Вы находитесь на странице: 1из 9

Control and Intelligent Systems, Vol. 36, No.

4, 2008

THE DESIGN OF SLIDING MODE CONTROL


SYSTEM BASED ON BACKSTEPPING
THEORY FOR BTT UAV
J. Yao, , X. Zhu, and Z. Zhou

and faster response than STT mode. However, the model


of BTT UAV is seriously nonlinear and coupled, and the
aerodynamic coecients vary greatly in the dierent ight
altitudes and speeds. Although the linear control theory
has been mature, it cant handle BTT UAV because its
model cant be linearized employing traditional method.
In the ight control system design for BTT UAV,
nonlinear control theories have more advantages than the
linear control theories. Among the nonlinear control theories, the dynamic inversion (DI) has been applied to design
ight control system [1]. In the approach employing DI,
the whole system is divided into two subsystems on the
basis of time-scale theory [2, 3]. The fast subsystem should
respond much faster than the slow subsystem. It is an
approximated method and the gains of the subsystems are
hard to choose because the analysis of stability is very
complicated [4].
Backstepping is another nonlinear control theory combined with DI, which is an approach guaranteeing the stability based on Lyapunov function [58]. It does not consider zero-dynamics and has more advantages than DI. In
the original backstepping algorithm, it assumes the model
is accurate without uncertainty. In addition, backstepping
is easy to combine PID to design control system [9, 10].
However, it has small robustness to attenuate the model
uncertainty. In the real ight of BTT UAV, a complete and
accurate dynamic model is dicult to obtain. Although
the aerodynamic coecients can be identied online, it is
not as accurate as we expect because the system is nonlinear. To get the robust control system, backstepping theory is usually combined with the robust control theories.
Many of these approaches have been studied. Backstepping theory combined with Neural Network (NN) methods
are presented in [8, 11, 12]; Backstepping theory combined
with sliding mode control (SMC) methods are presented
in [13, 14]; Backstepping theory combined with adaptive
control methods are presented in [8, 1517].
In this paper, we propose a novel nonlinear control
approach of ight control system design based on backstepping and SMC for BTT UAV. In this approach, it does
not require the separation of system variables and identication of aerodynamic coecients online. The eects of
parametric perturbation are compensated by SMC. Being

Abstract
A novel control system design approach is proposed based on
backstepping theory and sliding mode control (SMC) for BankTo-Turn (BTT) Unmanned Aerial Vehicle (UAV). It ensures BTT
UAV stable and accurate ight under large parametric perturbation.
In addition, the aerodynamic coecients are not necessary to be
identied online. This approach is based on backstepping theory
and the whole system is not divided into slow and fast subsystems.
However, backstepping cannot ensure the robustness of the closedloop system.

To solve this problem, SMC is employed, which

is designed in terms of the bounds of aerodynamic coecients.


To evaluate the performance of the ight control system using
the proposed approach, the three channels united simulation is
conducted considering the actuators rate and magnitude saturation.
The results show that the proposed approach is capable of handling
serious nonlinear and large model uncertainty.

Key Words
Unmanned Aerial Vehicle, Bank-To-Turn, backstepping, sliding
mode

1. Introduction
In the past few years, a variety of control theories have been
applied to the ight control system design of Unmanned
Aerial Vehicle (UAV). In the ight control, there are two
control modes: Skid-To-Turn (STT) and Bank-To-Turn
(BTT). In the STT mode, the UAV is controlled in pitch
and yaw separately, and the roll channel is stabilized. In
the BTT mode, the command of acceleration is produced in
the pitch channel and the direction is controlled by the roll
channel. Meanwhile, the sideslip angle is required as small
as possible. BTT mode can generate bigger acceleration

School of Astronautics, Northwestern Polytechnical University, Xian, Shaanxi 710072, P.R.China; e-mail: jianguo@
cs.mcgill.ca
School of Computer Science, McGill University, Montreal, QC
H3A2A7, Canada
UAV Research and Development Centre, Northwestern Polytechnical University, Xian, Shaanxi 710072, P.R.China; e-mail:
{zhuxiaoping, zhouzhou}@nwpu.edu.cn
Recommended by Prof. Jonathan Wu
(paper no. 201-2010)

347

dierent from the existing methods combining backstepping and SMC, the design of SMC in our approach is based
on the bounds of aerodynamic coecients. Finally, the stability of the proposed approach is proven using Lyapunov
function.
The remainder of this paper is organized as follows. In
Section 2, the problem of design BTT UAV ight control
system using backstepping is stated. In Section 3, a
new nonlinear control system design approach combining
backstepping and SMC is presented. In Section 4, the
proposed approach is applied to BTT UAV and the control
law is derived. A simulation study is conducted and the
results are summarized in Section 5. Finally, conclusions
are discussed in Section 6.

The control law employing the original backstepping


and PID is described as:
xd2 = g11 (x1 )[k1 z1 f1 (x1 ) + x d1 ]


u = g21 k2 z1 f2 (x) + x d2

where k1 , k2 R are positive parameters. We can prove the


system with the above control law is stable when the model
uncertainty d1 (x1 , x2 ) and d2 (x1 , x2 ) is zero or very small
[19]. Note that the PID controller has small robustness
to the model uncertainty. When the model uncertainty is
very large, the system performance may deteriorate. To
make the control system more robust, we employ SMC to
make the controller compensate the model uncertainty. In
the next sections, we will discuss our detailed approach.

2. Problem Statement

3. A Novel Control System Design Combining


Backstepping with SMC

This paper focuses on the ight control system design


for BTT UAV. In general, BTT guidance law generates
the commands including rolling angle c , and accelerations Ayc , Azc . However, the system with the outputs
[c , Ayc , Azc ] is a non-minimum phase phenomenon, which
results in instability when we exert DI. One way to solve
this problem is to redene the outputs as [c , c , c ] [18].
However, when BTT UAV ights with big angle of attack,
the control of c may lead to big sideslip angle. To ensure
the sideslip angle in the range of a small value, c is replaced by the bank angle command c . So the outputs of
BTT UAV is [c , c , c ]. Furthermore, this modication
ensures that the BTT UAV is stabler under the large angle
of attack. The detailed dynamic model can be referred in
the Appendix.
To present the problem conveniently, the model of
BTT UAV is described as:

In this section, we propose a novel control system design


approach based on backstepping and SMC for a general
nonlinear system under model uncertainty. Backstepping
is used to handle the model nonlinearity and SMC makes
the system robust in presence of the model uncertainty. In
the preliminary design in Section 3.1, there are derivatives
of the pseudo control variables. However, these derivatives
do not exist because they are not continuous. To deal
with this problem, we modify the pseudo-control signal in
Section 3.2. Finally, the stability of the control system is
proven in Section 3.3.
3.1 The Preliminary Design
Considering the following nonlinear system:

x 1 = f1 (x1 ) + g1 (x1 )x2 + d1 (x1 , x2 )


x 2 = f2 (x2 ) + g2 u + d2 (x1 , x2 )

x i = xi+1 + xi , i = 1, 2, . . . , n 1

(1)

x n = f (x) + g(x)u + d(x)


y = x1

n
]T and
where x1 = [ ]T , x2 = [p q r]T , u = [l m
T

a , r , e are calculated from the input u = [l m


n
] in (57).
d1 (x1 , x2 ) and d2 (x1 , x2 ) are the parametric uncertainty
aected by the aerodynamic coecients. The variation of
aerodynamic coecients may be great in the ight of BTT
UAV, and we assume that they have bounds as follows:
|d1 (x1 , x2 )| < dmax
, |d2 (x1 , x2 )| < dmax
1
2

(2)

si = xi xic , i = 1, 2, . . . , n

(7)

where x1c = y, x2c , x3c , . . . , xic are pseudo control variables.


To make the control system design approach easy to
understand, we introduce it in n steps.
Step 1. Dening the pseudo-control variable x2c as:

z1 = x1 xd1
(3)

x2c = k1 s1 + x 1c ud1

where xd1 and xd2 are the desired commands of x1 and x2 .


Using (1) and (3), we can get that:

(8)

where ud1 is the item corresponding to SMC, k1 > 0.


Calculating the time derivative of s1 :

z1 = f1 (x1 ) + g1 (x1 )x2 x d1 + d1 (x1 , x2 )


z2 = f2 (x2 ) + g2 u x d2 + d2 (x1 , x2 )

(6)

where x = [x1 , x2 , . . . , xn ]T is the state variable vector,


u, y are the input and output, respectively. xi is the
model uncertainty. f (x) and g(x) are continuous and
dierentiable with respect to x. d(x) is the disturbance.
Dening sliding planes:

where dmax
and dmax
are constant vectors.
1
2
We dene the error states variables z1 , z2 R3 as:

z2 = x2 xd2

(5)

(4)

s 1 = s2 + x2c + x1 x 1c
348

(9)

Step n: Dening the control input u as:

Substituting (8) into (9), we can get:


s 1 = s2 k1 s1 + x1 ud1

u = g 1 (x)[f (x) + x nc kn sn sn1 udn ]

(10)

Dening Lyapunov function as:


V1 =

1 2
s
2 1

where udn is the item corresponding to SMC, kn > 0.


Calculating the time derivative of sn :
(11)
s n = f (x) + g(x)u + d(x) x nc

Calculating the time derivative of V1 and substituting


(10), we can get:
V 1 = k1 s21 + s1 s2 + s1 (x1 ud1 )

s n = kn sn sn1 + d(x) udn

(12)

(13)

Vn =

V n =

n

j=1

(14)

i1
1 2 
si +
Vj
2
j=1

j=1

sj (xj udj )+sn (d(x)udn ) (24)

j=1

(25)

(16)

3.2 The Modification of Pseudo Control

(17)

In the control law design of Section 3.1, there are derivatives of the pseudo control variables. However, these
pseudo control variables are not dierentiable because they
are not continuous actually. To deal with this problem, the
pseudo control variables should be redened.
In this subsection, we redene pseudo control variables as:
xi+1,c = si1 ki si + x ic , i = 1, 2, . . . , n1, s0 = 0 (26)

(18)
and dene:

j=1

u = g 1 (x)[f (x) + x nc kn sn sn1 ud ]

We select ud1 in terms of the following inequation:


sj (xj udj ) < 0

n1


V n is negative denitive on the condition of (13),


(19) and (25). So the control system is stable and the
performance of system completely depends on the selection
of parameters ki and ui , i = 1, 2, . . . , n.

(15)

Calculating the time derivative of Vi and substituting


(16), we can get:
i
i


V i =
kj s2j + si si+1 +
sj (xi udj )

kj s2j +

sn (d(x) udn ) < 0

Dening Lyapunov function as:


Vi =

(23)

Selecting udn in terms of the following inequation:

Substituting (14) into (15), we can get:


s i = si+1 si1 ki si + xi udi

n1
1 2 
sn +
Vj
2
j=1

Calculating the time derivative of Vn and substituting


(22), we can get:

where udi is the item corresponding to SMC, ki > 0.


Calculating the time derivative of si :
s i = si+1 + x(i+1)c + xi x ic

(22)

Lyapunov function is dened as:

Apparently, if s2 = 0, the time derivative of Lyapunov


function V 1 < 0 so that the variable s1 is stable. However,
the assumption s2 = 0 is not ensured actually. So we should
design pseudo-control variable x3c to make s2 converge to
0 asymptotically in the next step.
Step i (i = 2,. . . ,n 1): Dening the pseudo-control
variable x(i+1)c as:
x(i+1)c = si1 ki si + x ic udi

(21)

Substituting (20) into (21), we can get:

We select ud1 in terms of the following inequation:


s1 (x1 ud1 ) < 0

(20)

(19)

(27)

where ud is the item corresponding to SMC. The model


uncertainty of the whole system can be compensated in
the last step of the control system design. The pseudo
control variables will have derivatives with the above modications.
To get the SMC item ud , we dene it in the following
formulation:

Apparently, if si+1 = 0, the time derivative of Lyapunov function V i < 0 so that the variables s1 , s2 , . . . , si
are stable. Similar with Step 1, we should design pseudo
control variable xi+1,c to make si+1 converge to 0 asymptotically.
349

4. The Flight Control System Design for BTT UAV


ud =

n


udj = udn +

j=1

n1


udj = udn +

j=1

n1


sn wj (x)

(28)

In this section, we design the ight control system exerting


the proposed approach in Section 3. In this control law,
there are two gain matrix K1 and K2 to choose. The item
of SMC is designed in terms of the bounds of aerodynamic
coecients.
We dene sliding planes as:

j=1

Selecting control law as follows:

wj (x) =

>

sj xj
s2n , sn

>

= 0, s <
n

udn

, j = 1, 2, . . . , n 1

> |d(x)|, s > 0


n
=
< |d(x)|, s < 0

(29)

S = [eT1

eT2 ]T = [xT1 xT1c

xT2 xT2c ]T

(35)

where x1c = [c c c ]T is input command vector.


Dening pseudo control as:
(30)

x2c = g11 (x1 )[K1 e1 f1 (x1 ) + x 1c ]

(36)

where K1 is diagonal and positive denite gain matrix.


The control law is dened as:

To prevent saturation of the actuators, wj (x) is set as


zero in (29) when sn < . Note that sj , j = 1, 2, . . . , n 1,
is also small when sn is small, so there is no need to add
SMC item.
Although the modication of pseudo control cannot
ensure V i < 0, i = 1, 2, . . . , n 1, the stability of the whole
system is ensured employing the proposed control law. In
the next part, the stability analysis will be given.

u = g21 [K2 e2 f2 (x2 ) + x 2c g1T (x1 )e1 ud ]

(37)

where K2 is diagonal and positive denite gain matrix,


ud is the item corresponding to SMC to compensate the
variation of aerodynamic coecients.
Substituting the control law (36) and (37) into the
system model (1), we can get:

3.3 The Stability Analysis


e 1 = K1 e1 + d1 (x1 , x2 )
Dening Lyapunov function of the whole system as:
1 2
1 T
s
S S=
2
2 j=1 j

e 2 = K2 e2 g1T (x1 )e1 + d2 (x1 , x2 ) ud

V =

Lyapunov function is dened as:

(31)

V =

where

1 T
S S
2

(39)

Calculating the time derivative of V and substituting


(38), we can get:
S = [s1

s2

sn ]

(32)
V = eT1 K1 e1 eT2 K2 e1 + eT1 d1 + eT2 d2 eT2 ud

Calculating the time derivation of Lyapunov function,


we can get:

V = 2

(38)

n

j=1

n1


kj s2j + 2

sj xj + 2sn d(x) 2sn ud

(40)

To make sure V < 0, we should choose the SMC item


in terms of the following inequation:
eT1 d1 + eT2 d2 eT2 ud < 0

(33)

(41)

j=1

To this end, we dene the formulation of ud as:


Substituting (28) and in terms of (29) and (30), we can
get, if sn > so that V < 0. It implies that the system is
stable; if sn < so sj , j = 1, 2, . . . , n 1, is also small.
n
n1


V < 2
kj s2j + 2
sj xj
j=1

ud = ud1 + ud2 = e2 (x) + ud2

(42)

Substituting (42) into (41):


eT1 d1 eT2 e2 (x) + eT2 (d2 ud2 ) < 0

(34)

(43)

j=1

We choose the sucient condition of (43) as follows:


If sj > xj /kj , j = 1, 2, . . . , n 1, so that V < 0. It
implies the system is Uniformly Ultimately Bounded
(UUB) [20].

eT1 d1 eT2 e2 (x) < 0


eT2 (d2
350

ud2 ) < 0

(44)
(45)

Calculating the parameters (x) of the SMC item from


(44). It is dened as:
(x) >

eT1 d1
e2 

(46)

Considering the bounds of aerodynamic coecients, so


(46) is written as:

T
> |e1 d1 |max , e  >
2
e2 
(x) =
= 0 ,e  <
2
Figure 1. Tracking responses of angle of attack.

1
1
|e11 ||L|max +
|e12 |(|D|max ||
MV
MV
1
+ |Y |max ) +
|e12 |(|D|max | cos tan |
MV
+ |Y |max |cos tan | + |L|max

|eT1 d1 |max =

|sin tan + |)

5.2 The Results Evaluation


Fig. 1 shows the tracking responses using the controller
based on backstepping and SMC. As we can see, the control
system ensures that the angle of attack tracks the reference
command without static error in both the conditions.

(47)

where e1i e2i are ith element of the error vectors e1 and e2 .
Calculating the parameters ud2 of the SMC item from
(45), it is presented as follows:

< c |l|
3
max c4 |n|max , e21 < 0
ud2 (1) =
> c |l|
+ c |n|
,e > 0
3

max

max

< c |m|
7
max , e22 < 0
ud2 (2) =
> c |m|
,e > 0
7

max

21

22

< c |l|
4
max c9 |n|max , e23 < 0
ud2 (3) =
> c |l|
+ c |n|
,e > 0
4

max

max

(48)

23

Figure 2. The responses of sideslip angle.


The responses of sideslip angle using the controller
based on backstepping and SMC are illustrated in Fig.
2. We can get that, when there is perturbation, the
sideslip angle is a bit bigger than the situation there is
no perturbation, but both of them are controlled to be
less than 2 . Note that one of the demanded control
performances in BTT UAV is to ensure sideslip in the
range of 3 , so that the performance of the controller is
eligible.
From Fig. 3, we can get the tracking responses of bank
angle using the controller based on backstepping and SMC.

The maximum perturbation of aerodynamic forces and


moments are dened in Appendix. From the above derivation, we can get the control law consisting of (36), (37),
(47) and (48).
5. Numerical Simulation
In this section, we make a simulation to evaluate the performance of the proposed control law based on backstepping and SMC applied to BTT UAV. The results show
that the backstepping combined with SMC controller can
handle serious nonlinearity and large model uncertainty.
5.1 The Simulation Setup
The simulation is conducted in Matlab [21]. In the simulation, the parameters are set as follows: the UAV ights at
the altitude of 12,000 m and the speed of 0.9 Ma; the magnitude limitation of actuators is 20 , and the maximum
rate limitation is 300 /s; the perturbation of aerodynamic
coecients is in the bound of 30%; the angle of attack
command is a step signal of 10 , the bank angle command
is a step signal of 90 and the sideslip angle is expected at
zero.

Figure 3. Tracking responses of bank angle.


351

c3 l + c4 n

d2 (x1 , x2 ) = c7 m

c4 l + c9 n

When there is no perturbation, the response is a little


better than the situation perturbation exists. However,
the overshoot is less than 11% in both the conditions, so
the control performance is acceptable. Fig. 4 shows the
deection responses of actuators using the controller based
on backstepping and SMC.

cos

g1 (x1 ) = sin

cos

c
3

g2 = 0

c4

1
0
0

cos

sin

(51)

0
c7
0

c4

c9

+ Px sin Pz cos )
M1V (L

(52)

+ g cos V cos

1
+ Y Px cos + Py )

(
D
MV

+ g sin V cos

f1 (x1 ) =

D cos tan + Y cos tan

+L(sin tan + ) + PX (sin sin

M V tan cos cos tan )

+PY cos sin PZ ( sin cos

tan + cos sin tan + cos )

g cosV cos
(53)

(c1 r + c2 p)q

f2 (x2 ) = c5 pr c6 (p2 r2 )
(54)

(c8 p c2 r) q

Figure 4. Tracking responses of actuator deections.


6. Conclusions
An SMC system combined with backstepping for BTT
UAV has been proposed. It can deal with nonlinearity and
model uncertainty. In this approach, it doesnt require that
the system be separated into slow and fast subsystems and
the aerodynamic coecients be identied online. In the
SMC design, the bounds of the aerodynamic coecients are
needed to make the system robust. To prove the stability of
the proposed approach, Lyapunov function is designed and
the proof is given. Finally, the numerical simulation shows
that the BTT UAV employing the proposed approach
maneuvers with good performance in the presence of large
parametric perturbation.
Acknowledgement
This work was supported in part by an NCET Grant
No.05-0867 and a National Study-Abroad Scholarship of
P.R.China under Grant No. [2007] 3020.

The nominal values of aerodynamic forces are dened as:




CA
CA

+ CL
+ CLe e qS
L = CL + CLq q
2V
2V


b
b

Y = CY + CY p p
+ CY r r
+ CY a a + CY r r qS
2V
2V
= (CD0 + CD )
D
qS
(55)

Appendix
When the UAV is in the BTT mode, the sideslip angle
is usually in the range of 3 , on this condition cos
1, sin tan = . The attack angle is usually 020 , so
sin sin tan sin 0. The function (1) is dened as
follows:

(50)

The perturbation value of lift force is dened as:



CA
CA
L = CL + CLq q
+ CL
2V
2V

(56)
+ CLe e qS

D + Y
1

d1 (x1 , x2 ) =

M V D cos tan + Y cos tan

+L(sin tan + )
(49)

Y, D are calculated similar with L.


The nominal values of aerodynamic moments are dened as:
352

Functions:
L, D, Y :
D,
Y :
L,



l = Cl + Clp p b + Clr r b + Cl a + Cl r qSb
a
r
2V
2V


CA
CA

m
= Cm0 + Cm + Cmq q
+ Cm
+ Cme e qSCA
2V
2V


b
b

n
= Cn + Cnp p
+ Cnr r
+ Cna a + Cnr r qSb
2V
2V
(57)

lift, drag, and side-force


nominal values of lift,
drag, and side-force
L, D, Y :
perturbations of lift,
drag, and side-force
|L|max , |D|max , |Y |max : maximum perturbations
of lift, drag, and side-force
l, m, n:
roll, pitch, and yaw
moment
l, m,
n
:
nominal values of roll,
pitch, and yaw moment
l, m, n:
perturbations of roll,
pitch, and yaw moment
|l|max , |m|max , |n|max : maximum perturbations
of roll, pitch, and yaw
moment
Px , Py , Pz :
thrust on three body axis

The perturbation value of roll moment is dened as:



b
b
l = Cl + Clp p
+ Clr r
+ Cla a
2V
2V

(58)
+Clr r qSb

Variables:
:
V:
q:
:

m, n are calculated similar with l.


The maximum perturbation of lift force is dened as:

|L|max


CA
= |CL |max || + |CLq |max |q|
2V

CA

+|CL |max ||
+ |CLe |max |e | qS
2V
(59)

:
:
p, q, r:
c :
Ayc , Azc :

|Y |max , |D|max are calculated similar with |L|max .


The maximum perturbation of roll moment is dened as:

c :
c :
c :
e , a , r :


b
|l|max = |Cl |max || + |Clp |max |p|
+ |Clr |max |r|
2V

b

+ |Cla |max |a | + |Clr |max |r | qSb


2V
(60)

References
[1] D.J. Bugajski, D.F. Enns, & M.R. Elgersma, A dynamic
inversion based control law with application to high angle
of attack research vehicle, Proc. AIAA Conf. on Guidance,
Navigation and Control, Portland, OR, 1990, 826839.
[2] P.K.A. Menon, M. Badgett, & R.A. Walker, Nonlinear ight
test trajectory controllers for aircraft, Journal of Guidance,
Control, and Dynamics, 10 (1), 1987, 6772.
[3] R. Srivastava, A. Sarkar, & S. Gollakota, Nonlinear inversion
ight control for a super maneuverable aircraft, Journal of
Guidance, Control, and Dynamics, 15 (4), 1992, 976984.
[4] C. Schumacher & P.P. Khargonekar, Stability analysis of a
missile control system with a dynamic inversion Controller,
Journal of Guidance, Control, and Dynamics, 21 (3), 1998,
508515.
[5] O. Harkegard & S. Torkel Glad, A backstepping design for
ight path angle control, Proc. 39th IEEE Conf. on Decision
and Control, Sydney, Australia, 2000, 35703575.
[6] M. Xin & S.N. Balakrishan, Nonlinear missile autopilot design
theta-D with technique, Journal of Guidance Control and
Dynamics, 27 (3), 2004, 406417.
[7] M.A. Khan & P. Lu, New technique for nonlinear control of
aircraft, Journal of Guidance, Control, and Dynamics, 17 (5),
1994, 10551060.
[8] M.M. Polycarpou & P.A. Ioannou, A robust adaptive nonlinear
control design, Automatica, 32 (3), 1996, 423427.

|m|max , |n|max are calculated similar with |l|max .


Nomenclature
Constants:
C# :
m:
g:
Ixx :
Iyy :
Izz :
Ixz :
ci :
CA :
b:

ight-path angle
air speed
dynamic pressure
roll angle about velocity
vector or bank angle
angle of attack
sideslip angle
roll, pitch, yaw angular
rates (body axis)
command of roll angle
command of acceleration
along the body axis
command of bank angle
command of angle of attack
command of sideslip angle
elevator, aileron, rudder
angle

aerodynamic coecient of #
mass
vertical component of gravity
roll moment of inertia
pitch moment of inertia
yaw moment of inertia
product moment of inertia
element of inverse-inertial matrix
mean geometric chord
wing span
353

Xiaoping Zhu was born in Hunan, P.R. China. He received


his Ph.D. degree from the Northwestern Polytechnical University (NPU), Xian, Shaanxi, P.R.
China, in 1992. He did postdoctoral research on ight control at
the Nanjing University of Aeronautics and Astronautics, Nanjing, Jiangsu, P.R.China, from
1993 to 1995. And then he did
postdoctoral research on rocket
design at the NPU from 1995 to 1997. Currently, he is a
Professor at the NPU. His research interests are conceptual
design, ight control and simulation for UAV.

[9] B. Hemici, L. Nezli, M. Tadjine, & M.S. Boucherit, Robust


PID/backstepping control design for permanent magnet synchronous motor drive, Control and Intelligent Systems, 34 (3),
2006, 194204.
[10] R. Skjetne & T.I. Fossen, On integral control in backstepping: analysis of dierent techniques, Proc. American Control
Conference, Boston, MA, 2004, 18991904.
[11] J.Y. Choi & J.A. Farrell, Adaptive observer backstepping
control using neural networks, IEEE Transactions on Neural
Networks, 12 (5), 2001, 11031112.
[12] O. Kuljaca, N. Swamy, F.L. Lewis, & C.M.A.K.C.M. Kwan,
Design and implementation of industrial neural network controller using backstepping, IEEE Transactions on Industrial
Electronics, 50 (1), 2003, 193201.
[13] J. Li, D. Xu, B. Song, & W. Yan, An easily implementable
backstepping sliding mode controller for nonlinear system with
unmatched uncertainties, Journal of Northwestern Polytechnical University, 22 (2), 2004, 145148.
[14] P. Carbonell, Z.P. Jiang, & D.W. Repperger. Nonlinear control
of a pneumatic muscle actuator: Backstepping vs. slidingmode, Proc. IEEE Conf. on Control Applications, Mexico City,
Mexico, 2001, 167172.
[15] Y. Zhang, B. Fidan, & P.A. Ioannou, Backstepping control
of linear time-varying systems with known and unknown parameters, IEEE Transactions on Automatic Control, 48 (11),
2003, 19081925.
[16] J. Zhou, C. Wen, & Y. Zhang, Adaptive backstepping control of
a class of uncertain nonlinear systems with unknown backlashlike hysteresis, IEEE Transactions on Automatic Control,
49 (10), 2004, 17511759.
[17] Y. Zhang, C. Wen, & Y.C. Soh, Adaptive backstepping control
design for systems with unknown high-frequency gain, IEEE
Transactions on Automatic Control, 45 (12), 2000, 23502354.
[18] C.-K. Lin & S.-D. Wang, An adaptive H/sup /spl inn//
controller design for bank-to-turn missiles using ridge Gaussian
neural networks, IEEE Transactions on Neural Networks,
15 (6), 2004, 15071516.
[19] A.R. Benaskeur & A. Desbiens, Backstepping-based adaptive PID control, IEE Transactions on Control Theory and
Applications, 149 (1), 2002, 5459.
[20] F.L. Lewis, C.T. Abdallah, & D.M. Dawson, Control of robot
manipulators (Englewood Clis, NY: Macmillan, 1993).
[21] B. Stevens & F. Lewis, Aircraft control and simulation (Englewood Clis, NY: Wiley, 1992).

Zhou Zhou was born in Hunan,


P.R. China. She received her
Ph.D. degree from the Northwestern Polytechnical University (NPU), Xian, Shaanxi, P.R.
China, in 1992. She did postdoctoral research on UAV conceptual
design at the Nanjing University
of Aeronautics and Astronautics,
Nanjing, Jiangsu, P.R. China,
from 1993 to 1995. And then
she did postdoctoral research on
integrated optimization design of aircraft aerodynamics
and stealth performances at the NPU. Currently, she is a
Professor at the NPU. Her research interests are conceptual
design, ight dynamics and ight control for UAV.

Biographies
Jianguo Yao was born in Hebei,
P.R. China.
He received his
B.E. and M.E. degrees from the
Northwestern Polytechnical University (NPU), Xian, Shaanxi,
P.R. China, respectively in 2004
and 2007. He received the Chinese State Scholarship to Study
Abroad in 2007. Currently, he is
a Ph.D. Student at the NPU, and
a Joint Education Ph.D. Student
at the McGill University, Montreal, QC, Canada. His research interests are robust and
nonlinear control, stochastic control and guidance system
for UAV.

354

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Вам также может понравиться