Escolar Documentos
Profissional Documentos
Cultura Documentos
AIAA 2010-1018
In the paper, we discuss an approach we used to design flapping wing Micro Air Vehicles
(MAV). The approach makes use of the conventional precision machining methods, such as
Rapid Prototyping 3D printing, Electrical Discharge Machining and Laser Micromachining
techniques, to manufacture the MAV parts. The current generation of MAVs has a total
weight of 10.8 grams (12.56 grams with one Camera) and a span of 20 cm and it can achieve
a payload of around 3 grams. The MAVs not only are capable of performing simple
acrobatic maneuver in the air but also can hover in place with very little drift. We are in the
process of redesigning the next generation of MAVs, which not only offer a reduction in size
but also make use of vision-based tracking technique to perform obstacle-avoidance flight
adjustments.
Nomenclature
P
Pl
Pr
Fl
Fr
El
Er
I
d
=
=
=
=
=
=
=
=
=
a given point
a pixel captured by left camera, the projection of P
a pixel captured by right camera, the projection of P
focal point
focal point
intersection between left image plane and the line of focal points
intersection between right image plane and the line of focal points
image
displacement
I. Introduction
The development of UAV/MAV has become increasingly important due to the recent successes of UAV in both
military and civilian uses. To name a few, US military UAVs from all services logged nearly 400,000 hours in 2008,
not counting man-portable aircraft and the US Custom and Border Protection Agency has employed more than
200 UAVs for border patrol. Although the deployment of UAV has been very successful, it was not designed to fly
in the urban and in-door environments. As the terrorism will be the main focus of the future warfare, the intelligence,
surveillance and reconnaissance mission will most likely be executed in an in-door or urban environment. For
various military applications, the US Air Force has set forth a goal of deploying a bird-sized MAV by 2015 and
insect size versions by 2030. The MAV potentially has a variety of civilian as well as military uses too difficult or
dangerous for humans, from searching buildings or caves for terrorists to probing damaged nuclear power plants for
radiation leaks or collapsed mine shafts for survivors.
The small scale of such vehicles poses a need for a dramatic change in the air vehicle design paradigm, one as
great as that faced by the Wright Brothers in which they identified that control was the missing link in a workable
aircraft. This new paradigm is simultaneous multi-disciplinary design of integrated multi-functional components and
systems. This new paradigm is simultaneous multi-disciplinary design of integrated multi-functional components
and systems. This small scale no longer affords the luxury of separately designing thrust, control, lift, actuation and
sensing with subsequent integration and adjustment of designs. For example, because the eddy created by the
1
Copyright 2010 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
dynamic flapping wings is as large as the vehicle itself, the exact full body shape becomes crucial to the
aerodynamic design of MAV's. This is the most fundamental difference between the design of MAV's and that of
the large fixed wing aircraft, in which the flow is mainly dominant by the 2-D airfoil shape and wind tunnel testing
can be conducted with a simplified geometry under steady-state conditions.
MAV work has been populated since early 90. A comprehensive review of the MAV activities can be seen in
the book by Mueller et al1. Pornsin-sirirak et al.2 created Caltechs Micro-Bat with a 6-min flight record of
flapping MAVs using MEMS process and the titanium-parylene material in 2001. Barrett et al.3 announced the
successful hovering of TU Delfts Delfly composed of a pair of dragonfly-like flexible wings in 2005. The team
also developed in 2008 the smallest ornithopter fitted with a camera, the Delfly Micro4. Although this version
measures 10 cm and weighs 3 grams, it still requires some work in flying outdoor. Jones et al.5 of Naval Graduate
School of US made a fixed-wing type MAV with a scissor-like clapping tail thruster in 2005. Banala et al.6 at
Delaware University employed a 5-bar mechanism for generating a prescribed wing motion taken from a hawk moth
kinematic flight data in 2005. McIntosh et al.7 designed a mechanism for biaxial rotation of a wing for a hovering
MAV in 2006. Robert Wood8 at Harvard University developed the smallest ornithopter, at just 3 cm, but this craft is
flying between the guided poles and it gets its power through a wire.
Figure 1. Study of the wing kinematics of a dragon fly using a 3-D high speed camera.
2
American Institute of Aeronautics and Astronautics
Figure 2. CAD model of the dragon fly and its CFD study.
Figure 6. Laser machined miniature gears: (a) 15 mm diameter polyimide gear, (b) 2 mm diameter stainless
steel gear.
4
American Institute of Aeronautics and Astronautics
Figure 7. Miniature assemblies and welding: (a) 3D drawing of assembly, (b) finished assembly, (c) detail of
laser-welded joints.
In this study, we propose a novel design of the three-way clap and fling mechanism, as shown in Figure 8.
Within one flapping period, the clap and fling motion was observed not only in left and right sides, but also exists on
the upper side of the wings. Other MAVs, such as i-Bird 7 and Delfly-II 8 also used the four wing mechanism but
did not seem to provide a full upper clap and fling motion. As will be demonstrated in this paper, the upper clap and
fling motion not only provides a wider flapping motion of the wings but also gives rise to a stronger forward motion
of the vehicle as compared to that without the full upper motion.
Table 1. Breakdown of the weight for each component of the MAV with camera.
Item
Weight (g)
No.
Item
Weight (g)
3-channel receiver
0.95
7
Tail
1.22
Actuator
0.98
8
Wings
1.05
Motor
2.80
9
2 Drive Gears
0.30
Frame
0.24
10 Drive Links
0.25
Battery (3.7v 30mAh)
1.25
11 Tape, Glue, Carbon Frame
1.00
Camera
1.85
12 Camera Battery
0.67
Total Weight
12.56 g
5
American Institute of Aeronautics and Astronautics
Figure 9. Wind tunnel test system: (a) 6-DoF load cell , (b) wind tunnel.
Since the MAV we have developed is small enough to fit into the test section of the wind tunnel, we are able to
conduct the experiments in full scale, as shown in the flow visualization of the smoke in Figure 10(a). The input
power to the motor and the wind speeds are varied from 1.5 to 2.2 Watts and 0 to 1.5 m/s, respectively. The
corresponding flapping frequency may range from 15 to 20 Hz. The data had a sample rate of 1,000 points per
second for a total of 13 seconds. This corresponds to 260 and 195 cycles per measurement for a flapping frequency
of 20 and 15 Hz, respectively.
The applied voltages for the MAV are 3.3, 3.7 and 4.1 volts. The angle of attack (AOA) is set at 0 degrees and
the wind speed ranges from 0 to 1.5 m/s. The plots for the aerodynamic performance vs. power are depicted in Fig.
10(b)-(d). Fig. 10(b) shows the mean lift force is positive even at 0 degree of AOA and it increases with increasing
wind speed and flapping frequency. Fig. 10(c) shows that the thrust force stalled at 15 gf under 18 Hz of flapping
frequency and it decreases with increasing wind speed. The stall phenomenon may be caused by aeroelastic effects
of the wing structure. Therefore, the maximum performance of the MAV is running at around 17 Hz. It should be
noted that when the MAV is hovering, it will subject to an angle of attack of 57 degree to maintain a total MAV
weight of 12.56 gram.
(a)
(b)
3.0
Lift force (gf) .
2.5
U= 0 m/s
2.0
U=0.5 m/s
1.5
U=1.0 m/s
1.0
U=1.5 m/s
0.5
0.0
1
1.5
2.5
Power (watt)
(d)
16
Thrust (gf) .
21
.
14
U= 0 m/s
U=0.5 m/s
U=1.0 m/s
U=1.5 m/s
12
10
8
Frequency (Hz)
(c)
U= 0 m/s
19
U=0.5 m/s
U=1.0 m/s
17
U=1.5 m/s
15
6
1
1.5
2.5
1.5
2.5
Power (watt)
Power (watt)
V. CFD Simulation
The Computational Fluid Dynamics (CFD) program, SC/Tetra10, developed by Software Cradle Company was
used to simulate the MAV design. The model is the exact replicate of the origin design shown in Figure 8 but it does
not include the electronics, motor, and gears. Excluding these from the model will simplify the geometry, making it
easier for gridding. Also, these features will not significantly affect the air flow generated by the wings. The main
goal of this research is to optimize shape and size of the MAV and compare with the experimental design. Our
overall goal is to reduce the wingspan while maintaining functionality and controllability. The results assume that
the MAV has rigid wings. Since the aeroelastic effects of the wings are not taken into account here, a direct
6
American Institute of Aeronautics and Astronautics
comparison with experiments is not possible. However, the simulation of a MAV with a flexible wing by coupling
SC/Tetra CFD and ABAQUS FEA software is currently in progress.
The overset mesh with moving elements feature in Sc/Tetra was used to simulate the flapping wings of the
MAV. This feature will also be implemented to study the effects of the control surfaces found in the tail of the
MAV. Currently, the simulation is of a stationary MAV with zero angle of attack and the flapping frequency
simulated was 100 radians per second (approximately 16 Hz). Particles were released from the wings surfaces
during the simulation. These particles are used to observe the vortices produced from the edge of the wing surfaces
shown in Fig. 11.
Figure 12. CFD results of the MAV with rigid flapping wing - Lift Coefficient and Surface Pressure.
7
American Institute of Aeronautics and Astronautics
(b)
Figure 13. Actual flight tests of the current MAV: (a) during an acrobatic maneuver & (b) during an ISR
mission.
Using image processing techniques, such as edge detection, the moving objects were tracked as shown in
Figure 15. Once the tracked objects are matched within the two sets of images captures by the pair of cameras using
image registration techniques, 3D location of the tracked objects relative to the camera setup can be computed based
on epipolar geometry. Obviously, a single camera can only provide a two-dimensional location of the tracked object
relative to the camera. When using a second camera, the distance to the camera can also be determined resulting in a
three-dimensional location. The principle of epipolar geometry is illustrated in Figure 16. In order to determine the
distance for a given point P that resulted in a pixel Pl captured by the left camera, the light ray is traced backward
from the focal point Fl of that camera through the pixel Pl. This ray eventually encounters the point P. However, the
exact distance is not known at this point. By using the second camera, this distance can now be computed. The line
represented by the connections between the points Er and Pr, where Er is simply the intersection between the image
plane of the right camera and the line connecting the focal points of both cameras; i.e., Fl and Fr; and Pr is the
projection of the point P as captured by the camera; i.e., the intersection between the image plane of the right camera
and the point P and the focal point Fr, now resembles the projection of the light ray that was traced back from the
focal point of the left camera Fl through the projected point Pl captured by the left camera all the way to the actual
point P. As a result, the distance between the left camera and the point P can now be computed based on the distance
between the points Er and Pr within the image captured by the right camera.
Salient features selected from the edge detection image processing techniques defined in the previous frames
are selected in the subwindow and tracked using the Lucas-Kanade feature tracker12. Given a feature point (x,y) and
a small window centered at this point in the current image I, the tracking algorithm gives the displacement
r
2
d = [d x , d y ] by minimizing the L2 norm: [d x , d y ] = arg min I ( x, y, t ) I [ x + d x , y + d y , t + t ) .
r
r
1
,
By assuming a small displacement vector d , the linear closed-form solution becomes: d = H b
I x I x I x I y
I x It
where H =
, b =
and I x , I y and I t are x, y, and t derivatives of the image.
I y I x I y I y
I y It
The equations will be solved by the iterative Newton-Raphson method in pyramids of images on a small subset of
distinctive pixel windows, as shown in Figure 17. From the correspondences of the tracked features, the algorithm
estimates the relative motion of the vehicle, which will then be used for MAV real-time control application.
(b)
Figure 15. Edge detection image processing techniques: (a) identify the location of the tracked objects, (b)
determine their shape and extent.
9
American Institute of Aeronautics and Astronautics
Pl
Pr
El
ER
Fl
Fr
Figure 17. The Lucas-Kanade tracker where the white dots are selected feature points.
10
American Institute of Aeronautics and Astronautics
The research for an autonomous MAV is just begun, what was described here is only the tip of the iceberg.
There are still many unexplored areas needing attention if our final goal is to mimic the nature such that an MAV
can be as agile as a dragon fly or a fruit fly.
Acknowledgments
The author would like to thank his MAV team members for their contributions to this paper: Alexander Feist,
Jaderic Dawson, Jeremy Crank, Dr. Larry Dosser of MLPC, Professor Joseph Shang, Professor Haibo Dong and
Professor Thomas Wischgoll. The participations of the US Air Force Wright Patterson Research Laboratory,
Software Cradle Co., Ltd. and Mount Laser & Photonics Center, Inc. to our MAV program are also appreciated.
References
1
Muller, T. J., Kellogg, J. C, Ifju, P. G. and Shkarayev, S. V., Introduction to the Design of Fixed-Wing Micro Air Vehicles,
American Institute of Aeronautics and Astronautics, Inc, 2006.
2
Pornsin-sirirak, T. N., Tai, Y. C., Nassef, H., and Ho, C. M., Titanium-Alloy MEMS Wing Technology for a Micro Aerial
Vehicle Application, Sensors and Actuators A: Physical, Vol. 89, No. 1-2, pp. 95-103, 2001.
3
Barrett, R., McMurtry, R., Vos, R., Tiso, P., De Breuker, R. Barrett, R., McMurtry, R., Vos, R., Tiso, P., and De Breuker, R.,
Post-Buckled Precompressed (PBP) Elements: A New Class of Flight Control Actuators Enhancing High-Speed Autonomous
VTOL MAVs, Proceedings of SPIE - The International Society for Optical Engineering, Vol. 5762, pp. 111-122, 2005.
4
TU Delft, website: http://www.defly.nl/
5
Jones, K. D., Bradshaw, C. J., Papadopoulos, J., and Platzer, M. F., Bio-Inspired Design of Flapping-Wing Micro Aerial
Vehicles, Aeronautical Journal, Vol. 109, No. 1098, pp. 385-393, 2005.
6
Banala, S. K. and Agrawal, S. K., Design and Optimization of a Mechanism for Out-of-Plane Insect Wing like Motion with
Twist, Journal of Mechanical Design/ Transactions of the ASME, Vol. 127, No. 4, pp. 841-844, 2005.
7
McIntosh, S.H., Agrawal, S.K., Khan, Z., Design of a Mechanism for Biaxial Rotation of a Wing for a Hovering Vehicle,
IEEE/ASME Transactions on Mechatronics, Vol. 11, No. 2, pp. 145-153, 2006.
8
Harvard Microrobotics Lab website: http://micro.seas.harvard.edu/
9
Liang, Z. and Dong, Computational Study of Wing-Wake Interactions between Ipsilateral Wings of Dragonfly in Flight,
AIAA 2009-4192, 2009.
10
Software Cradle Company, User Manuals and Guide, 2008.
11
Wischgoll, T., Hutchinson, T. C., Kster, F, Optical (Camera-Based) Technology for Seismic Risk Assessment, ASME
International Mechanical Engineering Congress & Exposition, Washington, D.C., 2003.
12
Kanade, T., Amidi, O., and Ke, Q, "Real-Time and 3D Vision for Autonomous Small and Micro Air Vehicles," 43rd IEEE
Conference on Decision and Control, December, 2004.
11
American Institute of Aeronautics and Astronautics