully cambered wing. The measured performance of the
cambering wings is shown in Fig. 9. While the camber provided
excellent and predictable performance for lower frequencies, at
higher frequencies the wings did not have enough time to change

JONES ET AL

BIO-INSPIRED DESIGN OF FLAPPING-WING MICRO AIR VEHICLES

389

Static

Figure 10. First fixed/flapping-wing configuration.

camber, and the performance degraded to that of the flat-plate wings.

With some additional work, perhaps this too could be resolved, but
due to the added weight and complexity of the cambering wings, for
the time being they were eliminated from the design.
All of these early test models used tiny, geared stepping motors,
which required large external power supplies and controllers. While
these drive systems were desirable for wind-tunnel models, they
were not suitable for flying models, and they were replaced by
brushed DC motors with custom gearboxes for the following
models.
3.2 Lift generation
The next step was the integration of the flapping-wing pair with a
fixed leading wing. The first test model, shown in Fig. 10, still
conformed with DARPAs 15cm square size criteria. This model
was tested in the wind-tunnel, and using flow visualisation with a
smoke wire, provided the first evidence of separation control using
the flapping wings. This is illustrated in Fig. 11, where the wings are
not flapping on the upper image, with the flow separating at the
leading edge, and on the lower image, with the wings flapping, the
flow appears to have reattached.
Tethered flight on a radial-arm test stand indicated that speeds on
the order of 2 to 3ms1 were possible while generating enough lift to
support the 7g weight. At that time commercial-off-the-shelf
(COTS) batteries and radio gear added up to a minimum of about
20g, so untethered flight was not feasible.
3.3 Energy storage
We are fortunate that some of the critical components we use are
also critical to the cell-phone industry, and hence performance
improvements have been tremendous; specifically, batteries, motors
(the motors that vibrate pagers and cell-phones), and DC-DC
converters to step-up a low voltage.
Battery technology has been driven by the cell-phone industry at
an extreme pace. Five years ago, about the best solution was a
50mAh nickel cadmium cell, providing 006Wh in a 35g cell an
energy density of about 17Wh/kg. The rechargeable Li-poly cells we
use now have a capacity of 135mAh and can provide 05Wh.
Weighing a mere 3g, they have an energy density of about
170Wh/kg, or ten times what the NiCd cell provided.

Flapping
Figure 11. Separation control at high AOA.

3.4 Motor/gear assembly

The motors are currently one of the weakest links. While the cellphone industry has pushed hard to make small, light and cheap
motors, they have done little to improve the efficiency. The motors
we used (Didel MK06L-10 or MK06-04.5) weighed about 1.3 to
15g, and used a two-stage gear system. For the first models we had
to fabricate our own gear systems, but now we use gear assemblies
built for the indoor model aircraft hobbyist (Didel 6R21). Using a
miniature dynamometer we measured the efficiency of the
motor/gear assembly, and at flight loads it was below 25%.
On the other hand, the motor/gear assemblies were very light,
about 22 to 25g, and they were able to deliver more than 0.4W of
shaft power, a power density of 180W/kg, which is quite good for
this scale.
3.5 Power conversion
A variety of motors were available, with nominal voltages between
08 and 24V. The first flying models built used motors designed for
12V with a rated power output of about 0.14W. In order to achieve

390

THE AERONAUTICAL JOURNAL

AUGUST 2005

3.7 Vehicle sizing

The next step in the design procedure was to size the configuration
for the required payload, in this case, just the propulsion and control
system. The components we used on our first model were slightly
heavier than those listed above, adding up to about 925g for a
battery, converter, drive system and receiver. From past building
experience, we estimated the weight of the structure to be about 5g,
yielding a total weight of just over 14g.
While little experimental data was available for low aspect-ratio
wings in the 2 104 to 5 104 Reynolds number range, papers by
Laitone(16) and Torres and Mueller(17) suggested that lift coefficients
on the order of about 06 were possible with an L/Dmax of around 5
on an aspect-ratio 2 wing. This indicated that a wing area of about
006m2 was needed.
Figure 12. First flying model.

4.0 DESIGN EVOLUTION

the power required to fly the MAV, we needed to increase the

power output considerably, and this was done by running the
motor at a much higher voltage. A miniature DC-DC step-up
circuit was assembled to bump the 30 to 42V from the battery up
to a regulated 5V. The circuit was based on an IC designed for
cell-phone type devices, but was miniaturised as much as possible,
with some sacrifice in efficiency. The circuit weighed about
0.35g, and produced a regulated 5V supply up to about 05A.
With some of the lower resistance motors now available, the
power converter is not required; although the flight performance
degrades somewhat as the battery life is exhausted and the voltage
drops.
3.6 Avionics
The radio gear is COTS hobbyist equipment, designed for parkflyers and indoor model aircraft. The three channel receiver
(Slowfly RFFS-100) weighs about 2g, slightly less if you remove
the bulky connector pins it comes with, and it includes an
electronic speed control and drivers for two magnetic actuators.
The receiver is not narrow band, and is limited to about 100m
range.
The magnetic actuators are comprised of small coils and a
magnet pair. The coil is glued to the aircraft frame, and the
magnets are glued to the control surface, such that they are inside
the coil. When the coil is energised, the magnets try to align with
the coil field, deflecting the control surface. If the control surface
is lightly spring loaded to return to center, then proportional
control can be achieved by varying the duty cycle of the coil.
The actuators are the other weakest link. They are not light,
typically 07 to 1g each, and they are not efficient, drawing as
much as 05W each, but they are relatively inexpensive, easy to
use and quite durable.

The first model, shown in Fig. 12, flew in December of 2002. It had a
main wing with a 30cm span and 145cm chord, and flapping wings
with a 25cm span and 4cm chord. It used a single-channel control,
throttle-only, with fixed rudder trim for a shallow turn, and fixed pitch
trim to give it a constant nose-up attitude. The flying weight was about
144g, and the model made a number of flights, the longest lasting
about three minutes before coming to rest high up in a tree.
Flight speed was only about 2ms1, and the ability of the model to
fly at very high angles of attack without stalling suggested that
separation control was in effect. With the power off, the model
would stall quite easily in response to gusts, but under power, it
would merely settle back into level flight without loosing any
altitude.
Several months later a second model was built; slightly smaller,
with a 27cm span, and including a rudder control. The weight was
reduced to 134g, and the model could now sustain longer flights, as
trees and buildings could be avoided. During an AIAA technical
seminar at NASA Ames on 12 February 2003, the model was flown
in the test section of the 80 by 120ft NFAC tunnel one of the
smallest aircraft flying the worlds largest wind-tunnel.
The second model was still without elevator control, so it was
trimmed with a nose-up attitude, and throttle was used to control
rate-of-climb. By changing the pitch trim, the flight speed could be
adjusted. The model flew well from speeds as low as 2ms1 up to
about 5ms1. Higher speeds would be possible with active pitch
control, but were risky with a preset pitch trim, as the model might
easily dive into the ground in response to a gust.
A Watt-meter was attached to the model on the bench top, and at
full power 15W was drawn from the battery while the model
produced about 10g of static thrust. Considering only the motor/gear
efficiency (ignoring losses through the DC-DC converter,
receiver/speed-controller, and crankshaft assembly), this suggests
that about 0375W shaft power was delivered, resulting in a figure of
merit (FOM) of about 30g/W at an effective disk-loading of about
6N/m2. This is about 60% higher than comparable rotary wing
vehicles(18).

Table 1
Model details

Model

Span

Length

Avionics
weight

Battery
weight

Motor/gear
weight

Structural
weight

Total
weight

1
2
3
4

30cm
27cm
25cm
23cm

20cm
19cm
18cm
17cm

27g
31g
27g
27g

38g
32g
32g
23g

27g
25g
23g
23g

52g
45g
42g
38g

144g
134g
124g
110g

JONES ET AL

BIO-INSPIRED DESIGN OF FLAPPING-WING MICRO AIR VEHICLES

391

Figure 13. Wind-tunnel variant of the MAV.

A third model, shown previously in Fig. 1, is slightly smaller yet,

with a 25cm span and 124g flying weight. It uses a lower voltage
motor, eliminating the need for the DC-DC converter, and it has
interchangeable parts allowing for more systematic optimisation.
The main wings on this model incorporate about 5 of leading-edge
sweep and a reduced dihedral-angle. The yaw-roll coupling on the
third model is higher than the previous models allowing for a very
tight turn radius, even without active pitch control.
A fourth model was assembled to compete in the 8th International
MAV competition, in April 2004. It had an even smaller span (as
well as the longest linear dimension) of 23cm, and weighed just 11g.
A summary of model dimensions and weight breakdown is shown in
Table 1.

Static

5.0 DESIGN ANALYSIS

With flying models in hand, we now went back to the wind-tunnel in
order to gain a better understanding of the flow physics; hopefully
allowing us to optimise the design. Since the pager motors used in
the flying models had a relatively short lifespan, a model was built
with the same fixed and flapping-wing geometry as the second radio
controlled model, but with a larger fuselage to house a bigger motor
and rotary encoder and with interchangeable parts. The new model,
shown in Fig. 13, was attached to a two-component force balance to
measure lift and thrust, and flow visualisation and unsteady LDV
experiments were run.
5.1 Flow visualisation
Streaklines were generated by a smoke wire which was constructed
from 025mm diameter NiCr beaded wire, heated by passing a
current through it, and using Rosco Fog Juice as the smoke agent.
Imagery was recorded using either a digital still camera or a digital
video camera with a high shutter speed to freeze the motion of the
wings and streaklines. Details of the methods can be found in Jones
and Platzer(6) and Papadopoulos(19).
Flow visualisation experiments were performed with the model
mounted at a 15 angle of attack, at a flow speed of about 2ms1,
approximating the low-speed flight conditions. Initially the flapping
wings were at rest, and they were then quickly accelerated to a
flapping frequency of about 30Hz. The results are shown in Fig. 14,
viewing the model from the left rear corner forward; an angle which
provides a good view of the flow over the upper surface of the left
wing. In the upper image, without wing flapping, it is clearly seen
that the flow separates at the leading edge, and the wing is fully
stalled. In the upper image, after just four flapping strokes, the flow
is already reattached. While the boundary layer appears to be very
thick and unsteady, the outer flow remains parallel to the upper wing
surface and reattaches at the trailing edge. Not only is the flow
entrainment sufficient to reattach the flow, but it requires only about
a tenth of a second to transition. The Reynolds number is about 2
104 for the main wing, and just 5 103 for the flapping wings.

Flapping
Figure 14. Separation control at high AOA.

5.1 Laser-doppler flow measurements

A TSI two-channel LDV system with a single probe was used, with
flow seeding provided by a Rosco fog generator. For unsteady
measurements, the signal from the rotary encoder was fed into a
rotary motion resolver (RMR), which allowed the LDV system to
record periodic data, synchronised with the wing flapping. Further
details about the setup and measurement procedures can be found in
Bradshaw(20).
Use of the RMR provided for higher fidelity velocity measurements, removing the effects of phase-biased flow seeding. Since
LDV is a statistical average of a large number of recorded events, in
an unsteady, periodic flow, the seeding density may fluctuate periodically, biasing the velocity prediction toward the more heavily
seeded parts of the cycle. This biasing was removed using the RMR,
resulting in somewhat higher velocity peaks.
The flow-entrainment effect is illustrated in Figs. 15 and 16. In Fig.
15 the time-averaged velocity profile just in front of the flapping
wings (x = 0) is shown for three cases. In the first case, the main wing
is removed, and the wings are flapped at 32Hz. In the second case the

392

THE AERONAUTICAL JOURNAL

AUGUST 2005

Figure 15. Time-averaged velocity at x = 0.

Figure 16. Time-averaged velocity at x = c.

main wing is included, but the wings are not flapped. In the third
case, the main wing is included and the wings are flapped at 32Hz. In
all three cases the freestream speed is 275ms1, and the model is set
at a 15 angle of attack. Unfortunately, the dihedral of the main wing
masked a large area above the symmetry plane, roughly where the
figure legend is placed, such that the effect of the upper flapping
wing is partially obscured.
Comparing the flapping cases with and without the main wing, the
entrainment effect is clearly seen with about a 30 percent overvelocity at the centerline of the lower flapping wing. Note that the
velocity profile is nearly unaffected by the inclusion of the main
wing. Without flapping the wings, a large velocity deficit is seen
near the stagnation point on the leading edge of the lower flapping
wing. Also note that without flapping the wings, a velocity deficit
appears more than a chordlength above the main wing, illustrating
the severity of the separated flow.
In Fig. 16, velocity profiles a chordlength upstream of the
flapping wings (x = c) are shown for the same three cases, and it
can be seen that the entrainment effect has diminished considerably,
indicating that the flapping wings must be quite close to the trailing
edge of the main wing to capitalise on this phenomenon.

separation suppression characteristic of the design.

Wind-tunnel measurements of lift and thrust, as well as flow
visualisation and unsteady LDV measurements of the surrounding
flowfield have verified the flow separation suppression phenomenon,
and now provide a means for the systematic analysis of the design,
which will eventually allow for design optimisation. While the
current design flies extremely well, it is by no means optimised.
However, operating with a static figure of merit of almost 30g/W, it
exceeds low Reynolds number (5 103 to 2 104) propeller performance at an equivalent disk loading by at least 60 percent.
Furthermore, the design should scale down well, allowing for much
smaller vehicles as power and avionics technology improves.
While there are several natural directions for the project to evolve,
its low speed performance makes it highly suitable for missions
where flight in confined areas is required. It is likely that the flight
speed can be reduced much further, perhaps eventually achieving
hovering flight. Along with the low flight speed comes the ability to
park the vehicle in strategic places, allowing it to gather and transmit
data long past its useful flight lifetime. There are certainly many
other possibilities, limited only be our creativity and perseverance.
Research is currently underway to investigate the complicated
interaction between the three wings. Low Reynolds number NavierStokes analysis of the aeroelastic pitching mechanism is in progress,
as well as simulations of the bi-plane interaction of the flapping
wings. Future work will include all three wing sections in order to
simulate the flow-entrainment and reattachment phenomenon.

6.0 SUMMARY & PROSPECTIVE

Almost a decade of research in flapping-wing propulsion and low
Reynolds number unsteady aerodynamics has culminated in the
development of an unusual flapping-wing propelled micro air
vehicle. The configuration consists of a biplane pair of flapping
wings, which move in counterphase, located just downstream of a
fixed wing. The relatively large fixed wing provides most of the lift,
while the flapping-wings provide the thrust and suppress flow
separation on the main wing due to their flow entrainment. The
symmetry of the flapping-wing pair provides mechanical and
aerodynamic balancing, and produces thrust more efficiently than
conventional flapping-wing systems.
The present model, using commercial off the shelf power and
avionics equipment, has a 25cm span, 18cm length and weighs a
paltry 124g, with two-channel control and enough battery capacity
for 15 to 20 minutes of flight. The model does not have active pitch
control, but the pitch may be manually trimmed to set the flight
speed. Speeds between about 2 and 5ms1 have been achieved, and
the model is virtually stall-proof while under power, due to the

ACKNOWLEDGMENTS
We are grateful for the support received from Richard Foch, head of
the Vehicle Research Section of the Naval Research Laboratory, and
project monitors Kevin Ailinger, Jill Dahlburg and James Kellogg.

REFERENCES
1.
2.

WOODS, M.I., HENDERSON, J.F. and LOCK, G.D. Energy requirements

for the flight of micro air vehicles, 2001, Aero J, 105, (1045), pp 135149.
KEENNON, M.T. and GRASMEYER, J.M. Development of the Black
Widow and Microbat MAVs and a vision of the future of MAV design,
2003, AIAA Paper No 2003-3327.

JONES ET AL
3.
4.
5.

6.
7.
8.
9.

10.
11.
12.
13.
14.

15.
16.

17.

18.

19.

20.

BIO-INSPIRED DESIGN OF FLAPPING-WING MICRO AIR VEHICLES

CHRONISTER, N. Ornithopter media collection, 2001, http://www.

ornothopter.org/rubber.html.
BRIDGES, A. Flying robots create a buzz, Monterey County Herald, 28
July, 2002.
JONES, K.D., LUND, T.C. and PLATZER, M.F. Experimental and
Computational Investigation of Flapping-Wing Propulsion for Micro
Air Vehicles, 2001, Chapter 16, Fixed and Flapping Wing
Aerodynamics for Micro Air Vehicle Applications, AIAA Progress in
Aerospace Sciences Series, MUELLER, T. (Ed), 195.
JONES, K.D. and PLATZER, M.F. Experimental investigation of the
aerodynamic characteristics of flapping-wing micro air vehicles, AIAA
Paper No 2003-0418, 2003.
JONES, K.D., DOHRING, C.M. and PLATZER, M.F. An experimental and
computational investigation of the Knoller-Betz effect, AIAA J, 1998,
36, (7), pp 1240-1246.
LAI, J.C.S., YUE, J. and PLATZER, M.F. Control of backward facing step
flow using a flapping airfoil, Experiments in Fluids, 2002, 32, pp 44-54.
DOHRING, C.M., FOTTNER, L. and PLATZER, M.F. Experimental and
numerical investigation of flapping wing propulsion and its application
for boundary layer control, 1998, ASME International Gas Turbine
Congress, 98-GT-046.
JONES, K.D., CASTRO, B.M., MAHMOUD, O. and PLATZER, M.F. A
numerical and experimental investigation of flapping-wing propulsion
in ground effect, 2002, AIAA Paper No 2002-0866.
CARR, L.W. and CHANDRASEKHARA, M.S. Compressibility effects on
dynamic stall, Progress in Aerospace Sciences,, 1996 32, pp 523-573.
MCCROSKEY, W.J. Unsteady airfoils, Ann Rev Fluid Mech, 1982, 14, pp
285-311.
TUNCER, I.H., LAI, J.C.S. and PLATZER, M.F. A computational study of
flow reattachment over a stationary/flapping airfoil combination in
tandem, 1998, AIAA Paper No 98-0109.
CASTRO, B.M. Multi-block Parallel Navier-Stokes Simulations of
Unsteady Wind Tunnel and Ground Interference Effects, 2001, PhD
thesis, Department of Aeronautics & Astronautics, Naval Postgraduate
School.
CYLINDER, D., SRULL, D. and KELLOGG, J. Biomorphic approaches to
micro air vehicles, 2003, UAV Asia-Pacific 2003 Conference.
LAITONE, E.V. Wind tunnel tests of wings and rings at low Reynolds
numbers, Chapter 5, Fixed and Flapping Wing Aerodynamics for Micro
Air Vehicle Applications, AIAA Progress in Aerospace Sciences Series,
MUELLER, T. (Ed), 195, pp 83-90.
TORRES, G.E. and MUELLER, T.J. Aerodynamic characteristics of low
aspect ratio wings at low Reynolds numbers, 2001, Chapter 7, Fixed
and Flapping Wing Aerodynamics for Micro Air Vehicle Applications,
AIAA Progress in Aerospace Sciences Series, MUELLER, T. (Ed), 195,
pp 115-142.
KORNBLUH, R. D., LOW, T. P., STANFORD, S. E., VINANDE, E., BONWIT,
N., HOLEMAN, D., DELAURIER, J. D., LOEWEN, D., ZDUNICH, P.,
MACMASTER, M. and BILYK, D. Flapping-wing propulsion using
electroactive polymer artificial muscle actuators, Phase 2: radiocontrolled flapping-wing testbed, 2002, SRI International Report ITAD3470-FR-03-009.
PAPADOPOULOS, J.P. An Experimental Investigation of the Geometric
Characteristics of Flapping-wing Propulsion for a Micro Air Vehicle,
2003, Masters thesis, Department of Aeronautics & Astronautics,
Naval Postgraduate School.
BRADSHAW, C.J. An Experimental Investigation of Flapping Wing
Aerodynamics in Micro Air Vehicles, 2003, Masters thesis,
Department of Aeronautics & Astronautics, Naval Postgraduate School.

393