P. 1
Paper 75

Paper 75

|Views: 19|Likes:
Published by Sparrow-hawk

More info:

Published by: Sparrow-hawk on Dec 17, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

12/17/2010

pdf

text

original

Derivation and simulation of the nonlinear dynamics of a

flapping wing micro-air vehicle
C. Orlowski

, A. Girard, and W. Shyy
University of Michigan, Department of Aerospace Engineering, 1320 Beal Ave, Ann Arbor, MI USA 48109
ABSTRACT
The paper presents a derivation of the equations
of motion for a flapping wing micro-air vehi-
cle. The equations of motion are derived using
D’Alembert’s Principle extended to rigid bod-
ies. The micro-air vehicle is modeled as a a sys-
tem of three rigid bodies with rigidly attached
wings. Each wing has three separate degrees
of freedom: flapping, lagging, and feathering.
The MAV is assumed to be operating in an envi-
ronment with a flat Earth, constant gravity, zero
wind, and the MAV has constant mass. The
model is simulated using time-averaged aerody-
namic force and moment data in a normal hover-
ing mode.
1 NOMENCLATURE
X, Y , Z: components of the position vector of the central
body in an inertial frame
ψ, θ, φ: 3-2-1 Euler Angles for the orientation of the central
body with respect to a inertial frame
α
R
, α
L
: angle of attack of the right and left wings, respec-
tively (rotation about
¯
b
y
axis)
δ
R
, δ
L
: flapping angle of the right and left wings, respec-
tively (rotation about
¯
b
x
axis, up and down flapping motion)
ζ
R
, ζ
L
: lagging angle of the right and left wings respectively
(rotation about
¯
b
z
axis, forward and back flapping motion)
u, v, w: components of the translational velocity vector of
the central body
p, q, r: components of the angular velocity vector of the
central body
˙ α
R
, ˙ α
L
: time rate of change of the angle of attack
˙
δ
R
,
˙
δ
L
: time rate of change of the flapping angle
˙
ζ
R
,
˙
ζ
L
: time rate of change of the lagging angle
m
1
, m
2
, m
3
: mass of the central body, mass of the right wing
and mass of the left wing
¯ ρ
c2
, ¯ ρ
c3
: position vectors from the hinge to the respective
wing’s center of mass
¯
R,
¯
L: position vectors of the hinge from the central body
center of mass in the B frame
¯
b
x
,
¯
b
y
,
¯
b
z
: unit vectors of the central body-fixed frame, the B
frame
Q
j
: generalized forces for each generalized coordinate

Email address(es): contact cptorlo@umich.edu
¯ γ
ij
: velocity coefficient of the ith body with respect to the
jth generalized speed
¯
β
ij
: angular velocity coefficient of the ith body with respect
to the jth generalized speed
[F
x
, F
y
, F
z
]: aerodynamics forces in body-fixed coordinates
[L, M, N]: aerodynamics moments in body-fixed coordinates
B: body-fixed frame for the central body
H
R
: body-fixed frame for the central body, origin at right
wing hinge
H
L
: body-fixed frame for the central body, origin at left
wing hinge
W
R
: body-fixed frame for the right wing
W
L
: body-fixed frame for the left wing
I
1
: inertia tensor for the central body in the B frame
I
2
: inertia tensor for the right wing in the W
R
frame
I
3
: inertia tensor for the left wing in the W
L
frame
I

2
, I

3
: inertia tensors for the right wing and left wing in the
H
R
and H
L
frames
R
B
: rotation matrix from inertial frame to B frame using
3-2-1 Euler Angles
R
R
: rotation matrix from W
R
frame to H
R
frame
R
L
: rotation matrix from W
L
frame to H
L
frame
R
ζ
R
, R
α
R
, R
δ
R
: individual rotation matrices for W
R
frame
to H
R
frame
R
ζ
L
, R
α
L
, R
δ
L
: individual rotation matrices for W
L
frame
to H
L
frame
2 INTRODUCTION
The goal of the presented research, as well as future work,
is to further the knowledge of the dynamic behavior of a flap-
ping wing vehicle and to develop effective control schemes
for the operation of a flapping wing vehicle in a relevant envi-
ronment. The research in the flapping wing field has greatly
increased over the past 10 years, especially in aerodynam-
ics [1]. Complex models of flapping wing aerodynamics, to
include 3-D effects, have been successfully developed, eval-
uated and implemented [2, 3]. The next step is to take the
knowledge of the complex and unstable aerodynamics and
incorporate that knowledge into a successful control scheme,
either by nonlinear methods, linear methods, or a combina-
tion of both.
In order to successfully implement the control scheme,
the development and thorough understanding of the nonlin-
ear dynamics and kinematics is paramount. The paper out-
lines the development of a dynamic model for a flapping-wing
1
MAV. The preliminary goal for vehicle size is a total mass
of 30 grams and a total wingspan of approximately 15 cen-
timeters. The method used to derive the equations of motion
is D’Alembert’s Principle for Multiple Rigid Bodies, which
is a hybrid of LaGrangian and Eulerian techniques [4]. The
flapping wing MAV is modeled as three rigid bodies (central
body, right wing, left wing) with a rigid airframe and rigidly
attached wings. The nonlinear dynamics will be incorporated
with lift, drag and aerodynamic moment data from models
developed in reference [5]. The combination of the nonlinear
dynamics and aerodynamics should allow for accurate study
of the motion of the vehicle. It will enable the discovery of
steady-state flight configurations and the eventual develop-
ment of effective control schemes.
The paper is organized in the following manner. Section
3 is a brief literature review focusing on the dynamic models
and control schemes for a flapping wing micro-air vehicle.
Section 4 presents a detailed development of the model with
relevant reference frames and the derivation of the nonlinear
equations of motion. Section 5 presents the simulation efforts
and results, accompanied by the dynamic and aerodynamic
models used in the simulation. Section 6 presents Conclu-
sions and Future Work.
3 LITERATURE REVIEW
The majority of the previous work in regards to flapping
wing vehicles focuses on the aerodynamics and the interac-
tions of the flapping wings with the fluid flow around the
wings. A limited amount of work has been completed in
regards to the development of dynamic models and control
schemes to successfully operate a flapping wing micro-air ve-
hicle.
Sun and Wang [1] acknowledge that the field of aerody-
namics, in regards to insect flight, is highly studied. The main
purpose of reference [1] is to produce a quantitative analy-
sis of the stability of hovering flight for a model insect. The
authors chose a dynamics model previously given in [6, 7].
The dynamics model chosen is the standard, linearized air-
craft dynamics that can be found in [8]. In contrast to the
Taylor and Thomas model [7], Sun and Wang use stability
and control derivatives calculated by CFD methods. Taylor
and Thomas used time average aerodynamics force and mo-
ment calculations for a wing beat cycle. The Sun and Wang
model limits the model to have six degrees of freedom, but en-
hances the linearization and simplification of the model with
improved accuracy of the aerodynamics, versus the models
used by Taylor and Thomas.
The authors choose the equilibrium point for their anal-
ysis to be motionless hovering, with no translational or rota-
tional accelerations. The wings need to accelerate to maintain
the flight condition. Sun and Wang make further simplifica-
tions, given the flight conditions, by only accounting for the
x and z translational motion of the model, in addition to the
pitching motion and pitch angle, q and θ, respectively. The
equilibrium point is determined by balancing the force and
moment equations. The stability derivatives are calculated
by only taking into account the motion of the wings, for at
hovering the aerodynamics of the central body are neglected
due to the absence of translational motion. Sun and Wang
concluded that the hovering motion can be stabilized using a
combination of two of the four available controls (δΦ, change
in position, δφ, change in mean position, α
1
, equal change in
angle of attack, and α
2
, differential change in the angle of
attack).
In reference [9], the authors derived the equations of mo-
tion for a flapping wing vehicle using Eulerian methods. The
central body is modeled as a point mass without mass mo-
ments of inertia. The hinge point for both wings is the same
and is positioned at the central body. The goal of the paper
is to study trajectory guidance, and not exactly implement a
control scheme, so the assumptions and model simplifications
seem justified. Modeling the central body as a point mass ef-
fectively eliminates the rotational considerations of the cen-
tral body and the coupling effects between the flapping mo-
tion of the wings and rotation of the central body. The authors
use two body-fixed reference frames, one for each wing. The
wings are modeled with three degrees of freedom: plunge
(up and down flapping), pitch (angle of attack), and sweep
(forward and back flapping). The disadvantage of the two
frames, and no central body frame, requires the tracking of
the absolute translational and rotational velocity of not only
the central body, but both wings as well. Furthermore, ve-
locities calculated in each of the wings’ frames will need to
be transformed back into the inertial frame, due to the lack
of a body-fixed frame at the central body point mass. Ad-
ditionally, the method chosen requires the calculation of the
reaction forces between the wings and the hinges.
The main advantage to the method of deriving the equa-
tions of motion chosen in [9] is that it reduces the non-
linearities of the system. There is still rotational motion of
the wing, so the nonlinearities associated with Euler’s Equa-
tions of Motion are still present in the equations of motion.
However, the coupling between p, q, and r is non-existent
and there is no need to track the Euler Angles of the cen-
tral body. Furthermore, as undertaken in the paper, the 3-2-1
Euler Angles tracking the orientation of each wing can be di-
rectly used as the time rate of change of the pitch, plunge
and sweep angles. The dynamics model does not require ad-
ditional variables to track and transform the motions of the
wing into an inertial frame.
In reference [10], Buler, et al., derive the nonlinear equa-
tions of motion using the Gibbs-Appel Equations by using ten
generalized coordinates, q, and ten quasi-velocities, w, listed
below:
q = [x, y, z, Φ, Θ, Ψ, β
L
, β
R
, θ
L
, θ
R
]
w = [U, V, W, P, Q, R,
˙
β
L
,
˙
β
R
,
˙
θ
L
,
˙
θ
R
].
The above equations use standard aircraft notation, except in
the case of β (used to denote the flapping angle - motion about
the longitudinal axis of the central body) and θ (used to de-
note the angle of attack of the wings). Buler, et al., limit the
flapping kinematics of their model and do not account, or al-
low for, forward and back (lagging) translation of the wings.
Strictly speaking, their model restricts the stroke-plane of the
wing beat to a single plane. The model results in a ten de-
gree of freedom, 2nd-order system (or, alternatively, a twenty
degree of freedom, 1st-order system). The method of deriv-
ing the equations of motion results in the preservation of the
numerous, inherent non-linearities present in a flapping wing
micro-air vehicle.
The second half of the paper, after outlining the deriva-
tion of the equations of model and the aerodynamic model,
presents an outline of Linear Control Theory. The authors
continue with an analysis of the control of the linearized sys-
tem previously derived, but do not provide an analytical so-
lution for the linearization. A desired trajectory is assumed
and a linearization, and associated Jacobian matrix, is numer-
ically calculated based on the said trajectory. The paper pro-
vides coverage of the various control methods at the authors’
disposal and their respective shortcomings. The authors use
a linear quadratic regulator problem to determine a nominal
control input for the flapping wings of the vehicle. In the end,
the authors produce simulations of the dynamics of a flapping
vehicle/animal, based on the solution of a Newton-Raphson
solver to produce the necessary control input vector for the
desired output vector.
In reference [11], Bolender derives the nonlinear equa-
tions of motion for a flapping wing MAV utilizing Kane’s
Equations. The main difference from the derivation in refer-
ence [11], from other models, is the modeling of the central
body as two separate rigid bodies. The central body has a
main body, B, and a tail section, T. The tail section is lim-
ited to one degree of freedom about the hinge point between
body B and body T. The tail section is used for pitch control
of the flapping wing MAV. Bolender treats the wings in two
separate manners. One treatment ignores the combined mass
of the wing. The other treatment accounts for the mass of the
wing in the dynamics of the central body. However, in either
case, the wings are not treated as having separate degrees of
freedom and are considered to be holonomically constrained
to the central body. The wings’ motion is treated as a pre-
scribed motion. The motion of the wings is a function of time
and a control input.
4 DERIVATION OF THE NONLINEAR EQUATIONS OF
MOTION
4.1 D’Alembert’s Principle for Multiple Rigid Bodies
The method chosen to derive the equations of motion is
D’Alembert’s Principle Extended to Multiple Rigid Bodies.
A detailed description of the method can be found in [12, 13].
The chosen method is also briefly covered in [4]. The Gibbs-
Appell Equations, presented in [10], and Kane’s Equations,
presented in [11], are both extensions of D’Alembert’s Prin-
ciple [4] and have their own strengths and weaknesses [11].
The statement of D’Alembert’s Principle, in [12], is
n

i=1
[m
i
(
˙
¯ v
i
+
¨
¯ ρ
ci
)·¯γ
ij
+(I
i
·
˙
¯ ω
i
+¯ ω
i
×I
i
·¯ ω
i
+m
i
¯ ρ
ci
×
˙
¯ v
i

¯
β
ij
] = Q
j
,
(1)
where i is the number of rigid bodies and j is the number of
generalized coordinates. The flapping wing vehicle is mod-
eled as a system of three rigid bodies (n = 3): a central
fuselage with two rigid wings attached at ideal hinges. The
wings and central body are modeled as rigid airframes. A
body-fixed frame, frame B, is attached to the central body
at its center of mass; the frame is aligned with the
¯
b
x
vector
through the nose of the micro-air vehicle,
¯
b
y
is perpendicular
to
¯
b
x
and pointed out of the right side of the central body, and
¯
b
z
pointed downward out of of the bottom of the micro-air
vehicle. The body-fixed frame is the same as for fixed-wing
aircraft and is presented in Figure 1.
Figure 1: Vehicle Model
The vectors
¯
R and
¯
L denote position vectors from the
central body center of mass to the hinge/attachment points of
the wings. At the hinge points, two central body-fixed frames
are attached, with directions parallel to the central body B
frame. The H
R
frame is fixed with origin at the right wing
hinge point. The H
L
frame is fixed with origin at the left wing
hinge point. Since the central body is modeled as a rigid body,
the magnitude and components of
¯
R and
¯
L are constant. We
define the individual components of
¯
R and
¯
L as
¯
R =

R
x
R
y
R
z

(2)
and
¯
L =

L
x
L
y
L
z

. (3)
The vectors ¯ ρ
c2
and ¯ ρ
c3
denote the vectors from the hinge
points on the central body to the respective centers of mass for
each wing (¯ ρ
c2
for the right wing and ¯ ρ
c3
for the left wing).
The magnitude of the ¯ ρ
ci
vectors is constant, based on the as-
sumption of rigid bodies and rigidly attached wings. A wing-
fixed frame is attached to each wing at the hinge point. The
origin of the winged fixed frames is the hinge point and when
the wings at are their respective initial positions, the axes of
the wing-fixed frames, W
R
and W
L
, are parallel with the cen-
tral body-fixed frame B, and are identical to the hinge frames,
H
R
and H
L
. The frames W
R
and W
L
and the vectors
¯
R,
¯
L,
¯ ρ
c2
, and ¯ ρ
c3
are presented in Figure 2.
Figure 2: Wing Frames
The motion of the center of mass of the wings is tracked
similar to a mass moving in spherical coordinates, with a con-
stant radius, fixed at the origin of the system. In this case, the
origin of the respective motion is the hinge point, which is co-
incident with the origin of the W
R
, H
R
, H
L
and W
L
frames.
The constant radius is the magnitude of the vectors, ¯ ρ
c2
and
¯ ρ
c3
. The center of mass of the wing is assumed to be along
the y-axis of the hinge frames when the wing angles are iden-
tically zero. The position of the centers of mass of each wing
are expressed by the following vectors:
¯ ρ
c2
= ρ
c2
[cosδ
R
sinζ
R
¯
b
x
+ cosδ
R
cosζ
R
¯
b
y
+ sinδ
R
¯
b
z
] (4)
¯ ρ
c3
= ρ
c3
[cosδ
L
sinζ
L
¯
b
x
−cosδ
L
cosζ
L
¯
b
y
+ sinδ
L
¯
b
z
]. (5)
Developing equations of motion using D’Alembert’s
Principle for Rigid Bodies requires the selection of reference
points for each rigid body. The reference point for the cen-
tral body, rigid body 1, is it’s respective center of mass. The
reference points for the right and left wing are chosen to be
the hinge attachment points, located on the central body. The
vectors tracking the center of mass of the wings in the hinged-
fixed frames are the vectors ¯ ρ
c2
and ¯ ρ
c3
. The required ac-
celeration vectors of
¨
¯ ρ
c2
and
¨
¯ ρ
c3
are derived using diligent
application of the chain rule.
There are three main advantages to selecting
D’Alembert’s Principle. First, by taking the hinge points as
reference points, the requirement to calculate the reaction
forces between the wings and the central body is eliminated.
Second, the mass moments of inertia of the wings need to
be initially calculated about the reference points. The mass
moments of inertia will remain constant in the wing-fixed
frames and do not need to be translated into the B frame [12].
Third, by choosing reference points on the central body, the
absolute velocity of the wings does not need to be tracked.
In order to successfully develop the equations of motion, the
generalized coordinates and quasi-velocities need to be spec-
ified. In this formulation, the twelve generalized coordinates
are chosen to be the following: X, Y , Z (inertial position of
central body), ψ, θ, φ (3-2-1 Euler Angles of central body),
δ
R
, α
R
, ζ
R
, δ
L
, α
L
, ζ
L
. The twelve quasi-velocities are the
following: u, v, w, p, q, r,
˙
δ
R
, ˙ α
R
,
˙
ζ
R
,
˙
δ
L
, ˙ α
L
, and
˙
ζ
L
. To
summarize, the generalized coordinates, q, are
q
j
= [X, Y, Z, ψ, θ, φ, δ
R
, α
R
, ζ
R
, δ
L
, α
L
, ζ
L
] (6)
and the quasi-velocites, u
j
, are
u
j
= [u, v, w, p, q, r,
˙
δ
R
, ˙ α
R
,
˙
ζ
R
,
˙
δ
L
, ˙ α
L
,
˙
ζ
L
]. (7)
The importance of flapping and lagging angles, and the asso-
ciated degrees of freedom, is important when biological fly-
ers transition from hovering to forward flight [2]. Figures 3,
4, and 5 show the vehicle model from different view points
and the relationship of the wing frames, W
R
and W
L
, and the
central body frame B. The left side of the vehicle, and α
L
,
is not depicted, but is simply the mirror image of Figure 4.
Dashed lines represent a negative direction in the respective
reference frame. The flapping angles, δ
R
and δ
L
, are posi-
tive in a downward motion (positive
¯
b
z
). The lagging angles,
ζ
R
and ζ
L
are positive in a forward flapping motion (positive
¯
b
x
). The hinge frames, H
R
and H
L
, are not depicted, but are
parallel to the B frame.
Figure 3: Front View - δ
R
and δ
L
4.2 Velocity and Angular Velocity Coefficients
The first step in the derivation is to determine the velocity
coefficients, ¯ γ
ij
=
∂¯ vi
∂uj
, and the angular velocity coefficients,
¯
β
ij
=
∂¯ ωi
∂uj
, where i denotes the rigid body and j denotes the
generalized speeds. The translational and rotational velocities
of each of the rigid bodies are
¯ v
1
= u
¯
b
x
+v
¯
b
y
+w
¯
b
z
¯ v
2
= ¯ v
1
+ ¯ ω
1
×
¯
R
¯ v
3
= ¯ v
1
+ ¯ ω
1
×
¯
L
(8)
Figure 4: Right Side View - α
R
Figure 5: Top View - ζ
R
and ζ
L
and
¯ ω
1
= p
¯
b
x
+q
¯
b
y
+r
¯
b
z
¯ ω
2
= (p +
˙
δ
R
)
¯
b
x
+ (q + ˙ α
R
)
¯
b
y
+ (r −
˙
ζ
R
)
¯
b
z
¯ ω
3
= (p −
˙
δ
L
)
¯
b
x
+ (q + ˙ α
L
)
¯
b
y
+ (r +
˙
ζ
L
)
¯
b
z
. (9)
The velocity and angular velocity coefficients are ob-
tained by taking the partial derivatives of ¯ v
i
and ¯ ω
i
with re-
spect to the generalized speeds. The velocity coefficients,
¯ γ
1j
, and the angular velocity coefficients,
¯
β
1j
for generalized
speeds j = 1 . . . 6 of the central body are
¯ γ
1j
=

¯
b
x
¯
b
y
¯
b
z
0 0 0

(10)
and
¯
β
1j
=

0 0 0
¯
b
x
¯
b
y
¯
b
z

. (11)
The velocity and angular velocity coefficients of the cen-
tral body for coordinates j = 7 . . . 12 are identically zero. The
velocity coefficients for the right wing, ¯ γ
2j
, and the left wing,
¯ γ
3j
, for j = 1 . . . 6 are
¯ γ
2j
= [
¯
b
x
¯
b
y
¯
b
z
. . .
−R
z
¯
b
y
+R
y
¯
b
z
R
z
¯
b
x
−R
x
¯
b
z
−R
y
¯
b
x
+R
x
¯
b
y
] (12)
γ
3j
= [
¯
b
x
¯
b
y
¯
b
z
. . .
−L
z
¯
b
y
+L
y
¯
b
z
L
z
¯
b
x
−L
x
¯
b
z
−L
y
¯
b
x
+L
x
¯
b
y
]. (13)
The velocity coefficients for the left and right wing, for
j = 7 . . . 12 are identically zero. The angular velocity co-
efficients of the right and left wings,
¯
β
2j
and
¯
β
3j
, respec-
tively, are identically zero for j = 1 . . . 3. For the right wing,
¯
β
2j
is identically zero for j = 10 . . . 12. For the left wing,
¯
β
3j
is identically zero for j = 7 . . . 9. For quasi-velocities
j = 4 . . . 6, the angular velocity coefficients are
¯
β
2j
=

¯
b
x
¯
b
y
¯
b
z

(14)
¯
β
3j
=

¯
b
x
¯
b
y
¯
b
z

. (15)
For the right wing, with respect to coordinates j = 7 . . . 9,
the angular velocity coefficients are
¯
β
2j
=

¯
b
x
¯
b
y

¯
b
z

. (16)
In regards to the left wing, the angular velocity coefficients
for j = 10 . . . 12 are
¯
β
3j
=


¯
b
x
¯
b
y
¯
b
z

. (17)
The accelerations of the three rigid bodies are obtained by
differentiating the translational and rotational velocities and
making proper use of the transport theorem [4]. For exam-
ple, the acceleration of the right wing,
˙
¯ v
2
in the central body
frame, is
˙
¯ v
2
=
˙
¯ v
1
+
˙
¯ ω
1
×
¯
R+ ¯ ω
1
× ¯ v
1
+ ¯ ω
1
×(¯ ω
1
×
¯
R). (18)
4.3 Inertia Tensors
For the central body, an x −z plane of mass symmetry is
assumed. No planes of mass symmetry are assumed for either
wing during the model development. As a result, the resulting
mass moments of inertia matrices for each rigid body are the
following:
I
1
=


I
xx,1
0 −I
xz,1
0 I
yy,1
0
−I
xz,1
0 I
zz,1


(19)
I
2
=


I
xx,2
−I
xy,2
−I
xz,2
−I
xy,2
I
yy,2
−I
yz,2
−I
xz,2
−I
yz,2
I
zz,2


(20)
I
3
=


I
xx,3
−I
xy,3
−I
xz,3
−I
xy,3
I
yy,3
−I
yz,3
−I
xz,3
−I
yz,3
I
zz,3


. (21)
4.4 Rotation Matrices
Although it is not necessary to continuously calculate the
mass moments of inertia for the wings in the B frame, it may
be necessary in the use of the model to rotate aerodynamic
forces generated by the flapping of the wings from the W
R
and W
L
frames into the B frame. R
R
is the rotation matrix
of the W
R
frame into the H
R
frame. Treating the angles de-
scribing the orientation of the right wing, ζ
R
, α
R
, and δ
R
, as
3-2-1 Euler Angles results in the calculation of R
R
.
R
ζ
R
=


cosζ
R
sinζ
R
0
−sinζ
R
cosζ
R
0
0 0 1


R
α
R
=


cosα
R
0 −sinα
R
0 1 0
sinα
R
0 cosα
R


R
δ
R
=


1 0 0
0 cosδ
R
sinδ
R
0 −sinδ
R
cosδ
R


R
R
= R
δ
R
R
α
R
R
ζ
R
. (22)
A similar procedure can be utilized to obtain the rota-
tion of aerodynamic forces and moments generated by the
left wing into the B frame. The rotation matrix for the left
wing is
R
L
= R
δ
L
R
α
L
R
ζ
L
. (23)
The rotation matrix R
α
L
has the same structure as it’s coun-
terpart on the right wing, R
α
R
. The rotation matrices R
ζ
L
and R
δ
L
are different and are defined as
R
ζ
L
=


cos(π −ζ
L
) sin(π −ζ
L
) 0
−sin(π −ζ
L
) cos(π −ζ
L
) 0
0 0 1


(24)
R
δ
L
=


1 0 0
0 cos(π + δ
L
) sin(π + δ
L
)
0 −sin(π + δ
L
) cos(π + δ
L
)


. (25)
By using the trigonometric rules for sine and cosine, the
rotation matrices R
ζ
L
and R
δ
L
simplify to
R
ζ
L
=


−cosζ
L
−sinζ
L
0
sinζ
L
−cosζ
L
0
0 0 1


(26)
and
R
δ
L
=


1 0 0
0 −cosδ
L
−sinδ
L
0 sinδ
L
−cosδ
L


. (27)
4.5 Force Description
The aerodynamics forces and moments produced by the
wings are assumed to act over the three bodies as a whole.
The aerodynamic forces,F
aero
, are defined as
F
aero
= F
x
¯
b
x
+F
y
¯
b
y
+F
z
¯
b
z
. (28)
The aerodynamic moments, M
aero
, are defined as
M
aero
= L
¯
b
x
+M
¯
b
y
+N
¯
b
z
. (29)
The generalized forces, Q
i
, are determined using the prin-
ciple of virtual work. Constraint forces do not perform virtual
work, therefore they are not considered in the formulations.
The generalized forces affecting the translation of the central
body are


Q
1
Q
2
Q
3


=


F
x
F
y
F
z


+ (m
1
+m
2
+m
3
)R
B


0
0
g


. (30)
The generalized forces/moments affecting the rotation of
the central body are


Q
4
Q
5
Q
6


=


L
M
N


+ (
¯
R+ ¯ ρ
c2
) ×(m
2
)R
B


0
0
g


+
. . . + (
¯
L + ¯ ρ
c3
) ×(m
3
)R
B


0
0
g


. (31)
The generalized forces Q
7
, Q
8
, and Q
9
are the control mo-
ments for the right wing, Q
10
, Q
11
, Q
12
are the control mo-
ments for the left wing. Q
8
and Q
11
control the angle of
attack of the wings, right and left wing respectively. Q
7
and
Q
10
control the flapping motion of the wings, while Q
9
and
Q
12
control the lagging angle of the wings.
4.6 Final Equations of Motion
The derived equations of motion, with all of the individual
pieces put together, are presented in vector/matrix notation.
The first three equations describe the translational velocity of
the central body.
m
1
˙
¯ v
1
+m
2
(
˙
¯ v
2
+
¨
¯ ρ
c2
) +m
3
(
˙
¯ v
3
+
¨
¯ ρ
c3
) =


Q
1
Q
2
Q
3


. (32)
Due to the continuous flapping of the wings, the rotational
dynamics of the central body are the most complex.
I
1
˙
¯ ω
1
+¯ ω
1
×I
1
¯ ω
1
+m
2
¯
R×(
˙
¯ v
2
+
¨
¯ ρ
c2
)+m
3
¯
L×(
˙
¯ v
3
+
¨
¯ ρ
c3
)+...
I
2
˙
¯ ω
2
+ ¯ ω
2
×I
2

¯ ω
2
+m
2
¯ ρ
c2
×
˙
¯ v
2
+...
I
3
˙
¯ ω
3
+ ¯ ω
3
×I
3
¯ ω
3
+m
3
¯ ρ
c3
×
˙
¯ v
3
=


Q
4
Q
5
Q
6


. (33)
The rotations of the right wing and and the left wing are de-
scribed by Equation 34 and Equation 35, respectively.
I
2
˙
¯ ω
2
+ ¯ ω
2
×I
2
¯ ω
2
+m
2
¯ ρ
c2
×
˙
¯ v
2
=


Q
7
Q
8
Q
9


(34)
I
3
˙
¯ ω
3
+ ¯ ω
3
×I
3
¯ ω
3
+m
3
¯ ρ
c3
×
˙
¯ v
3
=


Q
10
Q
11
Q
12


. (35)
5 SIMULATIONS AND RESULTS
5.1 Model
Through the derivations of the equations of motion, we
determined simplications could be made to the model in or-
der to reduce the coupling of the rotations of the wings with
the translational and rotational motion of the central body .
One of the simplications is to place the hinge points co-linear
with the y-axis of the central body. With the hinge points
along the y-axis, the x and z components of the
¯
R and
¯
L
vectors are identically zero.
5.2 Aerodynamic Data
The aerodynamic data is obtained by surrogate modeling
of two dimensional flow [5]. The model of the wings used to
obtained the time-averaged aerodynamics data is a flat plate,
with a prescribed chord length, c, and a span of 2c(for each
wing). The plate has a thickness of 2%. To model the mass
moments of inertia of the wings, all cross products of inertia
are assumed to be zero in the wing frame (when the δ and ζ
angles are identically zero). The resulting inertia tensor for
the right wing, with zero flapping motion, at the center of
mass of the wing in the wing-fixed frame, is
I
2
=


1
12
m
2
b
2
0 0
0
1
12
m
2
c
2
0
0 0
1
12
m
2
(b
2
+c
2
)


, (36)
where b is the span of the wing and c is the chord length.
The aerodynamic data [2, 5] prescribes the plunging am-
plitude as
ζ
R
(t) = h
a
sin(2πft)
ζ
L
(t) = h
a
sin(2πft)
. (37)
As previously stated, the wings motion is defined as positive
when the lagging motion is in the positive x direction in the
body frame. The angle of attack (or rotation of the wing) is
described as
α(t) = α
o
−α
a
sin(2πft + φ
α
). (38)
The phase lag between the translation and rotation of the wing
is denoted by φ
α
. In Equation 38, α
o
is the initial angle of
attack at the start of the flapping motion and α
a
is the time-
average angle of attack. For the aerodynamic model, the pa-
rameters h
a
, α
a
, and φ
α
are variable. For a given Reynolds
number, the frequency of the flapping can be calculated based
on the relationship, from reference [5, 2]
Re
hovering
=
(2πfh
a
)c
ν
. (39)
The current aerodynamic data is only for a normal hovering
mode, the wing translation is purely horizontal with the re-
spect to the B frame [2]. In the normal hovering mode, the
flapping angle (δ
R
, δ
L
) is identically zero for all time, t.
The control moments for each wing are calculated based
on the required accelerations, based on the flapping kinemat-
ics, and the inertia tensor in the B frame. They are related by
the relationship


Q
7
Q
8
Q
9


= I
2

¯ α
wing
+ ¯ ω
2
×I
2

¯ ω
2
(40)
where
¯ α
wing
=


0
α
a
(2πf)
2
sin(2πft + φ
α
)
−h
a
(2πf)
2
sin(2πft)


. (41)
SImilarly, the control moments for the left wing are


Q
10
Q
11
Q
12


= I
3

¯ α
wing
+ ¯ ω
3
×I
3

¯ ω
3
. (42)
5.3 Model Parameters
The initial simulations are conducted based off of the cur-
rent aerodynamic data and do not specifically replicate, or
model, any living organism. Based on the aerodynamic data
available in [5], a given set of wing kinematics is chosen. For
the purpose of the simulations, Case 11 from [5] is used. The
C
L
is 0.46 fromCase 11 and the plunging amplitude is related
to the chord length by the ratio [5]
2h
a
c
= 4.0. (43)
The chosen phase angle,φ
α
, is 90 degrees and the average
angle of attack, α
a
, is 62.5 degrees φ
α
and α
a
. The Reynolds
number for Case 11 is 100. The chord length is calculated
by choosing a frequency and utilizing Equation 39. Choos-
ing a frequency of 30 Hz, with a Reynolds number of 100,
results in a chord length, c, of 2 mm. The simulations are
intended to model a normal hovering mode. Based on the this
assumption, the time-averaged aerodynamic moments are as-
sumed to be identically zero. Additionally, the time-averaged
coefficient of drag, C
D
, is also zero, based on [5]. The lift
forces are assumed to counteract the gravity forces acting on
the central body.
5.4 Results
Figures 6 and 7 shows the position of the center of mass of
the MAV in the B frame. The simulations of the model pro-
duce symmetrical flapping. With flapping motion that con-
sists only of lagging and pitching motions, the center of mass
of the system symmetrically fore and aft in the B frame. The
shift of the center of mass is small. It is on the order of 10
−7
,
which is approximately four orders of magnitude smaller than
the vehicle scale.
The Euler angles of the central body are presented in Fig-
ures 8, 9 and 10. The results show that the central body
pitches up and down about the
¯
b
y
axis with each wing beat.
As the wings go forward, the nose pitches down. The hover
is unstable, as the angular velocity of the MAV continues to
increase. The bank angle, φ, and yaw angle, ψ, are effec-
tively zero. The values produced by the simulation are on the
order of 10
−17
and the error from zero is attributed to sim-
ulation error. The pitch angle change is approximately 0.01
radians per flapping cycle. The small effects on the body of
each wingbeat, coupled with unstable hover, are consistent
with the results in Reference [14].
!3 !2 !1 0 1 2 3
x 10
!7
!1
!0.8
!0.6
!0.4
!0.2
0
0.2
0.4
0.6
0.8
1
x(m)
z
(
m
)
Figure 6: MAV Center of Mass in B Frame - x v. z
!1 !0.8 !0.6 !0.4 !0.2 0 0.2 0.4 0.6 0.8 1
!3
!2
!1
0
1
2
3
x 10
!7
y(m)
x
(
m
)
Figure 7: MAV Center of Mass in B Frame - y v. x
0 5 10 15 20 25 30
!10
!8
!6
!4
!2
0
2
x 10
!17
t/T
!

(
r
a
d
ia
n
s
)
Figure 8: Yaw angle, ψ, of central body
0 5 10 15 20 25 30
!0.018
!0.016
!0.014
!0.012
!0.01
!0.008
!0.006
!0.004
!0.002
0
t/T
!

(
r
a
d
ia
n
s
)
Figure 9: Pitch angle, θ, of central body
0 5 10 15 20 25 30
!0.5
0
0.5
1
1.5
2
2.5
3
x 10
!17
t/T
!

(
r
a
d
ia
n
s
)
Figure 10: Roll angle, φ, of central body
6 CONCLUSION AND FUTURE WORK
The paper presented the derivation of the nonlinear, mul-
tiple body equations of motion for a flapping wing micro air
vehicle using D’Alembert’s Principle for Multiple Rigid Bod-
ies. The equations were derived using 12 generalized coordi-
nates and 12 quasi-velocities. Simulations were conducted
using time-averaged data for the aerodynamic forces and mo-
ments. Future Work will investigate the difference in perfor-
mance and simulation results between time-averaged data and
instantaneous lift and drag data. Previous work has shown in
[3, 15] that flexible wings produce more thrust. Humming-
birds control part of their flapping motion [16] by changing
the chord length, along the span of the wing, during flapping.
Simulations will be conducted with varying positions of the
hinge points, the conjecture is that stability may be improved
by ensuring that the time-variant system center of mass is for-
ward of the time-variant aerodynamic center. Investigation
of nonlinear and linear control schemes will be conducted.
Based on the work presented in [10], analytical equilibrium
solutions may not exist. Numerical solvers may be required
to obtain equilibrium solutions for desired maneuvers and tra-
jectories and to develop control schemes based on the numer-
ical solutions.
ACKNOWLEDGEMENTS
Thanks to Professor Donald T. Greenwood for assistance
with the development of the dynamical model. Thanks go
to Pat Trizila for assistance with the aerodynamic modeling
and integration of the data into the dynamic and kinematic
model.
REFERENCES
[1] M. Sun and J.K. Wang. Flight stabilization control of
a hovering model insect. The Journal of Experimental
Biology, 210:2714–2722, 2007.
[2] W. Shyy, Y. Lian, J. Tang, D. Viieru, and H. Liu. Aero-
dynamics of low Reynolds number flyers. Cambridge
University Press, New York, NY, 2008.
[3] W. Shyy, Y. Lian, J. Tang, H. Liu, P. Trizilia, B. Stan-
ford, L. Bernal, C. Cesnik, P. Friedmann, and P. Ifju.
Computational aerodynamics of low reynolds number,
plunging, pitching and flexible wings for mav applica-
tions. Acta Mechanica Sin, 24:351–373, 2008.
[4] H. Baruh. Analytical Dynamics. WCB McGraw-Hill,
Boston, MA, 2006.
[5] P. Trizilia, C. Kang, M. Visbal, and W. Shyy. Low
reynolds number hovering wing aeroydynamics: per-
formance, tip vortices, and induced jet. Submitted to
Journal of Fluid Mechanics, 2009.
[6] M. Sun and Y. Xiong. Dynamic flight stability of a hov-
ering bumblebee. The Journal of Experimental Biology,
208:447–459, 2005.
[7] G. Taylor and A. Thomas. Dynamic flight stability in
the desert locus Schistocerca gregaria. The Journal of
Experimental Biology, 206:2803–2829, 2003.
[8] B. Etkin and L.D. Reid. Dynamics of flight. John Wiley
and Sons, New York, NY, 1996.
[9] J. Jackson, R. Bhattacharya, and T. Strganac. Mod-
elling and suboptimal trajectory generation for a sym-
metric flapping wing vehicle. In Proceedings of the
AIAA Guidance, Navigation, and Control Conference
and Exhibit, Honolulu, Hawaii, USA, 18-21 August
2008. American Institute of Aeronautics and Astronau-
tics, 2008.
[10] W. Buler, L. Loroch, K. Sibilski, and A. Zyluk. Model-
ing and simulation of the nonlinear dynamic behavior of
a flapping wings micro-aerial-vehicle. In Proceedings
of the 42nd AIAA Aerospace Sciences Meeting and Ex-
hibit, Reno, Nevada, USA, 5-8 November 2004. Ameri-
can Institute of Aeronautics and Astronautics, 2004.
[11] M. A. Bolender. Rigid multi-body equations-of-motion
for flapping wing mavs using kane’s equations. In
Accepted for AIAA Guidance, Navigation and Control
Conference, 10-13 August 2009, Chicago, IL. American
Institute of Aeronautics and Astronautics, 2009.
[12] D. T. Greenwood. Principles of dynamics. Prentice
Hall, Upper Saddle River, NJ, 1988.
[13] D. T. Greenwood. Advanced dynamics. Cambridge Uni-
versity Press, 2006.
[14] M. Sun, P. Liu, and K.Wang. Dynamic flight stability of
a hovering hoverfly. In Proceedings of the Fifth Inter-
national Conference on Fluid Mechanics, Aug. 15-19,
2007, Shanghai, China. Tsinghua University Press and
Springer, 2007.
[15] H. Aono, S. Chimakurthi, C. Cesnik, H. Liu, and
W. Shyy. Computational modeling of spanwise flexi-
bility effects on flapping wing aerodynamics. In Pro-
ceedings of the 47th AIAA Aerospace Sciences Meeting
including the The New Horizons Forum and Aerospace
Exposition, Orlando, Florida, Jan 5-8, 2009, volume
2009-1270. American Institute of Aeronautics and As-
tronautics, 2009.
[16] B. Tobalske, D. Warrick, C. Clark, D. Powers,
T. Hedrick, G. Hyder, and A. Biewener. Three-
dimensional kinematics of hummingbird flight. The
Journal of Experimental Biology, 210:2368–2382,
2007.

as undertaken in the paper. 7]. It will enable the discovery of steady-state flight configurations and the eventual development of effective control schemes. is highly studied. The hinge point for both wings is the same and is positioned at the central body. Furthermore. and sweep (forward and back flapping). The dynamics model chosen is the standard. The flapping wing MAV is modeled as three rigid bodies (central body. equal change in angle of attack. the 3-2-1 Euler Angles tracking the orientation of each wing can be directly used as the time rate of change of the pitch. due to the lack of a body-fixed frame at the central body point mass. for at hovering the aerodynamics of the central body are neglected due to the absence of translational motion. β˙L . and not exactly implement a control scheme. in regards to insect flight. θ˙ ]. The central body is modeled as a point mass without mass moments of inertia. w. one for each wing. pitch (angle of attack).MAV. the method chosen requires the calculation of the reaction forces between the wings and the hinges. q and θ. However. The preliminary goal for vehicle size is a total mass of 30 grams and a total wingspan of approximately 15 centimeters. L R The above equations use standard aircraft notation. P. In contrast to the Taylor and Thomas model [7]. Buler. which is a hybrid of LaGrangian and Eulerian techniques [4]. listed below: q = [x. linearized aircraft dynamics that can be found in [8]. The method used to derive the equations of motion is D’Alembert’s Principle for Multiple Rigid Bodies. and α2 . and no central body frame. differential change in the angle of attack). Additionally. Sun and Wang concluded that the hovering motion can be stabilized using a combination of two of the four available controls (δΦ. The disadvantage of the two frames. The nonlinear dynamics will be incorporated with lift. Sun and Wang use stability and control derivatives calculated by CFD methods. The authors choose the equilibrium point for their analysis to be motionless hovering. The authors chose a dynamics model previously given in [6. In reference [10]. z. plunge and sweep angles. V. the coupling between p. Sun and Wang make further simplifications. derive the nonlinear equations of motion using the Gibbs-Appel Equations by using ten generalized coordinates. right wing. The wings need to accelerate to maintain the flight condition. 3 L ITERATURE R EVIEW The majority of the previous work in regards to flapping wing vehicles focuses on the aerodynamics and the interactions of the flapping wings with the fluid flow around the wings. left wing) with a rigid airframe and rigidly attached wings. R. q. The equilibrium point is determined by balancing the force and moment equations. A limited amount of work has been completed in regards to the development of dynamic models and control schemes to successfully operate a flapping wing micro-air vehicle. Sun and Wang [1] acknowledge that the field of aerodynamics. βL . Φ. Section 4 presents a detailed development of the model with relevant reference frames and the derivation of the nonlinear equations of motion. θR ] w = [U. respectively. so the assumptions and model simplifications seem justified. Section 5 presents the simulation efforts and results. but enhances the linearization and simplification of the model with improved accuracy of the aerodynamics. with no translational or rotational accelerations. Furthermore. Θ. versus the models used by Taylor and Thomas. βR . In reference [9]. Modeling the central body as a point mass effectively eliminates the rotational considerations of the central body and the coupling effects between the flapping motion of the wings and rotation of the central body. The paper is organized in the following manner. q. requires the tracking of the absolute translational and rotational velocity of not only the central body. The wings are modeled with three degrees of freedom: plunge (up and down flapping). and ten quasi-velocities. Section 3 is a brief literature review focusing on the dynamic models and control schemes for a flapping wing micro-air vehicle. y. by only accounting for the x and z translational motion of the model. and r is non-existent and there is no need to track the Euler Angles of the central body. The stability derivatives are calculated by only taking into account the motion of the wings. in addition to the pitching motion and pitch angle. but both wings as well. θ˙ . The authors use two body-fixed reference frames. Q. W. The combination of the nonlinear dynamics and aerodynamics should allow for accurate study of the motion of the vehicle. α1 . There is still rotational motion of the wing. δφ. et al. except in . The goal of the paper is to study trajectory guidance. The dynamics model does not require additional variables to track and transform the motions of the wing into an inertial frame. The main advantage to the method of deriving the equations of motion chosen in [9] is that it reduces the nonlinearities of the system. β˙R . so the nonlinearities associated with Euler’s Equations of Motion are still present in the equations of motion. drag and aerodynamic moment data from models developed in reference [5]. The main purpose of reference [1] is to produce a quantitative analysis of the stability of hovering flight for a model insect. change in mean position.. accompanied by the dynamic and aerodynamic models used in the simulation. given the flight conditions. The Sun and Wang model limits the model to have six degrees of freedom. Section 6 presents Conclusions and Future Work. Ψ. velocities calculated in each of the wings’ frames will need to be transformed back into the inertial frame. Taylor and Thomas used time average aerodynamics force and moment calculations for a wing beat cycle. the authors derived the equations of motion for a flapping wing vehicle using Eulerian methods. θL . change in position.

forward and back (lagging) translation of the wings.motion about the longitudinal axis of the central body) and θ (used to denote the angle of attack of the wings). the frame is aligned with the ¯x vector b through the nose of the micro-air vehicle. The method of deriving the equations of motion results in the preservation of the numerous. alternatively. The motion of the wings is a function of time and a control input. ¯ ¯ the magnitude and components of R and L are constant. is the modeling of the central body as two separate rigid bodies. B. A detailed description of the method can be found in [12. with directions parallel to the central body B frame. ¯y is perpendicular b to ¯x and pointed out of the right side of the central body. The body-fixed frame is the same as for fixed-wing aircraft and is presented in Figure 1.the case of β (used to denote the flapping angle . The main difference from the derivation in reference [11]. limit the flapping kinematics of their model and do not account. Bolender treats the wings in two separate manners. The chosen method is also briefly covered in [4]. after outlining the derivation of the equations of model and the aerodynamic model. The vectors ρc2 and ρc3 denote the vectors from the hinge ¯ ¯ points on the central body to the respective centers of mass for each wing (¯c2 for the right wing and ρc3 for the left wing). 2nd-order system (or. based on the as¯ sumption of rigid bodies and rigidly attached wings. and b ¯z pointed downward out of of the bottom of the micro-air b vehicle. The tail section is limited to one degree of freedom about the hinge point between body B and body T .. The authors use a linear quadratic regulator problem to determine a nominal control input for the flapping wings of the vehicle. The GibbsAppell Equations. The central body has a main body. is n i=1 ˙ ¯ ˙ ˙ [mi (vi +¨ci )·¯ij +(Ii ·ω i +¯i ×Ii ·¯i +mi ρci ×vi )·βij ] = Qj . A wingfixed frame is attached to each wing at the hinge point. The statement of D’Alembert’s Principle. but do not provide an analytical solution for the linearization. 4 D ERIVATION OF THE N ONLINEAR E QUATIONS OF M OTION presented in [11]. The flapping wing vehicle is modeled as a system of three rigid bodies (n = 3): a central fuselage with two rigid wings attached at ideal hinges. in [12]. the wings are not treated as having separate degrees of freedom and are considered to be holonomically constrained to the central body. The authors continue with an analysis of the control of the linearized system previously derived. a twenty degree of freedom. At the hinge points. and a tail section. A desired trajectory is assumed and a linearization. A body-fixed frame.1 D’Alembert’s Principle for Multiple Rigid Bodies The method chosen to derive the equations of motion is D’Alembert’s Principle Extended to Multiple Rigid Bodies. Since the central body is modeled as a rigid body. In reference [11]. Buler. frame B. ρ ¯ The magnitude of the ρci vectors is constant. Figure 1: Vehicle Model ¯ ¯ The vectors R and L denote position vectors from the central body center of mass to the hinge/attachment points of the wings. We ¯ and L as ¯ define the individual components of R ¯ R= and ¯ L= Rx Lx Ry Ly Rz Lz . two central body-fixed frames are attached. presented in [10]. and associated Jacobian matrix. and Kane’s Equations. In the end. is numerically calculated based on the said trajectory. However. based on the solution of a Newton-Raphson solver to produce the necessary control input vector for the desired output vector. The other treatment accounts for the mass of the wing in the dynamics of the central body. Bolender derives the nonlinear equations of motion for a flapping wing MAV utilizing Kane’s Equations. T . or allow for. the authors produce simulations of the dynamics of a flapping vehicle/animal. et al. are both extensions of D’Alembert’s Principle [4] and have their own strengths and weaknesses [11]. The HR frame is fixed with origin at the right wing hinge point. The . ¯ ρ γ ¯ ω ω ¯ ¯ ¯ (1) where i is the number of rigid bodies and j is the number of generalized coordinates. The paper provides coverage of the various control methods at the authors’ disposal and their respective shortcomings. their model restricts the stroke-plane of the wing beat to a single plane. One treatment ignores the combined mass of the wing. 13]. The model results in a ten degree of freedom. is attached to the central body at its center of mass. The wings and central body are modeled as rigid airframes. The HL frame is fixed with origin at the left wing hinge point. Strictly speaking. in either case. The wings’ motion is treated as a prescribed motion. (2) (3) 4. inherent non-linearities present in a flapping wing micro-air vehicle. The second half of the paper. 1st-order system). presents an outline of Linear Control Theory. The tail section is used for pitch control of the flapping wing MAV. from other models.

αL . p. Second. δ˙ . The frames WR and WL and the vectors R. ζL ] and the quasi-velocites. are not depicted. p. ζR .δR and δL 4. w. q. Figures 3. b ζR and ζL are positive in a forward flapping motion (positive ¯x ). the requirement to calculate the reaction forces between the wings and the central body is eliminated. ψ. αL . fixed at the origin of the system. θ. The reference points for the right and left wing are chosen to be the hinge attachment points. The translational and rotational velocities of each of the rigid bodies are ¯ v1 = u¯x + v¯y + w¯z b b b ¯ v2 = v1 + ω1 × R ¯ ¯ ¯ ¯ v3 = v1 + ω1 × L ¯ ¯ ¯ ∂uj (8) . WR and WL . δL . ζL . α˙L . ρc2 . ζR . with a constant radius. The hinge frames. but is simply the mirror image of Figure 4. The position of the centers of mass of each wing are expressed by the following vectors: ρc2 = ρc2 [cosδR sinζR¯x + cosδR cosζR¯y + sinδR¯z ] (4) ¯ b b b Developing equations of motion using D’Alembert’s Principle for Rigid Bodies requires the selection of reference points for each rigid body. by taking the hinge points as reference points. The constant radius is the magnitude of the vectors. and 5 show the vehicle model from different view points and the relationship of the wing frames. are positive in a downward motion (positive ¯z ). Dashed lines represent a negative direction in the respective reference frame. the mass moments of inertia of the wings need to be initially calculated about the reference points. Third. r. The left side of the vehicle. In this formulation. θ. To R R L L summarize. WR and WL . ¯ ¯ HR and HL . q. First. 4. αR . γij = ∂ui . L. is important when biological flyers transition from hovering to forward flight [2]. ¯ j ¯ ¯ βij = ∂ ωi . ζR . located on the central body. (7) (6) Figure 2: Wing Frames The motion of the center of mass of the wings is tracked similar to a mass moving in spherical coordinates. v. which is coincident with the origin of the WR . δL . and the central body frame B. are ˙ ˙ ˙ ˙ ˙ ˙ uj = [u. ¯ ¯ In order to successfully develop the equations of motion. the axes of the wing-fixed frames. v. δR . φ (3-2-1 Euler Angles of central body). (5) ¯ b b b The importance of flapping and lagging angles. is not depicted. rigid body 1. and ρc3 are presented in Figure 2. ρc3 = ρc3 [cosδL sinζL¯x − cosδL cosζL¯y + sinδL¯z ]. αR . αR . δR . Y . ρc2 and ¯ ρc3 .origin of the winged fixed frames is the hinge point and when the wings at are their respective initial positions. The mass moments of inertia will remain constant in the wing-fixed frames and do not need to be translated into the B frame [12]. HL and WL frames. The twelve quasi-velocities are the following: u. where i denotes the rigid body and j denotes the generalized speeds. are parallel with the central body-fixed frame B. HR and HL . δL . HR . q. The center of mass of the wing is assumed to be along ¯ the y-axis of the hinge frames when the wing angles are identically zero. and ζ˙ . and are identical to the hinge frames. Y. Z. αL . the absolute velocity of the wings does not need to be tracked. δR . are qj = [X. the origin of the respective motion is the hinge point. Figure 3: Front View . α˙R . ζL ]. and αL . is it’s respective center of mass. The flapping angles. δ˙ . and the associated degrees of freedom. The reference point for the central body. the twelve generalized coordinates are chosen to be the following: X. There are three main advantages to selecting D’Alembert’s Principle. The vectors tracking the center of mass of the wings in the hingedfixed frames are the vectors ρc2 and ρc3 . r. ψ. and the angular velocity coefficients. but are b parallel to the B frame. uj . Z (inertial position of central body). the generalized coordinates and quasi-velocities need to be specified. w. The required ac¯ ¯ ¨ ¨ celeration vectors of ρc2 and ρc3 are derived using diligent ¯ ¯ application of the chain rule. the generalized coordinates. ζ˙ . φ. The lagging angles.2 Velocity and Angular Velocity Coefficients The first step in the derivation is to determine the velocity ∂v ¯ coefficients. δR and δL . In this case. by choosing reference points on the central body.

β2j and β3j . For quasi-velocities j = 4 .ζR and ζL and ω1 = p¯x + q¯y + r¯z ¯ b b b ˙ b ˙ b ω2 = (p + δR )¯x + (q + αR )¯y + (r − ζR )¯z . for j = 1 .2 Izz.3 −Ixz. The velocity coefficients for the right wing. ¯ β2j is identically zero for j = 10 . it may be necessary in the use of the model to rotate aerodynamic forces generated by the flapping of the wings from the WR and WL frames into the B frame.1 0 Izz.2 Iyy. . −Ixz. the angular velocity coefficients are ¯ β2j = ¯ β3j = Figure 4: Right Side View . is ¯ ¯ ¯ ˙ ˙ ˙ ¯ ¯ ¯ v2 = v1 + ω 1 × R + ω1 × v1 + ω1 × (¯ 1 × R). (16) In regards to the left wing.1 0 I1 =  (19) −Ixz. (17) Figure 5: Top View . . .2  −Ixz. β1j for generalized ¯ speeds j = 1 . 12 are ¯ β3j = −¯x b ¯y b ¯z b .αR ¯x b ¯x b ¯y b ¯y b ¯z b ¯z b . the acceleration of the right wing. and the angular velocity coefficients. (14) (15) For the right wing. the angular velocity coefficients are ¯ β2j = ¯x b ¯y b −¯z b . .3  (20) The velocity and angular velocity coefficients of the central body for coordinates j = 7 . . (10) (11) Inertia Tensors For the central body. ¯ ˙ b ˙L )¯x + (q + αL )¯y + (r + ζL )¯z ˙ b ω3 = (p − δ b ¯ ˙ b The accelerations of the three rigid bodies are obtained by differentiating the translational and rotational velocities and making proper use of the transport theorem [4]. ¯ γ1j . respectively. for j = 7 . . RR is the rotation matrix .3 −Ixy. ¯ ¯ ω 4. For the left wing. . an x − z plane of mass symmetry is assumed. . For exam˙ ¯ ple. .3 Izz. .2 −Ixy. 3. The velocity coefficients. 12.1 0 −Ixz. .1  0 Iyy. . with respect to coordinates j = 7 .The velocity coefficients for the left and right wing.1  Ixx.2 I2 =  −Ixy. v2 in the central body frame. 9. 12 are identically zero. . . . γ2j . . 12 are identically zero. .3 −Iyz.2 −Iyz. and the left wing. . The angular velocity co¯ ¯ efficients of the right and left wings.3 (9) (18) The velocity and angular velocity coefficients are ob¯ tained by taking the partial derivatives of vi and ωi with re¯ spect to the generalized speeds.3 −Iyz. .2 −Ixz. 6 are ¯ γ2j = [¯x ¯ b −Rz ¯y + Ry ¯z b b −Lz ¯y + Ly ¯z b b γ3j = [¯x b ¯y b ¯y b ¯z . .3 I3 =  −Ixy. 6 of the central body are γ1j = ¯ and ¯ β1j = ¯x b 0 0 ¯y b ¯z b 0 ¯x b 0 ¯y b 0 0 ¯z b . No planes of mass symmetry are assumed for either wing during the model development. b − Ry ¯x + Rx¯y ] (12) b b (13) 4. ¯ β3j is identically zero for j = 7 . ¯ γ3j . are identically zero for j = 1 . For the right wing.3  . b ¯z . the angular velocity coefficients for j = 10 . As a result.2 −Iyz. 9. . b b Rotation Matrices Although it is not necessary to continuously calculate the mass moments of inertia for the wings in the B frame. 6.3 Iyy.4 (21) Rz ¯x − Rx¯z b b Lz ¯x − Lx¯z b b − Ly ¯x + Lx¯y ]. . . the resulting mass moments of inertia matrices for each rigid body are the following:   Ixx. .2   Ixx.

¯ ¯ ¯ ¯ ¯   Q4 ˙ ˙ I3 ω 3 + ω3 × I3 ω3 + m3 ρc3 × v3 =  Q5  ... RαR .5 Force Description The aerodynamics forces and moments produced by the wings are assumed to act over the three bodies as a whole. (23) The rotation matrix RαL has the same structure as it’s counterpart on the right wing. Constraint forces do not perform virtual work. + (L + ρc3 ) × (m3 )RB  0  .. (22) A similar procedure can be utilized to obtain the rotation of aerodynamic forces and moments generated by the left wing into the B frame. The aerodynamic forces. while Q9 and Q12 control the lagging angle of the wings. . The rotation matrix for the left wing is RL = RδL RαL RζL . Q8 . the rotational dynamics of the central body are the most complex. 4. b b b (29) (28) The rotations of the right wing and and the left wing are described by Equation 34 and Equation 35. therefore they are not considered in the formulations. the rotation matrices RζL and RδL simplify to   −cosζL −sinζL 0 −cosζL 0  RζL =  sinζL (26) 0 0 1 and 1 RδL =  0 0  0 −cosδL sinδL 0 −sinδL  . Q11 . ¯ ¯ ρ ¯ ¯ ρ ˙ ˙ ˙ ¯ I1 ω 1 +¯1 ×I1 ω1 +m2 R×(v2 +¨c2 )+m3 L×(v3 +¨c3 )+.. are defined as Faero = Fx¯x + Fy ¯y + Fz ¯z . as 3-2-1 Euler Angles results in the calculation of RR . respectively. Q12 are the control moments for the left wing. with all of the individual pieces put together. Treating the angles describing the orientation of the right wing. Q10 . Maero . b b b The aerodynamic moments. ¯ ω ¯ ¯ ˙ ˙ I2 ω 2 + ω2 × I2 ω2 + m2 ρc2 × v2 + . αR . ¯ ¯ ¯ ¯ ¯ Q6 (33) 4. .6 Final Equations of Motion The derived equations of motion. are presented in vector/matrix notation. Qi .  Q1 ˙ ˙ ˙ ¯ ¯ ρ ¯ ¯ ρ ¯ m1 v1 + m2 (v2 + ¨c2 ) + m3 (v3 + ¨c3 ) =  Q2  . The generalized forces affecting the translation of the central body are       Fx 0 Q1  Q2  =  Fy  + (m1 + m2 + m3 )RB  0  . are defined as Maero = L¯x + M ¯y + N ¯z . Q7 and Q10 control the flapping motion of the wings. (30) Fz g Q3 The generalized forces/moments affecting the rotation of the central body are       L 0 Q4 ¯ ¯  Q5  =  M  + (R + ρc2 ) × (m2 )RB  0  + g Q6 N   0 ¯ ¯ . right and left wing respectively. ζR . −sin(π + δL ) cos(π + δL ) (25) The generalized forces. and Q9 are the control moments for the right wing.Faero . −cosδL  (27) Due to the continuous flapping of the wings. The first three equations describe the translational velocity of the central body. The rotation matrices RζL and RδL are different and are defined as   cos(π − ζL ) sin(π − ζL ) 0 RζL =  −sin(π − ζL ) cos(π − ζL ) 0  (24) 0 0 1 RδL 1 = 0 0   0 0 cos(π + δL ) sin(π + δL )  . (35) ¯ ¯ ¯ ¯ ¯ Q12 .   cosζR sinζR 0 RζR =  −sinζR cosζR 0  0 0 1   cosαR 0 −sinαR  0 1 0 RαR =  sinαR 0 cosαR   1 0 0 RδR =  0 cosδR sinδR  0 −sinδR cosδR RR = RδR RαR RζR . Q8 and Q11 control the angle of attack of the wings.of the WR frame into the HR frame. (31) g The generalized forces Q7 . and δR . (32) Q3  By using the trigonometric rules for sine and cosine.   Q7 ˙ ˙ I2 ω 2 + ω2 × I2 ω2 + m2 ρc2 × v2 =  Q8  ¯ ¯ ¯ ¯ ¯ (34) Q9   Q10 ˙ ˙ I3 ω 3 + ω3 × I3 ω3 + m3 ρc3 × v3 =  Q11  . are determined using the principle of virtual work.

The CL is 0. The Reynolds number for Case 11 is 100. ν (39) The chosen phase angle.5 degrees φα and αa .2 Aerodynamic Data The aerodynamic data is obtained by surrogate modeling of two dimensional flow [5]. 0 0 (36) I2 =  12 m2 c 1 2 2 0 0 m2 (b + c ) 12 where b is the span of the wing and c is the chord length. δL ) is identically zero for all time. Choosing a frequency of 30 Hz. (37) ζL (t) = ha sin(2πf t) The control moments for each wing are calculated based on the required accelerations.0. the parameters ha .φα . CD . the control moments for the left wing are   Q10  Q11  = I3 αwing + ω3 × I3 ω3 . To model the mass moments of inertia of the wings. based on the flapping kinematics. With flapping motion that consists only of lagging and pitching motions. The Euler angles of the central body are presented in Figures 8. is also zero. The results show that the central body 5. from reference [5. The model of the wings used to obtained the time-averaged aerodynamics data is a flat plate.5 5. The shift of the center of mass is small. with a Reynolds number of 100. The plate has a thickness of 2%. 9 and 10. αa . The simulations are intended to model a normal hovering mode. the wing translation is purely horizontal with the respect to the B frame [2]. based on [5]. we determined simplications could be made to the model in order to reduce the coupling of the rotations of the wings with the translational and rotational motion of the central body . or model. For the aerodynamic model. The aerodynamic data [2. all cross products of inertia are assumed to be zero in the wing frame (when the δ and ζ angles are identically zero).46 from Case 11 and the plunging amplitude is related to the chord length by the ratio [5] 2ha = 4. For the purpose of the simulations. 5. In the normal hovering mode. The chord length is calculated by choosing a frequency and utilizing Equation 39. Additionally. and φα are variable. and a span of 2c(for each wing). The angle of attack (or rotation of the wing) is described as α(t) = αo − αa sin(2πf t + φα ). of 2 mm. They are related by the relationship   Q7  Q8  = I2 αwing + ω2 × I2 ω2 ¯ ¯ ¯ (40) Q9 where  0 2 =  αa (2πf ) sin(2πf t + φα )  . Based on the aerodynamic data available in [5]. is   1 2 0 0 12 m2 b 1 2 .1 Model S IMULATIONS AND R ESULTS Through the derivations of the equations of motion. the flapping angle (δR . 2 −ha (2πf ) sin(2πf t)  αwing ¯ (41) SImilarly. The resulting inertia tensor for the right wing. Case 11 from [5] is used. Results Figures 6 and 7 shows the position of the center of mass of the MAV in the B frame. For a given Reynolds number. is 90 degrees and the average angle of attack. One of the simplications is to place the hinge points co-linear with the y-axis of the central body. a given set of wing kinematics is chosen.4 The current aerodynamic data is only for a normal hovering mode. with a prescribed chord length. αa . ¯ ¯ ¯ Q12 5.3 (42) Model Parameters The initial simulations are conducted based off of the current aerodynamic data and do not specifically replicate. t. With the hinge points ¯ ¯ along the y-axis. The lift forces are assumed to counteract the gravity forces acting on the central body. the wings motion is defined as positive when the lagging motion is in the positive x direction in the body frame. c. c. the x and z components of the R and L vectors are identically zero. and the inertia tensor in the B frame. is 62. which is approximately four orders of magnitude smaller than the vehicle scale. (38) The phase lag between the translation and rotation of the wing is denoted by φα . The simulations of the model produce symmetrical flapping. 2] Rehovering = (2πf ha )c . at the center of mass of the wing in the wing-fixed frame. c (43) As previously stated. . results in a chord length. the frequency of the flapping can be calculated based on the relationship. Based on the this assumption. the time-averaged aerodynamic moments are assumed to be identically zero. any living organism. αo is the initial angle of attack at the start of the flapping motion and αa is the timeaverage angle of attack. the center of mass of the system symmetrically fore and aft in the B frame. It is on the order of 10−7 . the time-averaged coefficient of drag. 5] prescribes the plunging amplitude as ζR (t) = ha sin(2πf t) . In Equation 38. with zero flapping motion.

# $() $(* $(+ $(" ! !!&!!$ !!&!!) !!&!!( !-.01 radians per flapping cycle. the nose pitches down./ " 6 C ONCLUSION AND F UTURE W ORK !! !' !( !! !"#$ !"#% !"#& !"#' " . x $ )*#! !#+ ! !$ !( !' !& !#! ! " #! #" . as the angular velocity of the MAV continues to increase. Hummingbirds control part of their flapping motion [16] by changing the chord length. The bank angle. The values produced by the simulation are on the order of 10−17 and the error from zero is attributed to simulation error. $! $" %! ( )*!" !+ Figure 10: Roll angle. φ. analytical equilibrium solutions may not exist. ψ. Simulations were conducted using time-averaged data for the aerodynamic forces and moments. Previous work has shown in [3. of central body /. are effectively zero./01/234 !$() !# !! #&" !" !# $ %. The hover is unstable. Future Work will investigate the difference in performance and simulation results between time-averaged data and instantaneous lift and drag data.-. Numerical solvers may be required to obtain equilibrium solutions for desired maneuvers and trajectories and to develop control schemes based on the numerical solutions. $! $" %! Figure 8: Yaw angle. of central body The paper presented the derivation of the nonlinear. Based on the work presented in [10]. The equations were derived using 12 generalized coordinates and 12 quasi-velocities. b As the wings go forward. !*/01231456 . and yaw angle. the conjecture is that stability may be improved by ensuring that the time-variant system center of mass is forward of the time-variant aerodynamic center.-. # " %&#$ ! !' # !&" Figure 6: MAV Center of Mass in B Frame . % $ !$(" !$(+ $ !$(* $&" '(#! !#) !(-.x v.y v. $! $" %! Figure 9: Pitch angle. multiple body equations of motion for a flapping wing micro air vehicle using D’Alembert’s Principle for Multiple Rigid Bodies. during flapping. are consistent with the results in Reference [14]. Simulations will be conducted with varying positions of the hinge points. along the span of the wing./0120345 !!&!!' !!&!# !!&!#$ !!&!#) !!&!#( !!&!#' ! " #! #" *+. 15] that flexible wings produce more thrust. coupled with unstable hover./ "#' "#& "#% "#$ ! Figure 7: MAV Center of Mass in B Frame . The pitch angle change is approximately 0. φ. ψ. Investigation of nonlinear and linear control schemes will be conducted. The small effects on the body of each wingbeat. z ! !!&" ! " #! #" *+.-. θ. of central body ' ! )-.-.pitches up and down about the ¯y axis with each wing beat.

Wang. T. Greenwood for assistance with the development of the dynamical model.ACKNOWLEDGEMENTS Thanks to Professor Donald T. Trizilia. 2008. In Proceedings of the 47th AIAA Aerospace Sciences Meeting including the The New Horizons Forum and Aerospace Exposition. Shyy. Sun. Friedmann. Bernal. In Proceedings of the 42nd AIAA Aerospace Sciences Meeting and Exhibit. NY. Liu. [4] H. In Accepted for AIAA Guidance. 2009. Dynamic flight stability in the desert locus Schistocerca gregaria. Threedimensional kinematics of hummingbird flight. pitching and flexible wings for mav applications. plunging. 210:2368–2382. . Dynamic flight stability of a hovering hoverfly. 208:447–459. Visbal. Shyy. and W. Baruh. 2009. New York. 2009. Modeling and simulation of the nonlinear dynamic behavior of a flapping wings micro-aerial-vehicle.Wang. American Institute of Aeronautics and Astronautics. Upper Saddle River. The Journal of Experimental Biology. R EFERENCES [1] M. [9] J. [2] W. Zyluk. L. American Institute of Aeronautics and Astronautics. Tsinghua University Press and Springer. 1996. 2006. Hawaii. USA. 206:2803–2829. [10] W. Aug. USA. Lian. Aono. Nevada. Tang. The Journal of Experimental Biology. MA. [15] H. H. 2007. Lian. [12] D. T. Thomas. Sibilski. Jan 5-8. Navigation and Control Conference. Ifju. Hyder. Powers. Sun and Y. NY. Xiong. J. Rigid multi-body equations-of-motion for flapping wing mavs using kane’s equations. G. Biewener. 5-8 November 2004. Chimakurthi. 2007. Reno. IL. 2006. [3] W. American Institute of Aeronautics and Astronautics. C. and A. and K. Computational aerodynamics of low reynolds number. Buler. 2008. and induced jet. P. American Institute of Aeronautics and Astronautics. Strganac. B. Y. 10-13 August 2009. and W. and A. Taylor and A. [6] M. Shyy. [7] G. Clark. C. D. Cesnik. Viieru. Navigation. Warrick. Cambridge University Press.K. In Proceedings of the AIAA Guidance. Dynamic flight stability of a hovering bumblebee. Trizilia. Liu. and H. and T. Computational modeling of spanwise flexibility effects on flapping wing aerodynamics. P. tip vortices. New York. H. J. 2007. C. 2009. Kang. M. John Wiley and Sons. L. Etkin and L. Bhattacharya. D. Greenwood. Stanford. Honolulu. [16] B. The Journal of Experimental Biology. Submitted to Journal of Fluid Mechanics. Cambridge University Press. Sun and J.D. In Proceedings of the Fifth International Conference on Fluid Mechanics. Aerodynamics of low Reynolds number flyers. Orlando. Boston. [11] M. [5] P. 210:2714–2722. Acta Mechanica Sin. Y. Tobalske. 24:351–373. Modelling and suboptimal trajectory generation for a symmetric flapping wing vehicle. Reid. and Control Conference and Exhibit. 1988. 2004. 2007. Chicago. Hedrick. [8] B. [14] M. R. China. Principles of dynamics. Advanced dynamics. Shyy. Tang. D. 2005. The Journal of Experimental Biology. Dynamics of flight. Prentice Hall. K. Liu. S. [13] D. Florida. P. 18-21 August 2008. Flight stabilization control of a hovering model insect. 2003. 15-19. C. WCB McGraw-Hill. Liu. Loroch. Analytical Dynamics. Cesnik. and P. Greenwood. Shanghai. volume 2009-1270. Jackson. NJ. Low reynolds number hovering wing aeroydynamics: performance. 2008. T. Thanks go to Pat Trizila for assistance with the aerodynamic modeling and integration of the data into the dynamic and kinematic model. A. Bolender.

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->