Вы находитесь на странице: 1из 13

IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 21, NO.

6, NOVEMBER 2013 2251


Attitude Stabilization of Spacecrafts Under
Actuator Saturation and Partial Loss
of Control Effectiveness
Bing Xiao, Qinglei Hu, and Peng Shi, Senior Member, IEEE
AbstractA practical solution is presented to the problem
of fault tolerant attitude stabilization for a rigid spacecraft
by using feedback from attitude orientation only. The attitude
system, represented by modied Rodriguez parameters, is con-
sidered in the presence of external disturbances, uncertain inertia
parameters, and actuator saturation. A low-cost control scheme
is developed to compensate for the partial loss of actuator
effectiveness fault. The derived controller not only has the
capability to protect the control effort from actuator saturation
but also guarantees all the signals in the closed-loop system to be
uniformly ultimately bounded. Another feature of the approach
is that the implementation of the controller does not require any
rate sensor to measure angular velocity. An example is included
to verify those highly desirable features in comparison with the
conventional velocity-free control strategy.
Index TermsActuator saturation, attitude stabilization, fault
tolerant control (FTC), uniformly ultimately bounded, velocity-
free.
I. INTRODUCTION
O
NCE A spacecraft is launched, it is highly unlikely
that its hardware can be repaired if any fault occurs.
Thus any component or system fault cannot be xed with
replacement parts. This issue can potentially cause a host of
economic and safety problems. A classic example was the
accident of GPS BII-07, a spacecraft in the NAVSTAR GPS
constellation developed by the US Department of Defense.
It suffered a reaction wheel failure that led to three-axis
stabilization failure and a total loss of the spacecraft [1]. In
practical aerospace engineering, a common and cost-effective
practice is to plan space missions with the ability to execute
Manuscript received November 19, 2011; revised August 31, 2012; accepted
December 10, 2012. Manuscript received in nal form December 20, 2012.
Date of publication January 23, 2013; date of current version October 15,
2013. This work was supported in part by the National Natural Science Foun-
dation of China under Grant 61004072, Grant 61273175, Grant 61174058,
and Grant 61134001, the Program for New Century Excellent Talents in
University under Grant NCET-11-0801, the Heilongjiang Province Science
Foundation for Youths under Grant QC2012C024, the National Key Basic
Research Program of China under Grant 2012CB215202, and the 111 Project
B12018. Recommended by Associate Editor N. E. Wu.
B. Xiao and Q. L. Hu are with the Department of Control Science and
Engineering, Harbin Institute of Technology, Harbin 150001, China (e-mail:
bxiaobing@gmail.com; huqinglei@hit.edu.cn).
P. Shi is with the School of Engineering and Science, Victoria University,
Melbourne, 8001 VIC, Australia, and also with the School of Electrical
and Electronic Engineering, The University of Adelaide, Adelaide SA 5005,
Australia (e-mail: peng.shi@vu.edu.au).
Color versions of one or more of the gures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TCST.2012.2236327
a safe mode transition in which a detected anomaly deacti-
vates actuators and postpones the payload activities until the
situation is resolved by ground intervention. This solution is
sufcient to tolerate a variety of faults so long as the spacecraft
faces no immediate risk and provided the mission can be
continued following a potentially extended safe mode episode.
For instance, the far ultraviolet spectroscopic explorer satellite
went into a safe mode due to failure of two reaction wheels.
The mission engineers managed to develop a new control
law to reestablish ne pointing capability and to recover
the satellite mission by integrating the magnetic torquer
bars [2].
During the critical phases of some high real-time missions,
safe mode is not an option. Consider a military satellite
tasked with providing coverage of a specic high-priority
area. If this satellite goes to safe mode, substantial loss of
ground objectives could occur. As a result, on-line and real-
time fault tolerant control (FTC) design for spacecraft has
received considerable attention [3][5]. In [6], the problem
of automated attitude recovery for rigid and exible spacecraft
was investigated on the basis of feedback linearization control.
Variable structure reliable control design was presented in [7]
to perform attitude stabilization maneuver. Passive and active
reliable control laws were synthesized with an observer iden-
tifying actuator faults. Another fault-tolerant attitude-tracking
control to compensate for the loss of reaction wheel effec-
tiveness fault was discussed in [8]. Only the disturbance-free
case was considered; the knowledge of the spacecraft inertia
was needed to implement the controller. Cai et al. [9] derived
an adaptive controller to follow the desired attitude trajecto-
ries with thruster failures. Input constraint, uncertain inertia
parameters, and external disturbance were explicitly addressed.
In a related work [10], a sliding-mode-based adaptive fault-
tolerant controller was proposed for a satellite. The designed
scheme can achieve the attitude control in the presence of
unknown, slow-varying satellite mass distribution, and several
fault scenarios of rotating solar aps. In [11], an adaptive
fault-tolerant attitude-tracking controller was developed for a
exible spacecraft. Although terminal sliding-mode controller
was designed in [12] to achieve satellite formation ying, the
stability analysis was not carried out in the case of faults. To
recover control capabilities in a faulty situation as quickly as
possible, a terminal sliding mode control scheme was proposed
in [13] to perform the rest-to-rest maneuver of a satellite with
the degradation of actuation effectiveness.
1063-6536 2013 IEEE
2252 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 21, NO. 6, NOVEMBER 2013
The preceding FTC schemes assume that the measurement
of angular velocity is exactly available by using rate gyros. In
practice, however, the availability of angular velocity measure-
ments is not always satised due to either cost limitations or
implementation constraints. The occurrence of faults in sen-
sors also leads to wrong/imprecise measurements of angular
velocity [14]. Hence, the design of efcient and low-cost
attitude control without angular velocity measurements is
becoming a very challenging task. The earliest known result
in the eld of velocity-free control was examined in [15] and
[16]. A passivity lter was incorporated to generate a velocity-
related signal from attitude orientation measurements. Subse-
quent extension to this control design scheme was discussed
in [17]. Although the inevitable factors of uncertain system
parameters were explicitly addressed, external disturbance
was not considered. In [18], a new class of proportional
integral controller was derived on the basis of a rst-order
passivity lter to achieve attitude stabilization without velocity
feedback. The stability analysis in case of disturbance was
not carried out. In [19], an adaptive controller by employing
a ltering technique was proposed. System uncertainties and
external disturbance were estimated by using a Chebyshev
neural network.
Another solution to examine the challenge associated with
the unavailability of angular velocity is to construct model-
based nonlinear observers, as suggested in [20]. In that
approach, a class of velocity observers was introduced to esti-
mate the joint velocities in rigid manipulators. The controller
design was not discussed. In [21], Ren presented a solution to
the distributed cooperative attitude synchronization and track-
ing problems for multiple rigid bodies. The implementation of
the controller was independent of the absolute angular velocity
measurements. However, actuator faults were not considered.
In [22], an observer was designed to estimate the elastic
modes, their rates, and unknown angular velocity. A model-
independent and observer-based attitude control method was
explored in [23]. The controller met the objective of attitude
tracking. However, a full knowledge of inertia matrix was
needed, and it was assumed that the spacecraft is free from
external disturbances. In order to release these two restric-
tions, a new attitude-tracking control scheme without velocity
measurements was developed in [24]. The design was inertia-
independent and robust to external disturbances. In addition
to the above-mentioned two velocity-free design methods,
several other control schemes on the basis of quaternion output
only were also motivated with Lyapunov-based techniques
[25], [26].
Although various nonlinear control schemes are available
to accomplish various attitude maneuvers with fault-tolerant
capability guaranteed even in the absence of angular veloc-
ity measurements, few of the previously discussed literature
account for actuator saturation. Due to physical limitations,
spacecraft attitude control system is subject to actuator sat-
uration. The typical control actuators mounted in a space-
craft, such as the reaction wheel and thruster, have an upper
bound on the control torque they can exert on the attitude
system. If the attitude system is not equipped with a novel
control algorithm to solve this kind of nonlinearity, then
actuator saturation may lead to attitude control performance
deterioration or even system instability. Consequently, actuator
saturation is another key issue that needs to be addressed,
and a great deal of attention has been focused on controller
design with saturated actuators [27]. The endeavor to handle
actuator saturation is complex and multidimensional. The
most prominent method among the solutions for actuator
saturation is the antiwindup design, due to its simple structure.
In particular, a simple but effective modied proportional
integralderivative control was presented in [28] to achieve
large-angle attitude control of the spacecraft. In that paper,
the specic advantage for handling actuator saturation with
antiwindup control was veried. An alternative algorithm was
proposed in [29]. A new positive constant gain within the
framework of conventional backstepping-based control design
was derived to reduce the peak control torque. More recently, a
nonlinear adaptive control was investigated in [30]. A feedback
and a feed-forward component were incorporated to solve the
input saturation problem and to accommodate any linearly
parameterized disturbance.
It should be pointed out that almost all research until now
has dealt with attitude control design either under component
faults, or without angular velocity measurements, or with
actuator saturation. For instance, velocity-free attitude con-
troller was designed in [31] for spacecraft formation ying
subject to actuator constraints, but actuator faults were not
investigated. In [32], although the problem of robot tracking
control with bounded torque input was solved by using output
feedback only, the problem of actuator fault tolerance was
not solved. The integrated design to deal with actuator faults,
unavailable angular velocity, and actuator saturation still needs
further research. Actually, if an unknown fault occurs in an
actuator, then in spite of the fault, the attitude control system
would continue issuing its maneuver that may no longer be
achievable by the spacecraft. As a consequence, the required
control effort will quickly saturate the actuator while striving
to maintain the healthy attitude maneuvering performance,
and subsequently will destabilize the attitude. This situation
may quickly become mission critical. On the other hand, direct
high-precision velocity measurement is usually unavailable for
spacecraft in the presence of sensor faults. With a view to
tackle the above-stated challenges and potential problems, we
investigate in this paper the feasibility of performing desired
attitude maneuver by developing a dynamic control scheme
in the presence of partial loss of actuator effectiveness fault,
external disturbances, and uncertain inertia parameters.
The main contributions of this paper are as follows.
1) A fault-tolerant control scheme is proposed to compen-
sate for the partial loss of actuator effectiveness fault
without any fault detection and isolation mechanism.
The closed-loop attitude system is uniformly ultimately
bounded stable, when the external disturbances are
bounded.
2) In contrast to the attitude control schemes without
angular velocity measurements available in literature,
the proposed velocity-free control approach can tolerate
actuator fault with attitude successfully stabilized. More-
over, the approach can explicitly account for actuator
XIAO et al.: ATTITUDE STABILIZATION OF SPACECRAFTS 2253
saturation without any need of magnetic torquers to
dump the excess momentum of reaction wheels, even in
the case of actuator fault. Thus, the objective of efcient,
low-cost, and reliable attitude control design can be met.
The rest of this paper is organized as follows. In Section II,
some necessary notations, denitions, and preliminary results
are rst introduced. The mathematical attitude model of a
rigid spacecraft and the control problem formulation are
summarized in Section III. The control solution is presented
in Section IV, and a numerical example of the derived control
design is presented in Section V. Conclusions and future works
are given in Section VI.
II. NOTATIONS, DEFINITIONS, AND PRELIMINARY
RESULTS
Let (respectively,
+
) denote the set of real (respectively,
positive real) numbers,
mn
denote the set of m by n real
matrices, and I
n

nn
an n n identity matrix. For a
matrix A
mn
, A
T
denotes its transpose, A
1
for any
left inverse if A has full column rank, and A
2
= A
T
A.
For any vector x(t ) = [ x
1
x
2
. . . x
n
]
T

n
, we dene
vectors |x|

= [ |x
1
|

|x
2
|

. . . |x
n
|

]
T

n
and
|x| = [ |x
1
| |x
2
| . . . |x
n
| ]
T

n
with |x
i
|

= sup
t 0
|x
i
(t )|,
i = 1, 2, . . . , n. We also dene ||x||
1
=

n
i=1
|x
i
|. The
symbol || || denotes the Euclidean norm or its induced norm.
The set of all the signals which are bounded on [0, ) is
represented by L

[0, ).
A. Denitions
Our main results are based on the following stability den-
itions for a given nonlinear system:
_
x(t ) = f (x, t ) + g(x, t )d
y = h(x, t )
(1)
where f (x, t ) :
n

+

n
are locally Lipschitz and
piecewise continuous in t , d
s
is an exogenous disturbance,
whereas y
m
is the system output. If there exists a closed
ball B

= { x
n
| ||x|| } such that for all x
0
B

, then
there exists an > 0 and a number T() satisfying ||x||
for all t t
0
+T(), it is said that the solution x(x
0
, t
0
, t ) is
uniformly ultimately bounded (UUB) [19].
The following vectors sgn(), ln(), and Tanh()
n
,
and diagonal matrices Cosh() and Sech()
nn
are
dened as:
sgn(x) = [sgn(x
1
) sgn(x
2
) . . . sgn(x
n
)]
T
(2)
ln(x) = [ln(x
1
) ln(x
2
) . . . ln(x
n
)]
T
(3)
Tanh(x) = [tanh(x
1
) tanh(x
2
) . . . tanh(x
n
)]
T
(4)
Cosh(x) = diag(cosh(x
1
), cosh(x
2
), . . . , cosh(x
n
)) (5)
Sech(x) = diag(sech(x
1
), sech(x
2
), . . . , sech(x
n
)) (6)
where x = [ x
1
x
2
, . . . , x
n
]
T

n
, sgn(), ln(), tanh(),
cosh(), and sech() are the sign, logarithmic, standard hyper-
bolic tangent, cosine, and secant functions, respectively.
B. Preliminary Results
The following results are useful for the proof of the main
results in sequel.
Lemma 1 [33]: For x = [ x
1
x
2
. . . x
n
]
T

n
, the
following expressions hold:
1
2
tanh
2
(x
i
) ln(cosh(x
i
)) (7)
||Sech
2
(x)|| 1 (8)
Tanh
T
(x)Tanh(x) x
T
Tanh(x). (9)
Lemma 2: For x = [ x
1
x
2
... x
n
]
T

n
, it follows that
x x
T
sgn(x).
Lemma 3 [34]: Let and be real-valued functions dened
on
+
, and let b and c be positive constants. If they satisfy
the differential inequality
c +b(t )
2
, (0) 0 (10)
and L

[0, ), then L

[0, ). Moreover, (t ) is
further bounded by
(t ) (0)e
ct
+
b
c
||||
2

. (11)
III. MODEL DESCRIPTION AND PROBLEM FORMULATION
A. Attitude Model of a Rigid Spacecraft
In this section, the mathematical model of a rigid spacecraft
attitude system, which is given by the attitude kinematics and
spacecraft dynamics, is briey presented.
1) Spacecraft Dynamics: When all the actuators are fault-
free, with the assumption of rigid body movement, the space-
craft dynamics can be found from Eulers moment equation
as [25]
J = S()J + +d (12)
where = [
1

2

3
]
T

3
denotes the angular velocity of
the spacecraft with respect to an inertial frame I and expressed
in the body frame B, J
33
is the total inertia of the
spacecraft, = [ u
1
u
2
u
3
]
T

3
is the control torque
input, d = [ d
1
d
2
d
3
]
T

3
is the external disturbance,
and S() =

is the skew symmetric vector cross-product


operator.
The actuator failures commonly encountered in spacecraft
attitude system can be termed as [35]: 1) partial loss of
effectiveness (F1); 2) lock-in-place (F2); and 3) oat (F3).
In this paper, the spacecraft considered does not have actuator
redundancy. If one of the actuators undergoes F2 or F3, then
it will lead to the loss of three-axis attitude control. These
two catastrophic faults are not the point of this investigation.
Consequently, this paper mainly discusses partial loss of
effectiveness fault F1. Consider actuator fault F1, represented
by a multiplicative matrix E(t ). The faulty attitude dynamic
model is given by
J = S()J + E(t ) +d (13)
where E(t ) = diag(e
11
(t ), e
22
(t), e
33
(t ))
33
with 0 <
e
0
e
ii
(t ) 1 being the actuator health indicator for the i th
actuator. The case e
i
(t ) = 1 implies that the i th actuator is
2254 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 21, NO. 6, NOVEMBER 2013
healthy, 0 < e
i
(t ) < 1 corresponding to the case in which the
i th actuator partially loses its actuating power, but still works
all the time.
2) Attitude Kinematics: To represent the attitude of the
spacecraft, the modied Rodriguez parameters (MRPs) rep-
resentation is adopted. Given a Euler rotation angle (t )
about the Euler principle axis n
3
, the attitude orientation
of the spacecraft in B with respect to I can be represented by
the MRP vector = [
1

2

3
]
T

3
, which is dened by
= n tan
_
(t )
4
_
, (t ) [0, 360)deg. (14)
The spacecraft attitude kinematics can be summarized as [19]
=
1
4
[(1
T
)I
3
+2S() +2
T
] = T(). (15)
Remark 1: It should be mentioned that, as a complete
revolution is performed (i.e., (t ) 360deg), this particular
MRPs set goes singular. As shown in [36], it is possible to
map the original MRPs vector to its corresponding shadow
counterpart
s
= /(
T
). By choosing to switch the MRPs
when
T
> 1, the MRPs vector remains bounded within a
unit sphere. Consequently, the global rotation representation
without singularity can be guaranteed. Therefore, in order
to avoid the singularity problem, MRPs and its shadow
counterpart
s
are utilized to represent attitude rotation in this
paper.
To facilitate the subsequent control formulation, we com-
bine the faulty dynamics (13) and attitude kinematics (15) to
form the following second-order nonlinear dynamic equation
using appropriate procedures and denitions:
J

() +C(, ) = G
T
()E(t ) +

d (16)
where the matrices G, J

, C
33
, and the lumped
disturbance

d are dened as
G() = T
1
(), J

= G
T
J G,

d = G
T
d
C(, ) = G
T
J

G + G
T
S(G )J G. (17)
Note from (15) and (17) that J

and C are bounded,


provided that their arguments are bounded. Several basic
properties of the faulty dynamic model in (16) are listed as
follows.
P1: The inertia matrix J

is symmetric and positive denite.


Moreover, it is bounded by
J
min
||x||
2
x
T
J

()x J
max
||x||
2
x
3
(18)
where J
min
and J
max
are positive constants.
P2 [17]: The matrix

J

() 2C(, ) satises the follow-


ing skew-symmetric relationship:
x
T
(

J

() 2C(, ))x = 0 x
3
(19)
where

J

is the time derivative of J

().
P3 [37]: The matrix C(, ) satises the bounded condition
C
min
|| ||
2
||C(, ) || C
max
|| ||
2

3
(20)
where C
min
and C
max
are some positive known constants.
B. Problem Statement
The control objective can be stated as follows. Consider
the rigid spacecraft attitude system given by (12) and (15),
determine a control law such that:
1) the closed-loop attitude system is globally stable in that
all the internal signal variables are bounded;
2) the attitude orientation converges to a small set contain-
ing the origin, that is, ||(t )||

for t T(

).
Furthermore, the above objective is to be met with angular
velocity eliminated (i.e., is not required) and in the presence
of: 1) uncertain inertia parameters and unknown bounded
external disturbances; 2) partial loss of actuator effectiveness
fault E(t ); and 3) actuator saturation |
i
| u
max
(i = 1, 2, 3),
where the positive constant u
max
is the maximum control
torque that each actuator can generate.
IV. MAIN RESULTS
A. Velocity Filter Design
Because direct or accurate measurement of angular velocity
may be unavailable, a passivity lter whose structure is moti-
vated by the Lyapunov-like stability analysis is rst introduced
to generate a velocity-related signal from attitude measurement
only. The lter is dened as the following rst-order dynamics
[33]:
p = l
1
p +l
1
l
2
(21)
where l
1
, l
2

+
are lter gains. The output of the lter is
dened as

f
= p l
2
. (22)
To provide some insight on the above-designed lter, we now
derive the dynamics of the lter output
f
. From (21), the
time derivative of (22) can be calculated as

f
= p l
2
= l
1

f
l
2
. (23)
B. Fault-Tolerant Attitude Stabilization Control Design
In this section, it is rst shown that when all the actuators
are fault-free (i.e., E(t ) I
3
for all t 0), the states
and can be guaranteed to be stable through a novel control
design with angular velocity eliminated. Then, based on
the nominal controller, an auxiliary controller is designed
and added to the nominal controller to compensate for the
considered actuator faults.
1) Nominal Control Law Design: The external disturbances
d(t ) incorporate gravitational perturbations, atmospheric drag,
solar radiation pressure forces, magnetic forces, and aero-
dynamic drags. Those disturbances are bounded in practical
aerospace engineering. Thus, one has d(t ) L

[0, ). It
thus follows that

d L

[0, ) by using its denition in (17).


Theorem 1: Consider the nominal attitude system governed
by (12) and (15). The nominal control law is implemented by
= (G()
T
)
1
_
K
p
Tanh() K
I
Tanh()
+K
d
Tanh(
f
) + K
W

f
_
(24)
XIAO et al.: ATTITUDE STABILIZATION OF SPACECRAFTS 2255
with
=
2

_
t
0
Tanh((s))ds (25)
where K
p
, K
I
, K
d
, and K
W

+
are control gains. Suppose
that the control parameters are appropriately chosen to satisfy
K
p
4

1

2
J
max
> 0 (26)
2l
1
l
2
1 > 0 (27)
3K
W
4l
1
l
2

1
4
> 0 (28)
3l
2
K
W
4l
1

J
max


_
3K
W
2l
1
_
2

3C
max


1
4
1
> 0 (29)
1

K
P

1
4
K
d

l
2
K
W
l
1

2

1
4
2

2
> 0 (30)
where
1
,
2
, and
+
are prescribed constants specied
by the designer. Then, the nominal controller (24) guarantees
that the attitude orientation is UUB.
Proof: Refer to the Appendix.
Remark 2: Similar to the results in [9] and [24], the control
law (24) is independent on the precise knowledge of inertia
matrix J (particularly time varying and uncertain, due to
onboard payload motion, vibration of exible appendages,
or fuel consumption). Although the implementation of the
controller needs the upper bounds on J and C, which are
used to determine the control gains in (24), those two bounds
can be approximately estimated or be chosen larger value
before launch. Thus, all the control gains can be determined.
Therefore, from the standpoint of uncertainties and external
disturbances rejection, the derived control law has great sta-
bility robustness.
2) Fault-Tolerant Control Law Design: So far, all the
actuators are assumed to be healthy. This is not a real-
istic assumption in practice. Due to aging of components,
in general, actuator faults occur, especially partial loss of
effectiveness fault. This type of fault may deteriorate attitude
control performance and even result in system instability.
Taking partial loss of effectiveness fault into consideration,
the following fault-tolerant control law is proposed to perform
attitude stabilization maneuver:
=
N
+
F
(31)
where
N
is the nominal control (24) for the normal system,
and
F
is the fault-tolerant control part designed and added to
compensate for possible actuator faults effect on the system,
and it is given by

F
= sgn
__
+
1

Tanh()
_
T
G
T
_
T
(1 e
0
)
e
0
||
N
||,
> 1. (32)
The stability analysis of the closed-loop system under the
effect of the fault-tolerant control (31) can be stated in the
following Theorem.
Theorem 2: Consider the faulty attitude system governed
by (13) and (15) under partial loss of actuator effectiveness
fault. With the application of the control law (31), suppose
that the design parameters are chosen to hold (26)(30). Then,
the attitude orientation is UUB if the nominal control (24)
guarantees the nominal attitude system UUB.
Proof: When partial loss of effectiveness fault E(t )
occurs, it yields from (16), (A7), and (31) that
J

() +C(, ) = K
p
Tanh() K
I
Tanh()
+K
W

f
+

d + K
d
Tanh(
f
)
G
T
(I
3
E(t )) + G
T

F
. (33)
Then, it has

T
J

=
T
C(, ) + K
p

T
Tanh() K
I

T
Tanh() + K
d

T
Tanh(
f
)
+K
W

T

f
+
T

d +
T
G
T

F

T
G
T
(I
3
E(t )) (34)
1

Tanh()
T
J

=
1

Tanh()
T
[C + K
p
Tanh() +

d
K
I
Tanh() + K
d
Tanh(
f
) + K
W

f
]

Tanh()
T
G
T
(I
3
E(t ))
+
1

Tanh()
T
G
T

F
. (35)
With the same Lyapunov function candidate dened in
Theorem 1, substituting (34) and (35) into (A6) and following
the same lines as in (A12)(A16) result in

V (
1
+
2
)||

d||
2
m
1
|| ||
2
m
2
||
f
||
2
m
4
||Tanh(
f
)||
2
+(
T
+
1

Tanh()
T
)G
T

F
(
T
+
1

Tanh()
T
)(I
3
E(t )) m
3
||Tanh()||
2
.
(36)
Due to the inequalities (26)(30), it is obtained from (31) and
(32) and Lemma 2 that

V
_

1
+
2
_
||

d||
2
+(
T
+
1

Tanh()
T
)E(t )
F

_

T
+
1

Tanh()
T
__
I
3
E(t )
_

N
m
3
||Tanh()||
2
(
1
+
2
)||

d||
2

(
T
+
1

Tanh()
T
)G
T

1
(1e
0
)
e
0
||
N
||
+

(
T
+
1

Tanh()
T
)G
T

1
(1e
0
)||
N
||m
3
||Tanh()||
2
= (
1
+
2
)||

d||
2
m
3
||Tanh()||
2

(
T
+
1

Tanh()
T
)G
T

1
(1 e
0
)(1 )||
N
||
(
1
+
2
)||

d||
2
m
3
||Tanh()||
2
. (37)
With (37), the result is established using the same argument
as in the proof of Theorem 1. This completes the proof.
Remark 3: The controller is implemented with digital com-
puter in practical aerospace engineering. The value of
f
in
the time of (k + 1)T can be approximately estimated by
2256 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 21, NO. 6, NOVEMBER 2013
using the one-step previous information from attitude sensors
as

f
(kT) =

f
((k +1)T)
f
(kT)
T
(38)
where T is the control update period. By using (22), we can
calculate the value of
f
by

f
(kT) =
p((k +1)T) p(kT)
T
l
2
((k +1)T) (kT)
T
. (39)
Accordingly, it is obtained from (23) that
=
1
l
2

f

l
1
l
2

f
. (40)
Hence, can be numerically derived during the implementa-
tion of the control law (31), because
f
is obtained with (39),
and
f
is given by the attitude sensors. It can be summarized
that (32) is independent on angular velocity measurements,
although is involved. On the other hand,
N
is also angular
velocity-free. As a result, the proposed fault-tolerant control
(31) can be implemented without the need of any rate sensor
to measure angular velocity.
3) Analysis of the Upper Bound of the Control Effort: From
(21), we nd that the state of the velocity lter is bounded
by
p l
1
p +l
1
l
2
||, p(0) 0. (41)
According to the analysis in the proof of Theorem 2, we have
|||| L

[0, ). All conditions in Lemma 3 are satised.


Hence, the inequality (41) ensures that
p(t ) p(0)e
l
1
t
+l
2
||
2

. (42)
The inequality |
i
| 1, (i = 1, 2, 3) holds from Remark
1. Thus, by choosing p(0) = 0, one can evaluate p(t ) l
2
,
which leads to || p(t )||

3l
2
.
With the result in [36], direct calculation shows that the
matrix T() in (15) is such that
T
T
()T() =
_
1 +
T

4
_
2
I
3
. (43)
Thus
||(G()
T
)
1
|| = ||T()|| =
1 +
T

4

1
2
. (44)
On the basis of the preceding analysis, the upper bound of
the proposed fault-tolerant control law (31) is summarized in
the following Theorem.
Theorem 3: Consider the developed fault-tolerant control
law (31). Chose the control gains such that (26)(30), and
also satisfying the following inequality:
1
2
_
1 +
(1 e
0
)
e
0
__

3(K
p
+ K
I
+ K
d
) +(

3 +1)K
W
l
2
_
u
max
. (45)
The control output of each actuator is then rigorously bounded
by the actuator saturation value.
Proof: Together with (44), it leaves the nominal control
law (24) as
||||
1
2
[

3(K
p
+ K
I
+ K
d
) + K
W
||
f
||]. (46)
Next, from (22),
f
is bounded by
||
f
|| || p|| +l
2
|||| (

3 +1)l
2
. (47)
It follows that:
|
i
| ||
N
|| +|
Fi
| ||
N
|| +
(1 e
0
)
e
0
||
N
||

1
2
_
1 +
(1 e
0
)
e
0
__

3(K
p
+ K
I
+ K
d
)
+(

3 +1)K
W
l
2
_
(48)
for i = 1, 2, 3, and
Fi
is the i th argument of
F
. Thus, with
(45) and (48), it can be demonstrated that
|
i
| u
max
. (49)
Hence, actuator constraints can be satised.
From Theorem 2, it is known that the smaller e
0
in (32)
is selected, more severe actuator faults the controller (31)
can tolerate, and much more fault-tolerant capability results.
In Theorem 3, it is proved that once the value of e
0
is
determined (no matter how small the value of e
0
> 0 is),
the control with the control gains chosen from (26)(30)
and (45) can always guarantee that the closed-loop system
is UUB. It is further theoretically analyzed that the attitude
orientation is guaranteed to be bounded by ||(t )||

for
t T(

), and the control effort is such that |


i
| u
max
,
i = 1, 2, 3. Therefore, when implementing the fault-tolerant
controller (32), the selection of the value e
0
depends on the
tolerable level of actuator faults, which is imposed on the
spacecraft by the designers. For example, if the spacecraft
needs to tolerate 90% loss of control, then one should choose
0 < e
0
< 0.1. Moreover, the following two results should be
pointed out.
4) Small Value of e
0
Would Not Lead to Weak Control
Power: A conservative selection of e
0
may indeed lead to
small control gains. Then, it may follow weak nominal
control effort
N
. However, as shown in (32), small e
0
will
result in large fault-tolerant control effort
F
. This is due to
the term (1e
0
)/e
0
in
F
. As a result, the total control power
would not be weak, and the control objective can still be
achieved. For instance, assume that
N
= [ 1 1 1] 10
3
Nm
at t = 50 s, when e
0
= 0.01 and = 1.001 are chosen. One will
have
F
= [ 0.1683 0.1683 0.1683 ] Nm, if the vector sgn() in
(32) is equal to [ 1 1 1 ]. This leads to the total control
effort= [ 0.1673 0.1673 0.1673 ] Nm. This torque is not a
weak control power in practical aerospace engineering.
5) Unacceptably Long Time Would Not be Taken to Stabilize
the Attitude: Because the conservative selection of e
0
may
not lead to weak control, as shown in the above analysis, the
closed-loop system would be stabilized within a sufciently
reasonable amount of time. This is veried by using a numer-
ical example, as presented in Section V. On the other hand, as
shown in the proof of Theorem 2, the time T(

), within which
the attitude is governed to be ||(t )||

, is dependent on
XIAO et al.: ATTITUDE STABILIZATION OF SPACECRAFTS 2257
the attitude pointing accuracy

. The higher pointing accuracy


is, that is,

is smaller, much more time T(

) will be spent
to ensure the stabilization of the attitude system. Further, as
shown in (A19), one has

= D
0

(
1
+
2
)/m
3
. When
small control gains result in small value of m
3
, choosing small

1
and
2
can lead to small

; whereas the selection of large

1
and
2
will introduce large

. As a sequence, the attitude


pointing accuracy

is adjustable, and the time T(

) is then
selectable for the designer. Consequently, the proposed control
methodology will be practical and of real interest and use in
real-life scenarios.
Remark 4: A practical problem, namely, the chattering effect
that may be induced by the use of the discontinuity function
sgn() in (32), should be considered since it is impossible
to switch the control at innite time. This problem is prac-
tically undesirable because it may excite the neglected high-
frequency dynamics. One approach to reduce the chattering
is to approximate the function sgn() by using a continuous
function. Following this idea, saturation function is used to
approximate sgn(), (32) is thus modied as [38]

F
= sat
_
_
+
1

Tanh()
_
T
G
T
_
T
(1e
0
)
e
0
||
N
||, > 1
(50)
where we have slightly abused the notation by using sat()
to stand for both scalar-valued and vector-valued saturation
functions. For x = [ x
1
x
2
. . . x
n
]
T

n
, the i th argument
of sat() is dened by sat(x)
i
= min{x
i
/, sgn(x
i
)}, i =
1, 2, . . . , n, where is a small positive constant.
Remark 5: The denition of saturation function sat() leads
to |sat(x)
i
| 1, i = 1, 2, . . . , n in [39][43]. Hence, the
modied fault-tolerant control power
F
given in (50) is still
guaranteed to be such that |
Fi
| (1 e
0
)||
N
||/e
0
. With
the control parameters chosen in (45), Theorem 3 is still
valid, and thus the actuator upper bound condition is still
ensured.
Remark 6: Summarizing the analysis in Theorem 1 and
Theorem 3, the lter and the control parameters can be chosen
according to the following procedures.
Step 1: Choose e
0
> 0 as small as possible. The smaller
e
0
is, much more severe partial loss of actuator
effectiveness fault is tolerated.
Step 2: Choose > 1 as small as possible, and calculate
the value of

3(K
p
+K
I
+K
d
) +(

3 +1)K
W
l
2
by using (45).
Step 3: Estimate the value of J
max
and C
max
, choose large
and small
i
, i = 1, 2.
Step 4: Choose large , and l
1
/l
2
such that (27) is satised.
Step 5: Choose the control gain K
W
such that (28) and (29)
are satised.
Step 6: Choose the control gains K
p
and K
d
such that (26)
and (30) is satised.
Step 7: Choose the control gain K
I
such that (45) is satis-
ed.
Step 8: Check whether the chosen parameters satisfy the
value of in Step 2 or not. If the value of is not
met, then repeat Step 3 to Step 7.
TABLE I
MAIN PARAMETERS OF AN ON-ORBIT RIGID SPACECRAFT
Mass (kg) 75.6
Inertia moments (kgm
2
):
Principal moments of inertia
Products of inertia
J
1
= 20, J
2
= 20, J
3
= 30
Can be neglected
Orbit:
Type
Attitude (km)
Inclination (degree)
Right ascension of the ascending
node
Circular
672
97.4
10.30 am
Attitude control type
Three-axis control by three
reaction wheels
Maximum torque (Nm) of the
reaction wheel
u
max
= 0.6
TABLE II
CONTROLLER PARAMETERS CHOSEN FOR NUMERICAL ANALYSIS
Control Schemes Control Gains
VFFTC
K
P
= 0.15, K
I
= 0.01,
K
d
= 0.1,
K
W
= 0.005, l
1
= 0.5, l
2
= 30,
= 1.001, = 100
UQOFC

1
= 0.15,
2
= 0.35,

1
= diag(0.25, 0.25, 0.25)
V. NUMERICAL EXAMPLES
To demonstrate the effectiveness of the proposed fault-
tolerant control scheme, a rigid spacecraft with the orbital
parameters shown in Table I is numerically simulated using
the set of governing equations of motion (12) and (15) in
conjunction with the control law (31). A time-varying moment
inertia matrix as discussed in [24] is incorporated into the
model
J = 1 +e
0.1t
+2u(t 10) 4u(t 20)diag(3, 2, 1)
(51)
where u( ) is dened as u(t 0) = 1 and u(t < 0) = 0.
For the purpose of comparison, the proposed fault-tolerant
angular velocity-free control law (31) (it is called as
VFFTC here) and the unit quaternion output feedback control
(UQOFC) developed in [25] are carried out. Both controllers
are compared for a fault and external disturbance-free case
rst, and then for a faulty case in the presence of disturbances.
The control parameters were chosen according to Remark 10
by trial-and-error until a good stabilization performance was
obtained. The control gains are listed in Table II.
At time t = 0, the attitude orientation is such that (0) =
[ 0.3202 0.4850 0.3899 ]
T
with a zero initial body angular
velocity. For the consideration of engineering application,
attitude sensor noises (
2
p
): 0.0001(1) modeled as zero-mean
Gaussian random variable with variance
2
p
was considered in
the attitude measurement equations [8]

m,i
(t ) =
i
(t ) + N(0,
p
), i = 1, 2, 3 (52)
where
m,i
(t ) is the measured attitude from the attitude sensor,

i
(t ) is the real attitude orientation, and N(0,
p
) denotes a
zero-mean Gaussian white noise with variance
p
.
2258 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 21, NO. 6, NOVEMBER 2013
0 100 200
-0.2
0
0.2
0.4
0.6
0.8
t/sec

1
0 100 200
-0.8
-0.6
-0.4
-0.2
0
0.2

2
0 100 200
-0.2
0
0.2
0.4
0.6
0.8
t/sec

3
(a)
2900 2950 3000
-2
-1
0
1
2
3
x 10
-6
t/sec

1
2900 2950 3000
-3
-2
-1
0
1
2
x 10
-6
t/sec

2
2900 2950 3000
-2
-1
0
1
2
3
x 10
-6
t/sec

3
(b)
t/sec
Fig. 1. Spacecraft attitude orientation with VFFTC (solid line) and UQOFC
(dashed line) in the absence of faults and disturbances. (a) Initial response.
(b) Steady-state behavior.
0 100 200
-0.1
0
0.1
t/sec

1
(
r
a
d
/
s
)
0 100 200
-0.1
0
0.1
0.2
t/sec

2
r
a
d
/
s
)
0 100 200
-0.1
0
0.1
t/sec

3
(
r
a
d
/
s
)
(a)
2900 2950 3000
-4
-2
0
2
4
x 10
-6
t/sec

1
(
r
a
d
/
s
)
2900 2950 3000
-6
-4
-2
0
2
4
x 10
-6
t/sec

2
(
r
a
d
/
s
)
2900 2950 3000
-3
-2
-1
0
1
2
3
x 10
-6
t/sec

3
(
r
a
d
/
s
)
(b)
Fig. 2. Spacecraft angular velocity with VFFTC (solid line) and UQOFC
(dashed line) in the absence of faults and disturbances. (a) Initial response.
(b) Steady-state behavior.
A. Response With Fault-Free and Disturbance-Free Case
In this case, a relative ideal situation is simulated in which
not only no actuator fault occurs but also there is not any
0 100 200
-0.6
-0.3
0
0.3
0.6
t/sec

1
(
N
m
)
0 100 200
-0.6
-0.3
0
0.3
0.6
t/sec

2
(
N
m
)
0 100 200
-0.6
-0.3
0
0.3
0.6
t/sec

3
(
N
m
)
Saturation Saturation
Saturation
(a)
2900 2920 2940 2960 2980 3000
-2
0
2
x 10
-4

1
(
N
m
)
2900 2920 2940 2960 2980 3000
-2
0
2
x 10
-4

2
(
N
m
)
2900 2920 2940 2960 2980 3000
-2
0
2
x 10
-4
t/sec

3
(
N
m
)
(b)
Fig. 3. Time response of (t ) with VFFTC (solid line) and UQOFC (dashed
line) in the absence of faults and disturbances. (a) Initial response. (b) Steady-
state behavior.
external disturbance acting on the spacecraft. We rst present
the simulation results when applying VFFTC. It is shown
in Figs. 1 and 2 (solid line) that the velocity-free controller
managed to perform attitude stabilization maneuver with good
control performance. The attitude orientation is stabilized
within 100 s with high pointing accuracy even in the presence
of uncertain inertia parameters (51). Moreover, the results as
illustrated in Fig. 1 (solid line) further verify the conclusion
that the proposed control law (31) can achieve the attitude
control in the absence of actuator faults.
Although the application of UQOFC to the spacecraft can
achieve almost the same attitude control accuracy and stability
as VFFTC, as shown in Figs. 1 and 2 (dashed line), it requires
180 s to force the spacecraft attitude to a satised resolution.
Much more overshoot resulted with UQOFC than that by using
VFFTC. This is due to the fact that the proposed control
scheme can decrease the overshoot by tuning the proportional,
integral, and derivative gains K
P
, K
I
, and K
d
. It is also
interested to note that both controllers can protect the control
torque from actuator saturation magnitude, as we can see in
Fig. 3 (solid and dashed lines). However, compared with the
control power in Fig. 3, larger control effort of VFFTC is
observed than UQOFC. Indeed, this is due to the fact that
f
is always activated whether actuator fault occurs or not.
XIAO et al.: ATTITUDE STABILIZATION OF SPACECRAFTS 2259
0 300 600
0
0.2
0.4
0.6
t/sec

1
0 300 600
-0.8
-0.6
-0.4
-0.2
0
t/sec

2
0 300 600
-0.1
0
0.3
0.7
1
t/sec

3
(a)
2900 2950 3000
-10
-8
-6
-4
-2
0
2
x 10
-3
t/sec

2
2900 2950 3000
-2
0
2
4
6
8
10
x 10
-3
t/sec

3
2900 2950 3000
-5
0
5
10
15
20
x 10
-3
t/sec

1
(b)
Fig. 4. Spacecraft attitude orientation with VFFTC (solid line) and UQOFC
(dashed line) in the presence of faults and disturbances. (a) Initial response.
(b) Steady-state behavior.
0 300 600
-0.1
0
0.1
t/sec

1
(
r
a
d
/
s
)
0 300 600
-0.1
0
0.1
t/sec

2
r
a
d
/
s
)
0 300 600
-0.1
0
t/sec

3
(
r
a
d
/
s
)
(a)
2900 2950 3000
-4
-2
0
2
4
x 10
-6
t/sec

1
(
r
a
d
/
s
)
2900 2950 3000
-5
0
5
x 10
-5
t/sec

2
(
r
a
d
/
s
)
2900 2950 3000
-5
0
5
x 10
-5
t/sec

3
(
r
a
d
/
s
)
(b)
Fig. 5. Spacecraft angular velocity with VFFTC (solid line) and UQOFC
(dashed line) in the presence of faults and disturbances. (a) Initial response.
(b) Steady-state behavior.
B. Response With Actuator Fault and Disturbances Case
On-orbit spacecraft is inevitably under the effect of external
disturbances. Hence, a large constant external disturbance
0 300 600
-0.6
-0.3
0
0.3
0.6
t/sec

1
(
N
m
)
0 300 600
-0.6
-0.3
0
0.3
0.6
t/sec

2
(
N
m
)
0 300 600
-0.6
-0.3
0
0.3
0.6
t/sec

3
(
N
m
)
Saturation
Saturation
Saturation
(a)
2900 2920 2940 2960 2980 3000
-0.02
-0.0198
-0.0196

1
(
N
m
)
2900 2920 2940 2960 2980 3000
-0.02
0
0.02
0.04

2
(
N
m
)
2900 2920 2940 2960 2980 3000
-0.02
-0.01
0
0.01
t/sec

3
(
N
m
)
(b)
Fig. 6. Time response of (t ) with VFFTC (solid line) and UQOFC (dashed
line) in the presence of faults and disturbances. (a) Initial response. (b) Steady-
state behavior.
torque specied by d(t ) = [ 0.02 0.01 0.01 ]
T
is imposed in
this section. Moreover, the following partial loss of actuator
effectiveness fault is considered.
1) Actuator Fault Scenario: The fault scenarios occur under
these situations: 1) the reaction wheel mounted in line with
the roll axes decreases 50% of its normal value after 5 s;
2) the actuator mounted in line with the pitch axes loses its
power of 40% in 10 s; and 3) the reaction wheel mounted in
line with the yaw axes undergoes 50% loss of effectiveness
in 15 s. These are fairly severe faults that, if not compensated
for, will cause overall attitude system instability, which will
be discussed later.
When the VFFTC is implemented to the attitude system,
the time responses of attitude orientation and angular velocity
are presented in Fig. 4 (solid line) and Fig. 5 (solid line),
respectively. The driving torque is shown in Fig. 6 (solid line).
As expected, we clearly see that the proposed fault-tolerant
control scheme managed to compensate for the partial loss of
effectiveness fault. High attitude pointing accuracy and attitude
stability are still guaranteed without angular velocity measure-
ments. As shown in Fig. 6 (solid line), the control torque of
each reaction wheel is still within its maximum allowable limit
even in the presence of external disturbances and uncertain
inertia parameters. This is achieved by introducing the fault-
tolerant part
f
in (32). Due to the limited control power of
each reaction wheel, it demands longer time to stabilize the
attitude in the presence of actuator fault. As we can see in
2260 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 21, NO. 6, NOVEMBER 2013
TABLE III
PERFORMANCE SUMMARY UNDER DIFFERENT CONTROL SCHEMES
Control
Actuator
Status
Control Schemes
Performance Healthy VFFTC (31) UQOFC [25]
Attitude
Roll
Normal 2.0 10
6
3.0 10
6
Control
Fault 2.0 10
6
0.2
Accuracy
Pitch
Normal 3.0 10
6
1.0 10
6
Fault 4.0 10
5
0.008
Yaw
Normal 2.0 10
6
2.0 10
6
Fault 6.0 10
5
0.008
Slew
Roll
Normal 4.0 10
6
1.5 10
6
Rate
Fault 3.0 10
6
4.0 10
6
Accuracy
Pitch
Normal 4.0 10
6
2.0 10
6
(rad/s)
Fault 2.5 10
6
7.0 10
5
Yaw
Normal 2.5 10
6
2.0 10
6
Fault 1.0 10
6
8.0 10
5
Attitude
Stabilization
Normal 100 180
time (s) Fault 500 Innity
Fig. 4(a) (solid line), the whole attitude stabilization maneuver
is performed nearly 500 s.
The application of UQOFC leads to the attitude orientation
and angular velocity shown in Figs. 4 and 5 (dashed line). As
pointed out, that this control law can stabilize the attitude with
angular velocity eliminated only in the absence of external
disturbance and actuator fault. Therefore, when disturbance
d(t ) and actuator fault are introduced in the system, UQOFC
failed to achieve attitude stabilization control as we can see
in Fig. 4 (dashed line), although a relative higher slew rates
accuracy and limited control effort were still observed in
Figs. 5 and 6 (dashed line), respectively.
C. Summary of Simulation Results
On the basis of the above illustrated simulation results, the
steady attitude stability and control accuracy under VFFTC
and UQOFC are summarized in Table III.
1) In the absence of actuator fault and external disturbance,
both the proposed methodology and the scheme devel-
oped in [25] can accomplish the attitude stabilization
maneuver without the measurements of angular velocity.
Almost the same attitude pointing accuracy and attitude
stability are achieved. However, the developed controller
demands less time to govern the attitude, the comparison
with [25] shows that our solution provides a faster
response.
2) When actuator fault and external disturbances are con-
sidered, our presented control law can successfully
perform the attitude stabilization maneuver. While the
controller [25] failed to stabilize the attitude, although a
high slew rates accuracy is achieved.
3) When the developed strategy is implemented to the
spacecraft, comparison with the results of the actuator
normal case and the fault case shows that the slew rates
accuracy is met with the same order of magnitude, and
only the attitude control precision decreases one order
of magnitude in the presence of actuator fault.
VI. CONCLUSION
Although many nonlinear attitude control approaches were
available for spacecraft attitude control design in literature,
none of them have addressed fault tolerance, actuator satu-
ration, velocity-free control, and robustness, simultaneously.
In this paper, we developed a novel angular velocity-free
control scheme to achieve attitude stabilization maneuver in
the presence of partial loss of effectiveness fault, uncertainties
in the inertia parameters, external disturbances, and actuator
saturation. The control approach guaranteed the closed-loop
attitude system to be uniformly ultimately bounded stable.
Angular velocity sensors were not needed to implement the
control law, also magnetic torquers were not required to be
mounted to dump the excess momentum of reaction wheels.
The objective of developing a low-cost attitude control system
for spacecraft was realized. It thus let the developed con-
trol scheme be cost-effective for microsatellite applications.
However, the drawback of the scheme remains its dependence
on the lower bound e
0
of the actuator fault. As some of
future works, extension of the approach with elimination of
the requirement of such lower bound should be carried out.
The lock-in-place and oat faults also need to be addressed.
Further, extension to the fault-tolerant attitude control with
nite-time convergence should be investigated to perform time
critical aerospace mission, in which the attitude is stabilized
in nite time.
APPENDIX
PROOF OF THEOREM 1
The proof uses the elements of Lyapunov stability theory
and is organized as follows: we rst design a candidate
Lyapunov function, which is globally positive and radically
unbounded in the states; then we prove that the time derivative
of this Lyapunov candidate is negative denite along trajecto-
ries generated by (12) and (15). Finally, Barbalats lemma is
invoked to show that the closed-loop attitude system is stable.
A. Lyapunov Function Candidate
Consider the Lyapunov candidate function of the form
V(t ) =
1
2

T
J

Tanh()
T
J

+
K
d
l
2
3

i=1
ln[cosh(
f
i
)]
+K
p
3

i=1
ln[cosh(
i
)] +
1

2
Tanh()
_
0
s
T
K
I
Cosh
2
()ds +
K
W
2l
2

T
f

f
(A1)
where
Tanh()
_
0
s
T
K
I
Cosh
2
()ds =
3

i=1
tanh(
i
)
_
0
K
I
cosh
2
(
i
)s
i
ds
i
. (A2)
XIAO et al.: ATTITUDE STABILIZATION OF SPACECRAFTS 2261
Because K
I
and cosh
2
(
i
) are positive, it follows that:
Tanh()
_
0
s
T
K
I
Cosh
2
()ds > 0 = 0
3
. (A3)
Before proceeding with the time derivative computation of
the Lyapunov function, its positive deniteness is shown rst.
Using (7) of Lemma 1 and P1, one can get from (26) that
1
4

T
J

Tanh()
T
J

+
K
p
2
3

i=1
ln
_
cosh(
i
)
_
=
1
4
_

2

Tanh()
_
T
J

_

2

Tanh()
_

2
Tanh()
T
J

Tanh() +
K
p
2
3

i=1
ln
_
cosh(
i
)
_

K
p
2
3

i=1
ln
_
cosh(
i
)
_

2
Tanh()
T
J

Tanh()

i=1
_
K
p
4

J
max

2
_
tanh
2
(
i
) > 0. (A4)
In view of (A3) and (A4), rearranging (A1) yields
V
1
4

T
J

+
K
d
l
2
3

i=1
ln
_
cosh(
f
i
)
_
+
K
W
2l
2

T
f

f
+
K
p
2
3

i=1
ln
_
cosh(
i
)
_
+
1

2
_
Tanh()
0
s
T
K
I
Cosh
2
()ds

1
4

T
J

+
K
d
2l
2
3

i=1
tanh
2
(
f
i
)+
K
p
4
3

i=1
tanh
2
(
i
)
+
K
W
2l
2

T
f

f
+
1

2
_
Tanh()
0
s
T
K
I
Cosh
2
()ds >0 (A5)
for [
T

T

T
f
]
T
= 0. Hence, it can be concluded that
the selected Lyapunov function candidate V is continuously
differentiable, radially unbounded, and positive denite in the
states , , , and
f
.
B. Stability Analysis
The time derivative of V can be calculated as

V =
1
2

T

J

+
T
J

Tanh()
T

J

Tanh()
T
J

(Sech
2
() )
T
J


+
K
d
l
2

T
f
Tanh(
f
) + K
p

T
Tanh() +
K
W
l
2

T
f

f
+
K
I

2
Tanh()
T
Cosh
2
()
d[Tanh()]
dt
. (A6)
Substituting (24) into (16), one can use E(t ) I
3
for actuator
fault-free case to establish
J

() +C(, ) = K
p
Tanh() K
I
Tanh()
+K
d
Tanh(
f
) + K
W

f
+

d. (A7)
Through laborious yet relatively straightforward algebra
followed by the application of (23), and (25), (A7) yields

T
J

=
T
C(, ) + K
p

T
Tanh() K
I

T
Tanh()
+ K
d

T
Tanh(
f
) + K
W

T

f
+
T

d (A8)
Tanh()
T
Cosh
2
()
d[Tanh()]
dt
=
2
Tanh()
T
Tanh()
T
Tanh() (A9)

T
f
Tanh(
f
) = (l
1

T
f
+l
2
)Tanh(
f
). (A10)
With the inequality
Tanh()
T
Tanh(
f
)
1
4
||Tanh()||
2
+||Tanh(
f
)||
2
(A11)
imposing the bound ||Tanh()||

3, and using P1P3, the
time derivative of V in (A6) is simplied as

V =
1

_
Tanh()
T
C(, ) +
_
Sech
2
()
_
T
J

Tanh()
T

d
K
d

Tanh()
T
Tanh(
f
)

K
p

Tanh()
T
Tanh()+
T

d+
_
K
W

T
+
K
W
l
2

T
f
_

K
W

Tanh()
T

f

K
d
l
1
l
2

T
f
Tanh(
f
)

3C
max
+ J
max
_
|| ||
2

_
K
p


K
d
4
_
||Tanh()||
2

Tanh()
T
d
1

_
2l
1
K
d
l
2
K
d
_
||Tanh(
f
)||
2

K
W

Tanh()
T

f
+
_
K
W

T
+
K
W
l
2

T
f
_

f
+
T

d.
(A12)
In particular, using the Youngs inequality, it follows:

T
f

1
4
||
f
||
2
+|| ||
2
(A13)

T

d
1
4
1

T
+
1

d
T

d
1

Tanh()
T

d (A14)

1
4
2

2
Tanh()
T
Tanh() +
2

d
T

d. (A15)
Hence, for the last three items on the right-hand side of (A12),
it can be found that

K
W

Tanh()
T

f
+(K
W

T
+
K
W
l
2

T
f
)
f

l
1
K
W
4l
2
||
f
||
2
+
l
2
K
W
l
1

2
||Tanh()||
2
+ K
W

T

K
W
l
2
(l
1

f
+l
2
)
T

f
=
3l
1
K
W
4l
2

T
f

f
+
l
2
K
W
l
1

2
||Tanh()||
2
=
3K
W
4l
1
l
2
||
f
||
2

3l
2
K
W
4l
1
|| ||
2

3K
W
2l
1

T
f

+
l
2
K
W
l
1

2
||Tanh()||
2
2262 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 21, NO. 6, NOVEMBER 2013

_
3l
2
K
W
4l
1

_
3K
W
2l
1
_
2

_
|| ||
2

_
3K
W
4l
1
l
2

1
4
_
||
f
||
2
+
l
2
K
W
l
1

2
||Tanh()||
2
. (A16)
By using (26)(30) and (A14)(A16), the total derivative of
V along the closed-loop trajectories of (A12) can be further
established as

V
1

3C
max
+ J
max
)|| ||
2

_
K
p


K
d
4
_
||Tanh()||
2
+
l
2
K
W
l
1

2
||Tanh()||
2

1
2
_
2l
1
K
d
l
2
K
d
_
||Tanh(
f
)||
2

_
3l
2
K
W
4l
1

_
3K
W
2l
1
_
2

_
|| ||
2
+
1
4
2

2
Tanh()
T
Tanh()

_
3K
W
4l
1
l
2

1
4
_
||
f
|| +
1
4
1

T
+
1

d
T

d +
2

d
T

d
=
_

1
+
2
_
||

d||
2

_
K
p


K
d
4

l
2
K
W
l
1

2

1
4
2

2
_
. , .
m
3
||Tanh()||
2

_
2l
1
K
d
l
2
K
d
_
. , .
m
4
||Tanh(
f
)||
2

_
3K
W
4l
1
l
2

1
4
_
. , .
m
2
||
f
||
2

_
3l
2
K
W
4l
1

J
max


_
3K
W
2l
1
_
2

3C
max


1
4
1
_
. , .
m
1
|| ||
2
(
1
+
2
)||

d||
2
m
3
||Tanh()||
2
. (A17)
Due to

d L

[0, ), then there exists a D


0
> 0 such that
||

d|| D
0
. From inequality (A17),

V is bounded as

V (
1
+
2
)D
2
0
m
3
||Tanh()||
2
. (A18)
It is seen from (A18) that

V < 0 when are outside of the
set
D

||Tanh()||

_
(A19)
where

= D
0

(
1
+
2
)/m
3
. Equation (A19) implies that
V(t ) decreases monotonically outside the set D. Hence, all
the signal in the closed-loop system are bounded. Moreover,
one can choose small enough
0
to guarantee that
lim
t
|||| lim
t
||Tanh()|| D. (A20)
It can be concluded from (A20) that, there exists
a T(

) > 0 such that ||(t )||

for t T(

). This
shows that the attitude is UUB from Denition 2. Thereby,
the proof is completed.
ACKNOWLEDGMENT
The authors would like to thank the associate editor and
reviewers for their very constructive comments and sugges-
tions which have greatly helped to improve the quality and
presentation of the manuscript of this paper.
REFERENCES
[1] B. Robertson and E. Stoneking, Satellite GN&C anomaly trends, in
Proc. 26th Annu. AAS Rocky Mountain Guid. Control Conf., San Diego,
CA, 2003, pp. 115.
[2] B. A. Roberts, J. W. Kruk, T. B. Ake, T. S. Englar, B. F. Class, and D.
M. Rovner, Three-axis attitude control with two reaction wheels and
magnetic torquer bars, in Proc. AIAA Guid., Navigat., Control Conf.
Exhibit, Providence, RI, 2004, pp. 18.
[3] A. H. J. D. Ruiter, A fault-tolerant magnetic spin stabilizing con-
troller for the JC2Sat-FF mission, Acta Astronaut., vol. 68, nos. 12,
pp. 160171, 2011.
[4] Y. M. Zhang and J. Jiang, Bibliographical review on recongurable
fault-tolerant control systems, Annu. Rev. Control, vol. 32, pp. 229252,
Dec. 2008.
[5] B. Xiao, Q. L. Hu, and Y. M. Zhang, Adaptive sliding mode fault
tolerant attitude tracking control for exible spacecraft under actua-
tor saturation, IEEE Trans. Control Syst. Technol., vol. 20, no. 6,
pp. 16051612, Nov. 2012.
[6] S. Tafazoli and K. Khorasani, Nonlinear control and stability analysis
of spacecraft attitude recovery, IEEE Trans. Aerosp. Electron. Syst.,
vol. 42, no. 3, pp. 825845, Jul. 2006.
[7] Y. W. Liang, S. D. Xu, and C. L. Tsai, Study of VSC reliable designs
with application to spacecraft attitude stabilization, IEEE Trans. Control
Syst. Technol., vol. 15, no. 2, pp. 332338, Mar. 2007.
[8] J. Jin, S. Ko, and C. K. Ryoo, Fault tolerant control for satellites with
four reaction wheels, Control Eng. Pract., vol. 16, pp. 12501258,
Oct. 2008.
[9] W. C. Cai, X. H. Liao, and Y. D. Song, Indirect robust adaptive fault-
tolerant control for attitude tracking of spacecraft, J. Guid. Control
Dynamics, vol. 31, pp. 14561463, Sep.Oct. 2008.
[10] S. Varma and K. D. Kumar, Fault tolerant satellite attitude control using
solar radiation pressure based on nonlinear adaptive sliding mode, Acta
Astronaut., vol. 66, pp. 486500, Feb.Mar. 2009.
[11] Y. Jiang and Q. Hu, Adaptive backstepping fault-tolerant control for
exible spacecraft with unknown bounded disturbances and actuator
failures, ISA Trans., vol. 49, no. 1, pp. 5769, 2010.
[12] R. Godard and K. D. Kumar, Fault tolerant recongurable satellite for-
mations using adaptive variable structure techniques, J. Guid. Control
Dynamics, vol. 33, no. 3, pp. 969984, MayJun. 2010.
[13] H. Lee and Y. Kim, Fault-tolerant control scheme for satellite attitude
control system, IET Control Theory Appl., vol. 4, pp. 14361450,
Aug. 2010.
[14] M. Tafazoli, A study of on-orbit spacecraft failures, Acta Astronaut.,
vol. 64, pp. 195205, Jan.Feb. 2009.
[15] O. Egeland and J. M. Godhavn, Passivity-based adaptive attitude
control of a rigid spacecraft, IEEE Trans. Autom. Control, vol. 39,
no. 4, pp. 842846, Apr. 1994.
[16] F. Lizarralde and J. T. Wen, Attitude control without angular velocity
measurement: A passivity approach, IEEE Trans. Autom. Control,
vol. 41, no. 3, pp. 468472, Mar. 1996.
[17] H. Wong, M. S. de Queiroz, and V. Kapila, Adaptive tracking control
using synthesized velocity from attitude measurements, Automatica,
vol. 37, pp. 947953, Jun. 2001.
[18] K. Subbarao and M. R. Akella, Differentiator-free nonlinear
proportional-integral controllers for rigid-body attitude stabilization, J.
Guid. Control Dynamics, vol. 27, pp. 10921096, Nov.Dec. 2004.
[19] A. M. Zou and K. D. Kumar, Adaptive attitude control of spacecraft
without velocity measurements using Chebyshev neural network, Acta
Astronaut., vol. 66, pp. 769779, Mar.Apr. 2010.
[20] M. Namvar, A class of globally convergent velocity observers for
robotic manipulators, IEEE Trans. Autom. Control, vol. 54, no. 8,
pp. 19561961, Aug. 2009.
[21] W. Ren, Distributed cooperative attitude synchronization and tracking
for multiple rigid bodies, IEEE Trans. Control Syst. Technol., vol. 18,
no. 2, pp. 383392, Mar. 2010.
[22] S. Di Gennaro, Output attitude tracking for exible spacecraft, Auto-
matica, vol. 38, pp. 17191726, Oct. 2002.
[23] D. Seo and M. R. Akella, Separation property for the rigid body
attitude tracking control problem, J. Guid. Control Dynamics, vol. 30,
pp. 15691576, Nov.Dec. 2007.
[24] Y. D. Song and W. C. Cai, Quaternion observer-based model-
independent attitude tracking control of spacecraft, J. Guid. Control
Dynamics, vol. 32, pp. 14761482, Sep.Oct. 2009.
[25] A. Tayebi, Unit quaternion-based output feedback for the attitude
tracking problem, IEEE Trans. Autom. Control, vol. 53, no. 6,
pp. 15161520, Jul. 2008.
XIAO et al.: ATTITUDE STABILIZATION OF SPACECRAFTS 2263
[26] S. H. Li, S. H. Ding, and Q. Li, Global set stabilization of the spacecraft
attitude control problem based on quaternion, Int. J. Robust Nonlinear
Control, vol. 20, pp. 84105, Jan. 2010.
[27] B. Wie and J. B. Lu, Feedback control logic for spacecraft eigenaxis
rotations under slew rate and control constraints, J. Guid. Control
Dynamics, vol. 18, pp. 13721379, Nov.Dec. 1995.
[28] H. Bang, M. J. Tahk, and H. D. Choi, Large angle attitude control
of spacecraft with actuator saturation, Control Eng. Pract., vol. 11,
pp. 989997, Sep. 2003.
[29] I. Ali, G. Radice, and J. Kim, Backstepping control design with
actuator torque bound for spacecraft attitude maneuver, J. Guid. Control
Dynamics, vol. 33, pp. 254259, Jan.Feb. 2010.
[30] A. H. J. D. Ruiter, Adaptive spacecraft attitude control with actuator
saturation, J. Guidance Control Dynamics, vol. 33, pp. 16921695,
Sep. 2010.
[31] A. Mehrabian, S. Tafazoli, and K. Khorasani, Quaternion-based attitude
synchronization and tracking for spacecraft formation subject to sensor
and actuator constraints, in Proc. AIAA Guid., Navigat., Control Conf.,
Toronto, ON, Canada, 2010, pp. 121.
[32] J. Moreno-Valenzuela, V. Santibanez, and R. Campa, On output feed-
back tracking control of robot manipulators with bounded torque input,
Int. J. Control Autom. Syst., vol. 6, pp. 7685, Feb. 2008.
[33] Y. X. Su, P. C. Muller, and C. H. Zheng, Global asymptotic saturated
PID control for robot manipulators, IEEE Trans. Control Syst. Technol.,
vol. 18, no. 6, pp. 12801288, Nov. 2010.
[34] M. Krstic, I. Kanellakopoulos, and P. V. Kokotovic, Nonlinear and
Adaptive Control Design. New York: Wiley, 1995.
[35] R. Godard, Fault tolerant control of spacecraft, Ph.D. dissertation,
Dept. Aerospace Eng., Ryerson University, Toronto, ON, Canada, 2010.
[36] H. Schaub, M. R. Akella, and J. L. Junkins, Adaptive control of
nonlinear attitude motions realizing linear closed loop dynamics, J.
Guid. Control Dynamics, vol. 24, pp. 95100, Jan.Feb. 2001.
[37] S. Nicosia and P. Tomei, Nonlinear observer and output feedback
attitude control of spacecraft, IEEE Trans. Aerosp. Electron. Syst.,
vol. 28, no. 4, pp. 970977, Oct. 1992.
[38] M. Benosman and K. Y. Lum, Passive actuators fault-tolerant control
for afne nonlinear systems, IEEE Trans. Control Syst. Technol.,
vol. 18, no. 1, pp. 152163, Jan. 2010.
[39] Y. Liu, Y. Yin, and F. Liu, Continuous gain scheduled H-innity
observer for uncertain nonlinear system with time-delay and actuator
saturation, Int. J. Innovat. Comput., Inf. Control, vol. 8, no. 12,
pp. 80778088, 2012.
[40] H. Hamidi, A. Vafaei, and A. Monadjemi, A framework for fault toler-
ance techniques in the analysis and evaluation of computing systems,
Int. J. Innovat. Comput., Inf. Control, vol. 8, no. 7B, pp. 50835094,
2012.
[41] P. Shi, Y. Yin, and F. Liu, Gain-scheduled worst case control on
nonlinear stochastic systems subject to actuator saturation and unknown
information, J. Optim. Theory Appl., vol. 156, pp. 115, Aug. 2012.
[42] M. Liu, P. Shi, L. Zhang, and X. Zhao, Fault tolerant control for
nonlinear Markovian jump systems via proportional and derivative
sliding mode observer, IEEE Trans. Circuits Syst. I, Reg. Papers,
vol. 58, no. 11, pp. 27552764, Nov. 2011.
[43] B. Jiang, Y. Guo, and P. Shi, Delay-dependent adaptive reconguration
control in the presence of input saturation and actuator faults, Int. J.
Innovat. Comput., Inf. Control, vol. 6, no. 4, pp. 18731882, 2010.
Bing Xiao received the B.S. degree in mathematics
from Tianjin Polytechnic University, Tianjin, China,
in 2007, and the M.S. degree in engineering from the
Harbin Institute of Technology at Harbin, Harbin,
China, in 2010, where he is currently pursuing the
Ph.D. degree in control science and engineering.
His current research interests include spacecraft
attitude control, fault diagnosis, and fault tolerant
control for spacecrafts.
Qinglei Hu received the B.Eng. degree from the
Department of Electrical and Electronic Engineer-
ing, Zhengzhou University, Zhengzhou, China, in
2001, and the M.Eng. and Ph.D. degrees with
specialization in controls from the Department of
Control Science and Engineering, Harbin Institute
of Technology, Harbin, China, in 2003 and 2006,
respectively.
He was a Post-Doctoral Research Fellow with
the School of Electrical and Electronic Engineering,
Nanyang Technological University, Singapore, from
2006 to 2007. From 2008 to 2009, he was with the University of Bristol
as a Senior Research Fellow. He is currently an Associate Professor with
the Harbin Institute of Technology. His current research interests include
variable structure control and applications, spacecraft fault tolerant control
and applications, and spacecraft formation ying. He has authored or co-
authored more than 60 papers in journals and conferences.
Dr. Hu was a recipient of the Royal Society Fellowship for years 2008 to
2009. He was an Associate Editor of the Journal of the Franklin Institute.
Peng Shi (M95SM98) received the B.Sc. degree
in mathematics from the Harbin Institute of Technol-
ogy, Harbin, China, the M.E. degree in systems engi-
neering from Harbin Engineering University, Harbin,
the Ph.D. degree in electrical engineering from the
University of Newcastle, Newcastle, Australia, the
Ph.D. degree in mathematics from the University
of South Australia, Adelaide, Australia, and the
D.Sc. degree from the University of Glamorgan,
Pontypridd, U.K., in 1982, 1985, 1994, 1998, and
2006, respectively.
He was a Lecturer with Heilongjiang University, Harbin. He was a Post-
Doctoral Researcher and a Lecturer with the University of South Australia. He
was a Senior Scientist with the Defence Science and Technology Organisation,
Edinburgh, Australia. He was a Professor with the University of Glamorgan.
He is currently a Professor with Victoria University, Melbourne, Australia,
and the University of Adelaide, Adelaide, Australia. He has authored or co-
authored several papers in journals and conferences. His current research
interests include system and control theory, computational and intelligent
systems, and operational research.
Dr. Shi is a fellow of the Institution of Engineering and Technology, U.K.,
and the Institute of Mathematics and its Applications, U.K. He is on the
editorial board of a number of international journals, including Automatica, the
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, the IEEE TRANSAC-
TIONS ON FUZZY SYSTEMS, the IEEE TRANSACTIONS ON SYSTEMS, MAN
AND CYBERNETICSPART B: CYBERNETICS, and the IEEE TRANSAC-
TIONS ON CIRCUITS AND SYSTEMSPART I: REGULAR PAPERS.

Вам также может понравиться