Professional Documents
Culture Documents
a, *
KEYWORDS
Gain self-scheduled; H1 robust control;
Linear parameter varying; Morphing aircraft;
Wing transition
Abstract
This
article
investigat
es
gain
selfscheduled
H1 robust
control
system
design for
a tailless
fold-ingwing
morphing
aircraft in
the wing
shape
varying
process.
During
the wing
morphing
phase,
the
aircrafts
dynamic
response
will
be
co
mb
at
1. Introduction
mis
sion
Morphing aircraft can automatically s by
change its aerodynamic configuration to usin
adapt to different flight environments and g
advanced
materials
1,2
Pro
duct
ion
Co
mp
are
d
to
con
ven
tion
al
fixe
dwin
g
airc
1000-9361 2013 Production
a
n
access under CC
BY-NC-ND license.
response
of
morphing aircraft
will be governed
by
time-varying
aerodynamic
forces
and
moments
which
are related to the
wing
shape.
Dynamic models
for
morphing
aircraft must take
into account the
dynamic
http://dx.doi.org/10.1016/j.cja.2
013.06.004
910
T. Yue et al.
exploits
all
available
coupling between wing area and mass distribution change, aerody- information on h(t) to adjust
namics, structure, and control forces. Thus, the wing morphing the current plants dynamics.
57
aircraft needs a long time to achieve a new state of level flight without
P(h) with state-space
A
control. Consequently, the flying quality of morp-hing aircraft may
equations
Co 82 C1ii
deteriorate and flight safety may be threatened. To guarantee
C
>
2i
satisfactory flying quality, a morphing aircraft capa-ble of shape 8 x Ahx B1hw B2hu
>
11
12
reconfiguration requires a flight control system to maintain stability > C D D
:
< y 2 h x 21 h w 22
during the morphing transition phase.
<6
C1
>
1i
12
22
11i
21i
2i
12i
22i
7 2
5
h
3
7
; i 1; 2;
n
at the vertices h = hi of the
x e R , disturbance
mainly focuses on static configuration control.
There is a with state
parameter polytope.
m1
m2
w
2
R
,
control
input
u
2
R
notable lack of published work on flight control design for
p1
morphing aircrafts wing transition phase. When the aerody-namic , performance output z 2 R ,
p2
shape is changing, the controller should adapt online to maintain and measured output y 2 R .
The
plants
P(h)
can
be
stability in shape transition between different con-figurations. described as
Since the wing transition process is time-varying, it is obviously
P P
quite difficult to use general control techniques to control such a
11
12
2 Ah B1
P
complex dynamic behavior. One possible ap-proach is to assume
P P21
22
C1h D
that the time-varying dynamics could be represented by a linear
parameter varying (LPV) plant model that approximately captures
4
We seek an LPV controller K(
the wing transition phases com-plex behavior. In this LPV
framework, nonlinear dynamic equations of morphing aircraft xK
AKhxK BKhy
C h xK
DK
h y
could be simplified and trans811
C2
21
guarantees
H1
transition control. The gain self-scheduled control technique performance for the closedbased on the LPV model can rapidly change controller param- loop system in Fig. 1. The
eters to adapt to the aircrafts dynamic response. Additionally, this transfer function from w to z
control approach can guarantee stability of the closed-loop of the closed-loop system is
X
where ai is the weighting ratio.
For the LPV system Eq. (2), if:
the
parameter
The class of finite dimensional linear systems whose state-space (A)
dependence is affine, that is,
matrixes depend continuously on a time varying parameter vector the state-space matrices A(h),
h(t) is called linear parameter varying. In state-space form, an B1(h), B2(h), C1(h), C2(h),
LPV system model can be expressed as
D11(h), D12(h), D21(h) and
D22(h) depend affinely on h;
x Ahtx Bhtu
(B) the tine-varying parameter
y Chtx Dhtu
description
is
where the state matrixes A, B, C, Dan
explicit
vary with h(t), u is the con-trol input.function of h(t).
It is a time-varying vector that can Here
we
consist
of
system
outputs,focus on the
exogenous inputs, or combinationsstandard
H1
of both. LPV systems are linear control problem
systems where the state-spaceof LPV system.
As shown in
Fig.
1,
the
resulting
LPV
controller K(h)
F
i
;r 9
>
=
>
11
D21
When the
wing based
shape on
is an LPV
handmodel
side
Gain self-scheduled H1 control for morphing aircraft in the wing transition
process
911
the
dynamic
In the sequel the standard H 1 control problem can be de-changing,
of Eq. (11) can be
fined as: finding an internally stabilizing controller K that response of the morphing
expressed as
makes the closed-loop H1 gain from w to z less than c. Ifaircraft will be dependent on
8
aero-dynamic
Tzw(s) denotes the closed-loop transfer function w to z, thistime-varying
>
< Fx D cos a L sin a T
control objective can be formalized as
forces and moments, which
are both functions of the wing
16
shape. In Ref.
the folding15
wing
morphing
aircraft is reIt is known from Ref.
that there exists an LPV controller
Eq. (4) guaranteeing quadratic H1 performance c along allgarded as a variable geometry
parameter trajectories, if and only if there exist two symmetricrigid body, and a six-DOF
matrices R e Rnn and S e Rnn satisfying the system of 2r + 1 non-linear dynamic model in
linear matrix inequalities (LMI).
the wing folding process is
If h is in the polytope Co{h1, h2, . . ., hr}, that is,
founded. It also shows that
( r
)
the aerodynamic forces and
r
h 2 H : i 1 aihi : ai P 0; i 1 ai 1
9 moments of the folding-wing
morphing
aircraft
almost
X
X
linearly vary with the wing fold
Then the LPV controllers state-space matrices are given by
angle hfold in wing folding.
A B
r
Ki
Ki
AK h t
BK h t
Hence, we can regard the
D
a t
:
CKht DK h t
Ki
10 morphing process as an LPV
i 1 i CKi
system. In this LPV frame
X
work, the nonlinear dynamic
where AKi, BKi, CKi, DKi can be obtained off-line, and AK(h(t)),equations of the morphing airBK(h(t)), CK(h(t)), DK(h(t)) will update dependently on the
craft will be simplified and
parameter h(t) in real time.
transformed to an LPV model
LPV
modeling
3. Longitudinal LPV model in wing shape varying for folding- by
1720
approaches.
wing morphing aircraft
kTzwsk1 < c
mg sin h
> Fz D sin a L cos a mg cos h
My MA gSz sin h
where D, L, T are the drag,
lift and thrust, MA is the
pitching moment caused by
aerodynamic forces, a the
angle of attack, h the pitch
angle, Sz the z-component
of static momentum S = * r
dm in fixed-body axes, and
g the acceleration due to
gravity, respectively. The lift,
drag, and pitching moments
in Eq. (12) are given as
mqVz
Sz 2
Sz
>
>
z z
<
>
>
>
>
>
>
>
>
SC
MA
qV2 ScC m
>
>
>
where
is the air density, S the wing ar
>
q
:
>
<
Cm
CD a; V
D0
CDaa CDa2 a2
q Sz
Sz
Fz
_
2qSz
Fx
I yq
>
>h
>
_
mIy
Sz
SzqV
z
Cm a; q; V; de Cm0 Cm a C D
mV
>
>
:
12
mqVz
>
>
>
<
>
>
>
>
>
>
:
2
mIy Sz2
>D
>
>
>
>
>
>
>
>
>
8
S qV
Iyq
S
>
mIy
>
>
>
>
1
> Vz
mqVx
>
> m
>
>
2qSz Fx
>
>
>
L 1 qV2SCL
8 Vx mIy
>
>
912
T. Yue et al.
u
2w 3
6q 7
6_ 7
6
u u0
Dde
2 w w0 3
6
Aht
7 Bht DdT Wht 15
q
6h
h0 7
6
4 5
4
5
where u, w are the x and z components of airspeed in fixed-body
axes, u , w the x and z components of balanced airspeed in
0 0
w , DdT is the
propulsion control, and h0 the initial pitch angle. W(h(t)) is the force
and moment variations affected by wing folding, which can be
considered as a disturbance source in the wing morphing dynamic
response. In Eq. (15) the state-space matrices vary with
4. Multi-loop controller
design and simulation
for the wing morphing
process
We know from the dynamic
response of the folding-wing
morp-hing aircraft in transition
phase that the altitude and speed do change a lot. Moreover, it takes keep the altitude and speed
a long time for the aircraft to achieve new balance in stable flight. To stable when the system x
ensure that the morphing aircraft can maintain stable flight during Ahtx Bhtu is
wing shape varying, the wing morphing process needs a flight control disturbed by w(h(t)).
system to obtain a prospective objective. In this article, the control
According to the property
objective is to keep the altitude and speed constant during the of the LPV system, the inner
transition from extended-wing configuration to folded-wing loop of the control system is a
linear
quadratic
output
configuration.
tracking
performance
x Ahtx Bhtu B1wht
16 characteristics for the linear
system x Ahtx B
4
2
41
44
42
4
where x e R , u e R , w e R , A e R , B e R , B1 = I . Considering
htu. The outer loop is
w(h(t)) as a disturbance, then the system can be
composed of a gain selfregarded as an LPV system x Ahtx Bhtu with a
scheduled robust H1 controller
The longitudinal LPV model of the morphing aircraft in feedback con-troller which
provides stability, as well as
morphing can be rewritten as
in Fig. 4.
0
Gain self-scheduled H1 control for morphing aircraft in the wing transition process based on an LPV model
2 0
6 0
6
0
0 3
0 7
6 0
6
0
0
2
0
Ax
>
Bu
Cx
Fr
Gr
< z Hx
>
>z
>
<
>
&
18
DH Du
d T &T
u de
6
6
a3 1
6
6
6
6
1 0
u0
0
0
0
0
13
14
32
33 34
a
a
a
22
23
42
a41
6
6
12
24
43
44
11
12
31
32
b
b
21
22
41
42
0
20
6
6
6
0 07
7
0 0
7
0 07
0 0 0 77
0
1
7
0 07
0
7
7
1 0 0
0 0
0 1
0h
eH
0e
0 0
0 03
0
0 0 0 1
Kq K
G
0 0 0 0 0 0 0 0 0
According to Fig. 5, the control input
u KLQy
a21
0 0 0
1 0 0 0 0 0 0 0 0
0 0 1 0 0 0 T
&
>
0 0 0
>
F 0 0 0 0 1 0
0 0 0 0 0 0
>
where
Du Dw Dq Dh DH de
8 y Dq Dh eH eH
eu eu & T
17
0 1 0 0 0 0 0 0
0 0 1 0 0 0 0 0 3
0 0 0 1 0 0 0 0 7
7
0 0
0
0
0
0 1 0
6
6
6
7
7
7
7
7
7
7
0 5
0
2 0
C
6 0
6 0
20
4 0
Fig. 4 The multi-loop control structure.
0 7
6
6
6
6
6
6
6
913
19
T
T
T
J 2 Z0 e~ e~ u~ Ru~dt 2 e Ve
with R > 0, V P 0. The
optimization problem can be
solved
numerically
using
an
iterative procedure through
Lyapunov equations.22,23
As the state-space
matrices of the LPV model
change with
16 5
0
68: 7 0 14 0 03
0
:0
:0
15:8
18:9
27
914
T. Yue et al.
following formulas for the state-space data of the LPV controller are obtained:
8 z Hx
y Cx
where
>
32
CKi DKi
X
It is easy to check that ai are convex coordinates satisfying
h 2 H : f
ah :a P
0;
in
i1 i
performance
22
30
0:0344 &
&
&
So h(t) ranges in the polytope H e Co{hi, i = 1, 2, 3, 4}. The
vertices hi are the values of h(t) at the four vertices of the
h1
22min
Ki
eter variations.
a t
i 1 i
u KLPVhy
29
to make the altitude and speed remain constant during the
wing
morphing
process.
The reference
input
T
r rDH rDu & 0, and the outer-loop LPV control structure
is shown in Fig. 6.
When using convex hull algorithm to design the LPV controller, if there are n variable elements in Ac(h(t)) and B(h(t)),
n
the convex hull will have 2 vertices. The calculation will be
complicated if n is big. Therefore, we assume that the elements
in Ac(h(t)) and B(h(t)) do not change as they vary in very small
range. In other words, we only consider the elements that vary
in large ranges during the wing folding process.
From the simplified LPV equations of morphing aircraft we
find that the variable elements a22 and a32 in Ac(h(t)) vary in
large ranges, which both depend on the aerodynamic changes
during the wing folding process. Hereafter, the parameter vec-
parameter
Ki
a32min
:
DK h t
i1 i i
a32
CKht
28
>
<
BK h t
AK h t
box:
22min
a32min
22max
a32min
h3
22max
; h4 a32max
.It is clear that the state-space maa32max
trix Ac(h(t)) ranges in a polytope of matrices whose vertices
are the images of the vertices h1, h2, h3, h4. In other words,
Acht 2 CofAi : Ahi; i 1; 2; 3; 4g
31
Gain self-scheduled H1 control for morphing aircraft in the wing transition process based on an LPV model
wing folding process.
shape transition phase. The response of the aircraft in closedloop simulation is shown in Fig. 7. It is clear that the speed is
almost constant and the altitude decreases about 0.7 m to the
maximum during the wing folding process. They can both converge in 10 s after the wing finishes folding. Since the wing
area decreases after it folds, the angle of attack will increase
to achieve a new balance. In addition, the changes in elevator
deflection and throttle are both within acceptable ranges. It is
obvious that the gain self-scheduled H1 robust controller
based on the morphing aircrafts LPV model can eliminate the
disturbance caused by wing folding and guarantee con-stantspeed and altitude flight during the wing transition process.
915
p
0:8 0:5180 a CL
8
p
> CD0 1:2 0:5 180 a CD
>
>
< Cm0 0:8 0:5 p aCm
CL
>
>
180
>
>
>
:
aerodynamics.
5. Conclusion
(1)The
916
T. Yue et al.
Sz
a31 mIy
>
Sz2
>
>
>
>
>
>
qV 0SCDV
Da 0
Sz
mIy
>
>
>
>
>
qu0Sc Cm0
>
ma
qV0cSCmV
>
>
>
>
>
C a
qu0S CD0
C C a
L0 ma 0
La
mde e0
>
>
>
>
z
>
>
a32
>
>
>
>
>
>
mI
>
>
qV0SCDa
<
> 2
D0
qw0S
C a
Da 0 2
mIy Sz
>
>
Acknowledgements
C ma
>
>
>
>
>
qV0cSCma
qw0Sc Cm0
L0
ma 0
mde e0
La
>
>
>
Sz
>
>
>
mI
>
>
mw0
a3 3
>
2Sz
Iy
mI
z2
Sz w0
2
qV0c SCmq
>
>
>
>
>
Sz
>
2 mg cos h0
34
gS cos h
>
>
mIy S
>
>
mIy
>
>
Appendix A
>
>
A4
:
a
11
8
>
12
a21
>
>
>
>
>
>
>
>
>
>
>
>
>
>
23
42
41
11
b 11
21
24
Sz
2
mIy
43
44 7
12
qV cSCm
0
de
11
>
>
>
>
Lde
m2
Sz
>
22 3
qV SC
b21
>
>
>
>
A5
de
mIy
2
1 qV SCD
Iy
2
5>
43
a 34
33
42
1
;
6a
6
24
0;
41
14
>
>
22
a 32 a
31
>
13
0
;
>
A1
>B
>
>
b31
>
31
32
>
>
>
>
de
SCD
S 2
mIy
>
>
>
cSCm
de
Sz2
mIy
7>
b
7
7
41
42
>b
>
>
5>
Iy
<
w
11
>
dp
12
>
>
>
mIy
>
>
>
<
>
6
6
>
>
>W
>
>
Sz
>
21
31
>b
>
22
>
>
>
>
>
>
>
>
>
>
S
z
>
>
>
41
>
>
>
>
>
>
b32
dp
>
>
>
>
:
mI
>
Iy
>
>
b42
>
11
8
>
mIy
A6
Sz
qV0SCDV
>
C a
qu0S CD0
Iy
Da 0
>
11
mI
>
>
>
>
>
>
qV0cSCmV
>
>
>
S
z
mIy
qV
S CD0
T0
Da 0
mg sin h0
ma
>
qu0Sc Cm0
L0
ma 0
>
Sz
mde e0
>
CL
>
>
a
>
>
>
>
mI
>
>
Iy
S z
=m1
1x
2S S
1z
2x
2
qV cS
=m2
2S S
2z
>
>
>
>
>
>
Sz
>
>
>
>
>
a12
>
>
>
>
mIy
>
>
>
>
>
>
>
>
>
>
>
>
Sz >
qV0SCDa
>
>
>
>
Sz
qw0S CD0
Da 0
2
mIy
>
<
L0
ma 0
>
S CL0
CL
>
w Sc C
ma
m0
CL
mg cos h0
a0
>
cSC
q V0
gSz sin h0
e0
2
qV
w21
>
>
Sz
>
>
>
C a
mde
>
>
>
ma
CLa
>
>
>
m0
>
C a
>
L0
>
C mde e0
ma 0
<
Sz
>
w
SC
D0
T0
3
1
Da 0
mIy
Sz
mg sin
>
Iy
>
>
>
>
>
>
mw0
a 13
>
2Sz
>
>
>
mI
>
>
>
1>
>
>
>
>
>
>
mIy
>
w
z
2
V
>
2x
2z
S
z
mIy
c SC
>
mq
>
>
qV cS
0
ma
m0
>
L0
>
>
C La
a
>
>
>
gS
e0
>
2S S
1z
>
1x
=m2
2S S
>
>
>
>
=m1
>
sin h0
>
>
>
>
>
>
mg cos
a14
2
>
gS cos
0
2
>
>
>
>
>
S z mIy
mIy
>
Sz>
>
>
>
A7
>
>
>
>
>
A2
w41
1 1
>
>
>
>
<
1927; 1998.
u0
23
g sin
>a
>
>
>
>
>
A3
m 2
1. Wlezien RW, Horner GC, McGowan AR, Padula SL, Scott MA,
C Daa0
w S C D0
References
q V SC La
22
>
a21
>a
>
24
Am 2004;42(2):289.
h0
>
:
Gain self-scheduled H1 control for morphing aircraft in the wing transition process based on an LPV model
2008;31(6):168799.
3. Bowman
7.
8.
Seigler TM, Neal DA, Bae JS, Inman DJ. Modeling and flight control
of
9.
10.
Valasek J, Tandale MD, Rong J. A reinforcement learningadaptive control architecture for morphing. AIAA-2004-6220; 2004.
14.
Natesan K, Gu D,
Postlethwaite I. Design of
static H1 linear parameter
varying
controllers
for
unmanned
aircraft.
J
Guidance
Control
2007;30(6):18227.
Dyn
11. Baldelli DH, Lee D, Sanchez Pena RS. Modeling and control of 15.
an
Apkarian P, Gahinet
P, Becker G. Self-scheduled
H1
control of
linear
917
16.
17.
18.
19.
23.
Shin JY, Balas GJ, Kaya MA. Blending methodology of linear Yue Ting is a postdoctor at
School of Aeronautic Science
parameter varying control synthesis of F-16 aircraft system. J
and
Engi-neering,
Beihang
Guidance Control Dyn 2002;25(6):10408.
University. He received the B.S.,
21. Stevens BL, Lewis FL. Aircraft control and simulation. 2nd ed.M.S. and Ph.D. degrees in
New York: John Wiley & Sons, Inc.; 2003.
20.