Sie sind auf Seite 1von 6

2013 10th IEEE International Conference on Control and Automation (ICCA)

Hangzhou, China, June 12-14, 2013

Reference Tracking Control of Hypersonic Vehicles Using Switched


Linear Parameter-Varying Approach
Qiugang Lu1 , Lixian Zhang2 , Peng Shi3 , and Hamid Reza Karimi1
with dramatic parameter variations and large flight range,
a class of LPV modeling and control methodologies for
the hypersonic vehicles have been developed. For example,
[6] proposes a LPV framework, in which the structural
dynamics are represented with dependency on operating
parameters as a set of state-space matrices that are affine
functions of those parameters. For the over-actuated issue,
in [7], the controller design is given for the longitudinal
LPV model of the hypersonic vehicle to enforce an ideal
behavior in the presence of redundant actuation. Moreover,
with respect to large parameter variation and flight envelope,
some literature using switching-relevant techniques have also
been reported, refer to [3]. Recently, the switched LPV
approach has also been proposed with applications to such
as F-16 aircraft system [8], active magnetic bearing system
[9], and missile autopilot system [10]. However, surprisingly,
in the hypersonic vehicle regime, the utilization of switched
LPV technique still remains open, which inspires the study
in this paper. As far as the switching logic is concerned, the
typical average dwell time (ADT) switching logic has been
widely investigated [11] but revealed conservative, a more
flexible switching strategy called mode-dependent average
dwell time (MDADT) has been created in [12].
This paper mainly addresses the switched LPV system
modeling and controllers design for the hypersonic vehicle
under the ADT and MDADT switching logics. In Section II,
the detailed procedures to convert the nonlinear system of
the hypersonic vehicle into switched LPV system are thoroughly demonstrated. In section III, in the former part, some
necessary preparations for further derivation are illustrated.
Afterwards, the stability criterion and controllers design
approaches under the frameworks of ADT and MDADT are
explored respectively, which are expressed in the form of
linear matrix inequalities. The nonlinear simulations for the
switched LPV model of the hypersonic vehicle are conducted
under the ADT and MDADT switching logics in Section IV.
The conclusions of this paper are given in Section V.

Abstract This paper is concerned with the control problem


of the air-breathing hypersonic vehicle using switched linear
parameter-varying (LPV) systems approach. By linearizing the
nonlinear model of the hypersonic vehicle at trim points and
dividing the velocity scope into several partitions, a switched
LPV model for the underlying systems is derived. With aid
of multiple Lyapunov-like functions technique, the controllers
ensuring the tracking of given commands of the closed-loop
system are designed under the mode-independent and modedependent switching logics, respectively. Simulation results for
the hypersonic vehicle show the effectiveness of the proposed
method and the advantages of the mode-dependent switching
logic.
Index Terms Average dwell time; Hypersonic vehicle;
Mode-dependent average dwell time

I. I NTRODUCTION
The air-breathing hypersonic vehicles have received extensive attention in the world during the past decades.
However, the control of hypersonic vehicles is considerably
challenging due to the strong parameter coupling, large
modeling uncertainty and the slender geometries required
for the aircraft [1]. To overcome such impediments, the
longitudinal dynamic equations including rigid and flexible
models are widely utilized to address the control problems of
the hypersonic vehicles [2]. In practice, for convenience, the
rigid longitudinal models are more frequently employed to
verify the control methodologies although they are relatively
simplified.
The characterized dramatic parameter variations and wide
velocity scope of the hypersonic vehicles render their stability and dynamic performance different from other aircrafts.
With such properties, the hypersonic vehicle aircrafts are
sensitive to the flight condition changes and it is difficult
to measure and estimate the aerodynamic characteristics. To
solve such problems, numerous effective control strategies
have emerged in the recent literature, refer to [3] and [4] for
the flexible and rigid model of the hypersonic vehicles.
Besides, the LPV system has drawn considerable attention
in the past decade [5]. Because of its merits in dealing

II. S WITCHED LPV M ODEL FOR H YPERSONIC V EHICLES

1 Qiugang

Lu and Hamid Reza Karimi are with the Department of Engineering, University of Agder, N-4898 Grimstad, Norway

The switched LPV model for the rigid longitudinal dynamics of the hypersonic vehicle is presented in this section.
Firstly, we need choose the scheduling parameters and divide the scope of the scheduling variables into partitions.
Secondly, in each partition zone, we need to select and
calculate the set point so as to achieve the LPV local
model of the hypersonic vehicle in this partition. Thirdly,
via Jacobian linearization method and tensor-product (TP)
model transformation approach, we can achieve the local

qiugang.lu@uia.no, hamid.r.karimi@uia.no

2 Lixian Zhang is with the Space Control and Inertial Technology Research Center, Harbin Institute of Technology, Harbin 150001, Heilongjiang
Province, P.R.China lixianzhang@hit.edu.cn
3 Peng Shi is with the Department of Computing and Mathematical
Sciences, University of Glamorgan, Pontypridd, CF37 1DL, U.K., with
the School of Engineering and Science, Victoria University, Melbourne,
Vic. 8001, Australia, and also with the School of Mathematics and
Statistics, University of South Australia, Adelaide, S.A. 5001, Australia

pshi@glam.ac.uk

978-1-4673-4708-2/13/$31.00 2013 IEEE

670

polytopic LPV model in each partition. Then by combining


all the local LPV models using the switching signal, which
is determined by the scheduling parameters, we can obtain
the overall switched LPV model of the hypersonic vehicle.
The hypersonic vehicle model considered in this paper is
developed by [13] and [14]. TH , DH , LH , GH are trust,
drag, left, graviton force, respectively. The rigid nonlinear
longitudinal model of a generic hypersonic vehicle can be
described as follows [4], [15],

model of the hypersonic vehicle. The overall switched LPV


model of the hypersonic vehicle is described as (3),

x(t)

= A(t) (VH , h)x(t) + B(t) (VH , h)u(t)


(3)
y(t) = C(t) (VH , h)x(t) + D(t) (VH , h)u(t)
where x = [ VH h H q ]T is the state vector, y =
[ VH h ]T is the output vector, u = [ e ]T denotes
the control input of the system.
In this paper, we consider the tracking control of hypersonic vehicle. The reference signal is chosen as r(t) =
[ Vr hr ]T , and the actual output is y(t) = [ Vt ht ]T .
The tracking problem can be addressed by state-feedback
controllers such that the output of the closed-loop system
could track the given reference signal,

TH cos H DH
sin
f1 = V H =
E
m
(RE + h)2
LH + TH sin H
E cos
VH cos
f2 = =

+
2
mVH
VH (RE + h)
RE + h
f3 = h = VH sin
(1)

lim [y(t) r(t)] = 0

f4 = H = q
Myy
f5 = q =
Iyy

Defining the following error integral action:


Z t
xe =
[y( ) r( )] d
0

where VH , , h, H , q are state variables, representing


velocity, flight path angle, altitude, angle of attack and
pitch rate, respectively. E , m, RE , Iyy are, respectively,
gravitational constant, mass, radius of the earth and y-axis
inertia. The coefficients can be expressed as

we can obtain the augmented system,

(t) (VH , h)u(t)


x a (t) = A(t) (VH , h)xa (t) + B
+B1 r(t)

y(t) = C(t) (VH , h)xa (t)

(4)

where xa (t)
= [ x(t) xe ]T , A(t) (VH , h) =

A(t) (VH , h) 0
(t) (VH , h) = B(t) (VH , h) ,
,B
C(t) (VH , h) 0
0

C(t) (VH , h) = [ C(t) (VH , h) 0 ], B1 = [ 0 I ]T .


Based on TP transformation [17], the LPV augmented
model (4) of the hypersonic vehicle can be rewritten as

x a (t)
xa (t)
B1
= SN
A
(p
(t))
+
r(t)
n
(t)n
n=1
y(t)
u(t)
0
(5)
where row vector A(t)n RIn (n = 1, 2, ..., N , N
is the dimension of the parameter vector p(t)) contains
one bounded variable and continuous weighting function
(t),n,in (pn ) (in = 1, 2, ..., In ), In is the number of the
weight function used in the ndimension of the parameter
vector p(t). S RI1 I2 IN OI is constructed from the
linear time-invariant (LTI) vertex systems Si1 i2 iN ROI
and p(t) P = [VH min ,VH max ] [hmin , hmax ].
From (5), the augmented switched LPV system (4) can be
transformed into the following polytopic form:

x a (t) =
(t),i (t) A(t),i xa (t) + B(t),i u(t)

i=1
+ B1 r(t)
(6)

y(t) = P

(t)C
x (t)

LH = 0.50 VH2 SH CL
DH = 0.50 VH2 SH CD
TH = 0.50 VH2 SH CT
Myy = 0.50 VH2 SH cCM
where LH , DH , TH , Myy are lift, drag, thrust and pitching
moment, 0 is the density of air. The aerodynamic coefficients and thrust coefficients which are the functions of Ma
(Ma = VH /a, a is the speed of sound), angle of attack (H ),
fuel rate () and pitch control surface ( e ) are provided in
[4] and [16].
We choose the velocity (VH ) and altitude (h) as the
scheduling variables, since velocity and altitude are the
main concerns in the flight of the hypersonic vehicle, p =
[ VH h ]T . Then we divide the velocity scope into three
partitions [7500, 8500], [8500, 9000], [9000, 9500]. In each
partition, the LPV model can be written as the following
form by Jacobian linearization method:
x = A(VH , h)x + B(VH , h)u
where x = [ VH h H q ]T , u =
[ e ]T . The specific expression of elements of matrices A(VH , h), B(VH , h) and C(VH , h) are omitted here.
The LPV model in the (t)-th partition zone is formulated
as
x = A(t) (p(t))x + B(t) (p(t))u
(2)

where

Ps

i=1

i=1

(t),i

(t),i a

(t),i (t) = 1, s =

Q
n

In = 2N .

III. C ONTROLLER D ESIGN UNDER ADT AND MDADT


S WITCHING L OGICS
In this section, the switched controllers design methodologies are given for the general switched LPV systems under
the frameworks of ADT and MDADT respectively, which
are expressed in the form of linear matrix inequalities.

(t) I, is the switching signal governing the switchings of


the system between subsystems and I is the set comprised
of all subsystems. (t) is determined by the parameter
vector p(t). In this paper, the velocity VH is the parameter
determining the switching sequence of the switched LPV
671

TABLE I
T RIM CONDITION

1
2
3

h
90000 ft
90000 ft
90000 ft

VH
7500 ft/s
8700 ft/s
9500 ft/s

0
0
0

H
3.6262
2.7424
2.3044

q
0
0
0

0.2919
0.2169
0.1947

Lemma 2: Consider the nonlinear switched system (9), let


0 < i < 1, i 1 and > 0 be given constants. Suppose
there exist positive definite C1 functions V(t) : Rn R,
(t) I, with zero initial condition V(t0 ) = 0 such that
(i, j) I I, i 6= j, Vi (x(t)) i Vj (x(t)), i I.
Denoting (t) = y T y 2 wT w
V i (x(t)) + (t) i Vi (x(t))
(16)

e
5.1795
2.3033
1.2777

the system is GUAS for any switching signal satisfying


MDADT
ln i
ai ai =
(17)
i
with weighted L2 -gain
Z
Z
max ( t0 )
T
e
y( ) y( )d
2 w( )T w( )d

A. Preliminaries
Definition 1: [18] For a switching signal (t), (t) I,
and each t2 t1 0, let N (t2 , t1 ) denote the number of
discontinuities of (t) in the open interval (t1 , t2 ). We say
that (t) has an average dwell time a if there exist two
positive numbers N0 (we call N0 the chattering bound here)
and a such that

where =

N (t2 , t1 ) N0 + (t2 t1 )/ a , t2 t1 0
(7)
Definition 2: [12] For a switching signal (t), (t) I,
and any T t 0, let N(t)i (t, T ) be the switching
numbers of the ith subsystem activated during the interval
[t, T ], Ti (t, T ) is the total running time of the ith subsystem
in [t, T ], and i I. We say that (t) has a modedependent average dwell time ai , if there exist positive
mode-dependent chattering bounds N0i and ai such that
the following inequality is satisfied
N(t)i (t, T ) N0i + Ti (t, T )/ ai , T t 0
Definition 3: For > 0, > 0, system

x = f(t) (x(t), w(t))


y = h(t) (x(t), w(t))

(t) I

the

continuous-time

x = f(t) (x(t)), (t) I

max

(18)
= max{i }, min =
iI

Proof. For the stability, it is assumed that w(t) = 0,


from Lemma 1, we can easily prove that if (17) holds,
the GUAS of (9) can be guaranteed. Now we turn to the
case w(t) 6= 0, integrating both sides of (16), and from
Vi (x(t)) i Vj (x(t)), we can get
V (t)

V (ti )ei (tti )

(9)

i ei (tti ) V (ti1 )ei (ti ti1 ) i ei (tti )


Z ti
Z t
ei1 (ti ) ( )d
ei (t ) ( )d

ei (t ) ( )d

ti

ti1

i 1 ei (tti ) ei1 (ti ti1 ) e0 (t1 t0 ) V (t0 )


Z t1
i i1 1
ei (tti ) e0 (t1 ) ( )d
t
Z 0t2
i i1 2
ei (tti ) e1 (t2 ) ( )d
t1
Z t

ei (t ) ( )d
ti

(11)

V i (x(t)) i Vi (x(t))

(13)

ti

nonlinear

(12)

Since V (t0 ) = 0, the inequality above can be rewritten as


Z t1
i i1 1
ei (tti ) e0 (t1 ) y yd
t0
Z t2
ei (tti ) e1 (t2 ) y yd
+ i i1 2
t1
Z t
i (tti )
e
y yd
+ +
ti
Z t1
ei (tti ) e0 (t1 ) 2 wT wd
i i1 1
t0
Z t2
ei (tti ) e1 (t2 ) 2 wT wd
+ i i1 2
t1
Z t
+ +
ei (tti ) 2 wT wd

and ((t) = i, (t ) = j) I I, i 6= j,
(14)

the system is GUAS for any switching signal with MDADT


ln i
i

N0i
,
iI i

(8)

1i (kx(t)k) Vi (x(t)) 2i (kx(t)k)

ai ai =

max
min

iI

let i > 0, i 1, i I be given constants. Suppose


there exist positive definite C1 functions V(t) : Rn R,
(t) I, and two class K functions 1i , 2i , i I such
that i I

Vi (x(t)) i Vj (x(t))

min{i }.

is said to be globally uniformly asymptotically stable


(GUAS) with weighted L2 -gain no greater than if under
zero initial condition, system (9) is GUAS and the following
inequality holds for all nonzero w(t) L2 [0, +)
Z
Z
s
T
e y(s) y(s)ds
2 w(s)T w(s)ds
(10)
Lemma 1: Consider
switched system

(15)

ti

672

It is easy to get
Z t
t0
t

B. Controllers Design
emax (t )

emin (t )

t
Z 0t

iI

t0

emax (t )

Consider the following switched LPV system:

x(t)

= A(t) ((t))x(t) + B(t) ((t))u(t)

+ E(t) ((t))w(t)
z(t) = C(t) ((t))x(t) + D(t) ((t))u(t)

+ F(t) ((t))w(t)

Ni ( ,t) T

iI

y yd

Ni ( ,t) 2

Y
iI

wT wd

Ni (t0 , ) T

y yd

where x(t) Rn is the state, w(t) is the disturbance input,


z(t) Rnz is the controlled output, u(t) Rnu is the
control input. The system matrices are parameter-dependent
matrices of compatible dimensions of time-varying parameter (t) = [ 1 (t) 2 (t) N (t) ]T RN . To
proceed, we make the following assumptions:
1) The state-space matrices are continuous and bounded
functions and depend affinely on time-varying parameter
(t).
2) The parameter (t) is measurable online and varies in
a polytopic :

emin (t ) 2 wT wd

t0

From (8), we have


Ti (t0 , )i
ln i
Hence, the following inequality can be obtained
Z t
emax (t ) emax ( t0 ) y T yd
t0
Z t
Y
0i

N
emin (t ) 2 wT wd
i
Ni (t0 , ) N0i

iI

(t)

t0

Integrating both sides from t = t0 to , we can obtain


Z Z t
emax (t ) emax ( t0 ) y( )T y( )d dt
t=t0 =t0
Z Z t
Y
0i

N
emin (t ) 2 w( )T w( )d dt
i
iI

t=t0

i=1

The objective here is to design a set of parameter-dependent


switching controllers

t=

which is equivalent to
Z
emax ( t0 ) y( )T y( )d
t0
Z
max Y
N0i

2 w( )T w( )d
iI i
min
t0

K(t) ((t)) =

where =

(21)

Theorem 1: (MDADT case) Consider the switched LPV


system (22) and let 0 < i < 1, i I, > 0 and i > 1, i
I, be given constants. It is assumed that all the system matrices belong to the convex hull, (Aiz , Biz , Ciz , Diz , Eiz , Fiz )
are the vertices of the ith subsystem, z z := {1, 2, ..., s}.
If there exist matrices Si > 0 and Uiz , i I, z z, such
that (i, j) I I, Si i Sj , the following linear matrix
inequalities are satisfied

(Aiz Si + Biz Uiz )T + (Aiz Si + Biz Uiz ) + i Si

(Ciz Si + Diz Uiz )T


Eiz
I
Fiz < 0
(23)

2I

Letting t0 = 0, the proof can thus be finished.


Remark 1: Lemma 2 presents the weighted L2 -gain of
a general nonlinear switched system under the MDADT
switching signals. From the derivation process, it is straightforward that the weighted L2 -gain under the ADT switching
logic can be obtained by letting max = min = 0 .
Therefore, the weighted L2 -gain for the general nonlinear
switched system can be written in the following form if
corresponding conditions similar to Lemma 2 are satisfied,
Z
Z
e0 ( t0 ) y( )T y( )d
2 w( )T w( )d
p

(t),i (t)K(t),i

such that the following switched LPV closed-loop system is


exponentially stable with specified L2 -gain ,

x(t)

= A(t) ((t))x(t) + E(t) ((t))w(t)


(22)
z(t) = C(t) ((t))x(t) + F(t) ((t))w(t)

t0

s
X
i=1

The inequality above can be expressed in the form of (10),


i.e.,
Z
Z
max ( t0 )
T
e
y( ) y( )d
2 w( )T w( )d
t0

:= Co{ (t),1 , (t),2 , , (t),s }


( s
)
X
N
=
(t),i (t) = 1, s = 2

In this sense, the switched LPV system (20) can be


expressed as:

A(t) ((t)) B(t) ((t)) E(t) ((t))


C(t) ((t)) D(t) ((t)) F(t) ((t))

s
X
A(t),i B(t),i E(t),i
=
(t),i (t)
C(t),i D(t),i F(t),i

=t0

=t0

i=1

Rearranging the integral area, we have


Z
Z
emax ( t0 ) y( )T y( )d
emax (t ) dt
=t0
t=
Z
Z
Y
N0i
2
T

i
w( ) w( )d
emin (t ) dt
iI

(20)

(19)

N0 .
673

for all z z, there exist a set of switched controllers that


can guarantee the switched LPV system to be GUAS with a
weighted L2 -gain in the form of (18) under the MDADT
switching signal

9200
Cmd
ADT
MDADT

9100
9000

ai ai

Velocity(ft/s)

8900

ln i
=
i

(24)

the controllers can be obtained by

8700
8600

Kiz = Uiz Si1


(25)
Theorem 2: (ADT case) Consider the switched LPV system (22) and let 0 < 0 < 1, > 0 and > 1 be
given constants. It is assumed that all the system matrices
belong to the convex hull, (Aiz , Biz , Ciz , Diz , Eiz , Fiz ) are
the vertices of the ith subsystem, z z. If there exist
matrices Si > 0 and Uiz , i I, z z, such that
(i, j) I I, Si Sj , the following linear matrix
inequalities are satisfied

(Aiz Si + Biz Uiz )T + (Aiz Si + Biz Uiz ) + Si

(Ciz Si + Diz Uiz )T


Eiz
I
Fiz < 0
(26)

2I

8500
8400
8300

Fig. 1.
logics

10

15

20
Time(s)

25

30

35

40

The responses of the system under ADT and MDADT switching

9.5

x 10

Cmd
ADT
MDADT

9.4
9.3

Altitude(ft)

9.2
9.1
9
8.9
8.8

for all z z, there exist a set of switched controllers that


can guarantee the switched LPV system to be GUAS with
a weighted L2 -gain in the form of (19) under the ADT
switching signal
ln
a a =
(27)
0

8.7
8.6
8.5

10

15

20
Time(s)

25

30

35

40

Fig. 2. The altitude response of the system under ADT and MDADT
switching logics

the controllers are given by


Kiz = Uiz Si1

8800

(28)

chosen as square waves which cross the three subsystems


back and forth with its ADT or MDADT satisfying the
requirements of Theorem 1 and Theorem 2 (note that this
command may seem impractical, but it is difficult enough
to test our approach, since actually, the switching behavior is
much more smooth than this constructed command signal).
The command of altitude is chosen as shown in Fig. 2.
Fig. 1 shows the nonlinear responses of the hypersonic
vehicle aircraft under the command input. The solid line
is the command input. The dash line stands for the output
trajectory under the ADT switching logic, while the dot
line shows the velocity response under MDADT switching
logic. Both the two tracking performances are acceptable
over the entire time range. However, there is still difference
between the response curves. From Fig. 1, it can be observed
that the velocity response of MDADT switching logic is
faster than that of the ADT switching logic, which indicates
that the MDADT switching logic outperforms the ADT in
achieving the transient performance. Fig. 2 is the altitude
tracking responses of the two switching logics. The response
under MDADT framework is better than that under ADT
framework. It should be pointed out that both the tracking
responses are acceptable. Fig. 3 and Fig. 4 display the fuel
rate and the actuator inputs respectively. Note that the control

IV. S IMULATION R ESULTS


Based on the established switched LPV model of the rigid
longitudinal dynamics of the hypersonic vehicle in Section
II and the theorems proposed in Section III, the following
simulations are conducted to validate the effectiveness of the
former results. In this section, the flight envelop is chosen
as VH [8000, 9500] and altitude h [85000, 95000], and
the velocity scope is divided into three partitions. The trim
conditions are illustrated in Table 1. For the ADT framework,
0 = 0.1, = 1.01, and for the MDADT framework, =
[ 0.1 0.11 0.12 ], = [ 1.01 1.01 1.01 ]. By solving
the LMIs (23)-(25) and (26)-(28), respectively, we can obtain
the controller gains. Via Theorem 1, for the MDADT case,
we have a1 = 2.7763 s, a2 = 2.5239 s, a3 = 2.3136 s.
From Theorem 2, we can get a = 2.7763 s.
Next, we mainly focus on the simulations of switched LPV
model of the hypersonic vehicle under ADT and MDADT
switching logics. To test the tracking performance of the
closed-loop nonlinear system of the hypersonic vehicle, a
velocity command input is constructed as the solid line in
Fig. 1. To fulfill the command tracking and illustrate the
effectiveness of the controllers, we employ similar techniques
as in [8]. In particular, the trajectory of the command is
674

193821/S60 from the Research Council of Norway (RCN).


NORCOWE is a consortium with partners from industry and
science, hosted by Christian Michelsen Research.

1
ADT
MDADT

fuel rate

0.8

0.6

R EFERENCES

0.4

[1] B. Fidan, M. Mirmirani, and P. Ioannou. Flight dynamics and control


of air-breathing hypersonic vehicles: review and new directions, AIAA
Guidance, Navigation and Control Conference, Norfolk, VA, AIAA
paper 20037081, 2003.
[2] D. Schmidt. Optimum mission performance and multivariable flight
guidance for airbreathing launch vehicles, Journal of Guidance, Control, and Dynamics, vol. 20, no. 6, pp. 11571164, 1997.
[3] H. Hughes. LPV H control for the longitudinal dynamics of a
flexible air-breathing hypersonic vehicle. Ph.D. dissertation, North
Carolina State University, 2011.
[4] C. Marrison and R. Stengel. Design of robust control systems for a
hypersonic aircraft, Journal of Guidance, Control, and Dynamics, vol.
21, no. 1, pp. 5863, 1998.
[5] L. Wu, P. Shi, H. Gao, and C. Wang. Delay-dependent robust H and
L2 -L filtering for LPV dystems with both discrete and distributed
delays, IEE. ProceedingControl Theory and Application, vol. 153,
no. 4, pp. 483492, 2006.
[6] R. Lind. Linear parameter-varying modeling and control of structural
dynamics with aerothermoelastic effects, Journal of Guidance, Control, and Dynamics, vol. 25, no. 4, pp. 733739, 2002.
[7] D. Sigthorsson, A. Serrani, M. Bolender, and D. Doman. LPV control
design for over-actuated hypersonic vehicles models, AIAA Guidance,
Navigation, and Control Conference, Chicago, Illinois, AIAA Paper
20096280, 2009.
[8] B. Lu, F. Wu, and S. Kim. Switching LPV control of an F-16 aircraft
via controller state reset, IEEE Transactions on Control Systems
Technology, vol. 14, no. 2, pp. 267277, 2006.
[9] B. Lu and F. Wu. Switching LPV control designs using multiple
parameter-dependent Lyapunov functions, Automatica, vol. 40, no. 11,
pp. 19731980, 2004.
[10] S. Lim and J. How. Modeling and H control for switched linear
parameter-varying missile autopilot, IEEE Transactions on Control
Systems Technology, vol. 11, no. 6, pp. 830838, 2003.
[11] L. Zhang, E. Boukas, P. Shi, and Z. Chen. A -dependent approach to
H control of uncertain switched linear systems with average dwell
time, Optimal Control Applications and Methods, vol. 32, no. 1, pp.
1527, 2011.
[12] X. Zhao, L. Zhang, P. Shi, and M. Liu. Stability and stabilization
of switched linear systems with mode-dependent average dwell time,
IEEE Transactions on Automatic Control, vol. 57, no. 7, pp. 1809
1815, 2012.
[13] M. Bolender and D. Doman. A non-linear model for the longitudinal
dynamics of a hypersonic air-breathing vehicle, AIAA Guidance,
Navigation and Control Conference, San Fransisco, CA, AIAA paper
20056255, 2005.
[14] M. Bolender and D. Doman. A nonlinear longitudinal dynamical
model of an air-breathing hypersonic vehicle, Journal of Spacecraft
and Rockets, vol. 44, no. 2, pp. 374387, 2007.
[15] L. Fiorentini, A. Serrani, M. Bolender, and D. Doman. Robust
nonlinear sequential loop closure control design for an air-breathing
hypersonic vehicle model, Proceedings of the 2008 American Control
Conference, Seattle, WA, pp. 34583463, 2008.
[16] J. Parker, A. Serranit, S. Yurkovich, M. Bolender, and D. Doman.
Control-oriented modeling of an air-breathing hypersonic vehicle,
Journal of Guidance, Control, and Dynamics, vol. 30, no. 3, pp. 856
869, 2007.
[17] Petres Z. Polytopic decomposition of linear parameter varying models
by tensor-product model transformation, Ph.D. Thesis Booklet. Budapest: Budapest University of Technology and Economics, 2006.
[18] J. Hespanha and A. Morse. Stability of switched systems with average
dwell-time, Proceedings of the 38th IEEE Conference on Decision and
Control, Phoenix, AZ, pp. 26552660, 1999.

0.2

0.2

0.4

10

15

20
Time(s)

25

30

35

40

The fuel rate input under ADT and MDADT switching logics

pitch control surface(deg)

pitch control surface(deg)

Fig. 3.

15
ADT

10
5
0
5
10
15

10

15

20
Time(s)

25

30

20

35

40

MDADT

10
0
10
20
0

10

15

20
Time(s)

25

30

35

40

Fig. 4. The pitch control surface deflection under ADT and MDADT
switching logics

inputfuel ratio and control surface deflection have large


peaks when switchings happen, since the command input
has a steep change at switching time point. It is easy to find
that the larger change of command velocity input implies the
larger control input.
V. C ONCLUSION
This paper explored the controller design problem for
the switched LPV model of the hypersonic vehicle under
the frameworks of ADT and MDADT switching logics. By
Jacobian linearization and TP transformation, the switched
LPV model was obtained to cope with this problem. With
regard to this switched LPV system, the controllers design
approaches under both ADT and MDADT switching logics
were given through linear matrix inequalities. By comparison, the MDADT logic had an advantage of more flexibility
in manipulating than the ADT switching logic. Finally, the
nonlinear simulation of hypersonic vehicle system was given
to verify the validity and effectiveness of the proposed ADT
and MDADT controllers design method, respectively, and the
merits of the MDADT switching logics.
VI. ACKNOWLEDGMENT
This work was partially supported by the Norwegian
Center of Offshore Wind Energy (NORCOWE) under grant
675

Das könnte Ihnen auch gefallen