Magnetic torque attitude control of a satellite using the state-dependent

Riccati equation technique

Mohammad Abdelrahman, Insu Chang, Sang-Young Park
n
Astrodynamics and Control Laboratory, Department of Astronomy, Yonsei University, Seoul 120-749, Republic of Korea
a r t i c l e i n f o
Article history:
Received 18 November 2009
Accepted 24 February 2011
Available online 4 March 2011
Keywords:
Satellite attitude control
Magnetic torque
Non-linear optimal control
State-dependent Riccati equation
Pseudo-linear systems
a b s t r a c t
A non-linear attitude control method for a satellite with magnetic torque rods using the state-
dependent Riccati equation (SDRE) technique has been developed. The magnetic torque caused by the
interaction with the Earths magnetic eld and the magnetic moment of torque rods plays a role of the
control torque. The detailed equations of motion for this system are presented using angular velocity
and quaternions. The SDRE controller is developed for the non-linear systems which can be formed in
pseudo-linear representations using the state-dependent coefcient (SDC) method without lineariza-
tion procedure. The aim of this control system is to achieve a stable attitude within 51, and minimize
the control effort. The stability regions for the SDRE controlled satellite system are estimated through
the investigation of the stability conditions developed for pseudo-linear systems and the application of
Lyapunovs theorem. For comparisons, the Linear Quadratic Regulator (LQR) method using the solution
of the algebraic Riccati equation (ARE) is also applied to this non-linear system. The performance of the
SDRE non-linear control system demonstrates more robustness and stability than the LQR control
system when subjected to a wide range of initial conditions.
& 2011 Elsevier Ltd. All rights reserved.
1. Introduction
A number of approaches for control designs of highly non-linear
systems have been examined in past years. These approaches
include feedback linearization, adaptive control, gain scheduling,
non-linear model predictive control, and sliding mode control
methods [1]. Unlike these methods, the state-dependent Riccati
equation (SDRE) technique proposed by Cloutier et al. in 1996 [2], is
well known as an optimal control method which can be applied for
non-linear systems. Recently, the SDRE technique is being applied
actively to various elds of non-linear applications [3]: autopilot
design [4], guidance law design [5], missile guidance and control
system [6], simultaneous state and parameter estimation [7], and
control design of various non-linear systems such as a continuously
stirred tank reactor [8], an under-actuated robot [9], magnetically
levitated ball [10], system with parasitic effects [11], an articial
human pancreas [12], ducted fan [14], benchmark problems [15],
etc. Typically, the SDRE technique is also used in the control and
estimation of non-linear system related to satellites. A new non-
linear lter, state-dependent Riccati equation lter (SDREF), is
introduced by Mracek et al. [16], and applied to angular rate
estimation by Harman and Bar-Itzhack [17]. In these two papers,
the performance of the SDREF is compared with the performance of
the linearized Kalman lter (LKF) and the extended Kalman lter
(EKF). Also, Luo and Chu [18] used the SDRE method to achieve a
large-angle attitude maneuver control, and French [19] designed
control algorithm for rapid, large angle slewing maneuver of a
satellite, and it has been implemented on SIMSAT with both
thrusters and reaction wheels as control inputs. Stansbery and
Cloutier [20] applied the SDRE technique to the development of a
controller to maneuver and orient a spacecraft in the proximity of a
tumbling target. In this study, they used a simple equation neglect-
ing gravity, and assumed control actuators such as thrusters, rotors
and reaction wheels. Attitude control problems based on the SDRE
technique also have been applied to satellites with attitude hard-
ware such as internal momentumrotor [21] and reaction wheel [22].
In addition, Irvin [23] proposed the SDRE application for the
reconguration of satellite formations, and compared with linear
quadratic regulator (LQR) and sliding mode controller. Chang et al.
proposed attitude control methods in satellite formation ying [24]
and satellite tethered formation ying [25]. However, no previous
research has been presented on the application of the SDRE
technique on attitude control based on only magnetic torque rods.
Small satellites have the advantage in low cost missions due
to their minimum hardware congurations and low-power
consumption. These missions include those for scientic purposes
and remote sensing in a solo or formation ying [26,27]. The
attitude control method for these small satellites is required to
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/nlm
International Journal of Non-Linear Mechanics
0020-7462/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijnonlinmec.2011.02.009
n
Corresponding author. Tel.: 82 2 2123 5687; fax: 82 2 392 7680.
E-mail address: spark624@yonsei.ac.kr (S.-Y. Park).
International Journal of Non-Linear Mechanics 46 (2011) 758771
have simple control logic, less computing efforts and low data
storage. So, in general, control methods such as gravity-gradient
stabilization joined with magnetic control and three-axis attitude
stabilization using magnetic torquer are applied with pointing
accuracy within 51 [28]. Previous researches by Wisniewski [29]
and Psiaki [30], dealt with the problem of three-axis attitude
stabilization using magnetic torquer. In these papers, Wis-
niewski [29] presented a simple constant gain controller (CGC)
via the LQR method and proposed an innite horizon and a nite
horizon controller using periodic characteristic of the Earth
magnetic eld. Similarly, Psiaki [30] presented a time-varying,
full-state-feedback LQR method based on a constant approximate
solution of the time-varying Riccati equation. However, both
papers considered linear time-varying system taking the periodic
nature of the magnetic eld as an advantage to approximate
the solution of the Riccati equation. Lovera et al. [31], Silani and
Lovera [32], and Yan et al. [33] also studied the attitude control
of satellites using magnetic torquers. They applied linearized
dynamic models to construct the controllers. It is obvious that
the most of the studies about attitude control of satellite using
magnetic actuators are based upon using the linear time-invar-
iant or time-dependent models.
This current paper proposes an attitude control method of a
satellite with only magnetic torquers as actuators via the SDRE
technique based on pseudo-linear time-varying modeling of the
spacecraft non-linear dynamics. The aim of this control system is
to achieve a stable attitude within 51 pointing accuracy, and to
minimize the control efforts under the limitation of the low-cost
technology. However, as similar to any control method that uses
magnetic actuation, this control strategy is recommended for
satellites in relatively high inclined orbits. Based on these char-
acteristics, magnetic-torquer-based attitude control design is
studied. In the equations of motion, the gravity-gradient torque
is considered as the dominant disturbance torque. Also, the
magnetic control torque is generated by the interaction with
the Earths magnetic elds and the magnetic moment of torque
rods. The current study has addressed multiple challenges, which
are described here. First, a SDRE attitude control algorithm for
satellite with magnetic torque rods has been developed. Second,
the stability region is estimated to guarantee global stability for
the SDRE controller developed in the current research. Third,
linearized equations of the attitude motion are formulated for the
LQR controller. Fourth, the performance of the SDRE controller is
compared with that of LQR controller to give insights into the
advantages of the SDRE controller.
This paper consists of four main sections. Section 2 describes
the development of the SDRE controller and addresses the
stability issues regarding pseudo-linear systems in addition to
the application of Lyapunovs stability criteria. The dynamics and
kinematics of satellite are presented in Section 3. In this section
the equations of motion of the satellite are structured in a
pseudo-linear form to suite the application of the SDRE controller.
Also, another linearized model for the satellite motion is pre-
sented and used through the application of the LQR controller for
the sake of comparison with the candidate controller. In Section 4,
simulation results of the SDRE technique for non-linear system
are compared with results of the LQR method. The concluding
remarks are presented in Section 5.
2. SDRE technique
The SDRE technique is a non-linear control design method
for the direct construction of non-linear feedback controllers.
Through the state-dependent coefcient (SDC) factorization,
designers can represent the non-linear equations of motion as
pseudo-linear structures having state-dependent coefcients.
+ Consider the pseudo-linear system
_ x =A(x)xGu (1)
and a quadratic cost function J is dened as
J =
1
2
Z
t
f
t
0
x
T
Qxu
T
Ru

dt (2)
The Hamiltonian of the system
H(x,p,u) =p
T
(A(x)xGu)
1
2
(x
T
Qxu
T
Ru) (3) .
.
1
1
+
+

Controller Plant
Fig. 1. Plant and optimal feedback controller for the SDRE technique.
Fig. 2. Denition of the coordinate systems; a reference coordinate system (RCS),
an orbit-xed coordinate system (OCS) and a satellite body-xed coordinate
system (SCS).
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 759
The necessary conditions for optimality are
@H
@u
=0 = G
T
pRu =0 = u =R
1
G
T
p (4)
@H
@p
= _ x =A(x)xGu (5)

@H
@x
= _ p =A
T
(x)p
@A(x)
@x
i
x

T
pQx (6)
where
@A(x)
@x
i
x

T
=
@A(x)
@x
1
x
@A(x)
@x
2
x
@A(x)
@x
n
x
!
T
(7)
Dene
p(t) = P(t)x(t) (8)
where P(t) is a positive denite matrix.
Then
_ p =
_
PxP_ x (9)
From the above, a modied Riccati equation can be derived

_
P =PAA
T
PPGR
1
G
T
P
@A
@x
i
x

T
PQ (10)
with a boundary condition
P(t
f
) =0 (11)
Eq. (10) is solved backwards for P then the control gain K is
obtained according to
K =R
1
G
T
P (12)
Then the control signal u is derived as
u = Kx (13)
The mechanism of the SDRE technique is shown in Fig. 1.
The stability regions estimation of SDRE controller can be shown
as follows:
+ Consider the Pseudo-linear system
_ x =
~
A(x)x (14)
First, we shall introduce the results of Banks and Al-Jurani [13].
Suppose that
~
A(x) is continuous and let m(x) denote the largest
eigenvalue of
1
2
~
A(x)
~
A
T
(x)
h i
. If m(x)o0 for all x then the pseudo-
linear system is asymptotically stable.
Proof. Let x
0
be arbitrary and consider the ball S ={x : :x:r:x
0
:].
Since S is compact and m(x) is continuous, it follows that m attains a
maximum on S. Let
m
0
=max
xAB
m(x) (15)
Then m
0
o0. Now,
d
dt
:x:
2
= _ x
T
xx
T
_ x =x
T ~
A(x)
~
A
T
(x)
h i
x
rm(x)x
2
(16)
and so x
2
decreases. Hence, if (0)=x
0
, then
d
dt
:x:
2
rm
0
:x:
2
(17)
and :x:
2
re
m
0
t
:x
0
:
2
, so that :x:-0, for any x
0
.
Second, let us apply the Lyapunovs method by choosing
V(x) =x
T
Px (18)
0 5 10
-50
0
50
R
o
l
l

(
d
e
g
)
10 15 20
-10
0
10
20 25 30
-5
0
5
0 5 10
-50
0
50
P
i
t
c
h

(
d
e
g
)
10 15 20
-10
0
10
20 25 30
-10
0
10
0 5 10
-50
0
50
100
Y
a
w

(
d
e
g
)
10 15 20
-50
0
50
Orbit
20 25 30
-50
0
50
Fig. 3. Pointing error based on LQR controllerSatellite A.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 760
0 5 10
-20
0
20
40
R
o
l
l

(
d
e
g
)
10 15 20
-5
0
5
20 25 30
-2
0
2
0 5 10
-50
0
50
P
i
t
c
h

(
d
e
g
)
10 15 20
-5
0
5
20 25 30
-0.5
0
0.5
0 5 10
-50
0
50
Y
a
w

(
d
e
g
)
10 15 20
-10
0
10
20
Orbit
20 25 30
-5
0
5
Fig. 5. Pointing error based on LQR controllerSatellite B.
0 5 10
-0.05
0
0.05
W
x

(
d
e
g
/
s
)
10 15 20
-2
0
2
x 10
-4
20 25 30
-2
0
2
x 10
-4
0 5 10
-5
0
5
x 10
-4
W
y

(
d
e
g
/
s
)
10 15 20
-2
0
2
x 10
-4
20 25 30
-2
0
2
x 10
-4
0 5 10
-1
0
1
x 10
-3
W
z

(
d
e
g
/
s
)
10 15 20
-5
0
5
x 10
-4
Orbit
20 25 30
-5
0
5
x 10
-4
Fig. 4. Angular velocity error based on LQR controllerSatellite A.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 761
as a Lyapunov function candidate, where P is the positive denite
solution of the Riccati equation
_
V(x) = _ x
T
Pxx
T _
Pxx
T
P_ x (19)
Substituting with Eq. (1), where
u =R
1
G
T
Px (20)
and using the modied Riccati equation
_
V(x) =x
T
PGR
1
G
T
Pxx
T
Qxx
T
@A
@x
i
x

T
Px (21)
Let z =R
1
G
T
Px be some vector, then
_
V(x) =z
T
Rzx
T
Qxx
T
Lx (22)
where L=((qA/qx
i
)x)
T
P
Since Q and R are positive denite matrices then stability
regions should be estimated that lead to a positive denite matrix
L and hence,
_
V(x)o0 (23)
To estimate the stability regions of the candidate SDRE con-
troller, uniformly random distributed initial conditions Y and O
on the intervals 7Y
*
1 and 7O
*
1/s for each axis, are generated
and the following conditions are investigated:
Rel
max
1
2
(AGK)(AGK)
T
h i

o0 (24)
V(x)40;
_
V(x)o0 (25)
3. Problem statements
The coordinate systems used in the current study are a satellite
body-xed coordinate system (SCS), an orbit-xed coordinate sys-
tem (OCS), and a reference coordinate system (RCS), under the
circular orbit assumption [26]. Each coordinate system is illustrated
in Fig. 2. The SCS is chosen such that it has the origin at the center of
mass of the satellite and coincides with the principal axes. When the
satellite has the desired attitude, this SCS is aligned with the OCS.
The OCS is a right-hand orthogonal system with the origin xed at
the center of mass of the satellite. In the OCS, z-axis points at the
center of the Earth and its y-axis is parallel to the opposite direction
of the angular momentum of the orbital plane. Then, the x-axis is
dened by using the right-hand rule. The RCS is an Earth-centered
inertially xed coordinate system. The z-axis is aligned with the
rotation axis of the Earth and the x-axis is points towards Vernal
Equinox. The y-axis is dened by the right-hand rule.
The equations of motion for a satellite are expressed by two sets
of kinematic and dynamic equations. The kinematic equation can be
described by quaternions (q
so
=
q
1
q
2
q
3

T
, q
2
1
q
2
2
q
2
3
q
2
4
=
1) as attitude coordinate, and the dynamic equation is dened by
Eulers equation
_ q
so
=
1
2
(q
4
o
so
o
so

q
so
)
_ q
4
=
1
2
(o
so
)
T
q
so
(26)
d
dt
H
i
=
d
dt
H
s
o
si

h i
H =L
gg
L
ctrl
, H =Io
si
(27)
where superscripts i, o, and s denote the inertial-reference, the orbit-
xed coordinate, and the body-xed coordinates, respectively. [o]
denotes the skew symmetric matrix of the angular velocity. o
si
is
0 5 10
-0.05
0
0.05
W
x

(
d
e
g
/
s
)
10 15 20
-2
0
2
x 10
-4
20 25 30
-5
0
5
x 10
-5
0 5 10
-5
0
5
x 10
-4
W
y

(
d
e
g
/
s
)
10 15 20
-2
0
2
x 10
-4
20 25 30
-1
0
1
x 10
-5
0 5 10
-1
0
1
x 10
-3
W
z

(
d
e
g
/
s
)
10 15 20
-2
0
2
x 10
-4
Orbit
20 25 30
-5
0
5
x 10
-5
Fig. 6. Angular velocity error based on LQR controllerSatellite B.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 762
the angular velocity vector of the body-xed coordinate with respect
to the inertial-reference coordinate. o
so
is the angular velocity
vector of the body-xed coordinate with respect to the orbit-xed
coordinate. Also, H is the angular momentum vector of the rigid
body, L
gg
is the disturbance torque (gravity gradient torque), L
ctrl
is
the control torque, and I is the constant moment of inertia matrix as
seen from the body-xed coordinate for the rigid body. Substituting
H=Io
si
into Eq. (27), then
I _ o
si
=[o
si
](Io
si
) L
gg
L
ctrl
(28)
Here, Eq. (28) should be expressed as the equation using o
so
in
order to represent the equation describing the motion of the body-
xed coordinate with respect to the orbit-xed coordinate. There-
fore, by substituting the equation, o
si
=o
so
o
oi
=o
so
C(q
so
)o
oi
o
into Eq. (28), the dynamic equation can be expressed as follows:
_ o
so
= (C(q
so
)o
oi
o
)
h i
o
so
I
1
[o
so
]Io
so
I
1
(IC(q
so
)o
oi
o
)
h i
o
so
I
1
(C(q
so
)o
oi
o
)
h i
Io
so
I
1
(C(q
so
)o
oi
o
)
h i
IC(q
so
)o
oi
o
I
1
L
gg
I
1
L
ctrl
(29)
where o
oi
o
= 0 o
o
0

T
rad=s is the angular velocity of the
orbital-reference coordinate with respect to the inertial-reference
coordinate expressed in the orbital-reference coordinate and C(q
so
)
denote the circular orbital rate and the direction cosine matrix from
the orbit-xed coordinate to body-xed coordinate [34]. Based on
the satellite model of Wisniewski [29] and Psiaki [30], the following
gravity gradient torque is considered as important disturbance
torque [35]
L
gg
=3
m

R
3
CM
(C
3
IC
3
) =3o
2
o
C
3
IC
3
(30)
where C
3
= c
31
c
32
c
33

T
is the third column vector in the
direction cosine matrix and m

is the Earth gravitational constant.

R
CM
is the distance between the center of mass of the Earth and the
center of mass of the satellite. The control torque L
ctrl
is generated by
the interaction with the Earths magnetic eld and magnetic
moment caused by magnetic torque rods. Thus, the control torque
is given by the following relation:
L
ctrl
=m B (31)
where m= m
1
m
2
m
3

T
is the magnetic dipole moment of the
torque rods and B = b
1
b
2
b
3

T
is the Earths magnetic eld
vector.
The necessary condition for the power optimality of a control
law is that the magnetic moment lies on a plane perpendicular to
the geomagnetic eld vector. Therefore, through the following
mapping, Eq. (32), m is exactly perpendicular to the local
geomagnetic eld vector. The direction of the control signal
vector m can be chosen arbitrarily by the controller [29]
m/m, m=
B m
B
2
(32)
where m= m
1
m
2
m
3

T
represents a new control signal for
the satellite. From Eqs. (31) and (32), the magnetic control vector
0 5 10
-50
0
50
100
R
o
l
l

(
d
e
g
)
10 15 20
-2
-1
0
1
20 25 30
-1
0
1
0 5 10
-20
0
20
40
P
i
t
c
h

(
d
e
g
)
10 15 20
-1
0
1
20 25 30
-1
0
1
0 5 10
-50
0
50
100
Y
a
w

(
d
e
g
)
10 15 20
-5
0
5
10
Orbit
20 25 30
-5
0
5
Fig. 7. Pointing error based on SDRE controllerSatellite A.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 763
0 5 10
-0.1
0
0.1
W
x

(
d
e
g
/
s
)
10 15 20
-2
0
2
x 10
-3
20 25 30
-2
0
2
x 10
-3
0 5 10
-0.05
0
0.05
W
y

(
d
e
g
/
s
)
10 15 20
-1
0
1
x 10
-3
20 25 30
-1
0
1
x 10
-3
0 5 10
-0.1
0
0.1
W
z

(
d
e
g
/
s
)
10 15 20
-5
0
5
x 10
-3
Orbit
20 25 30
-5
0
5
x 10
-3
Fig. 8. Angular velocity error based on SDRE controllerSatellite A.
0 5 10 15 20 25 30
-10
0
10
m
x

(
A
.
m
2
)
0 5 10 15 20 25 30
-10
0
10
m
y

(
A
.
m
2
)
0 5 10 15 20 25 30
-10
0
10
Orbit
m
z

(
A
.
m
2
)
Fig. 9. Dipole moment history based on SDRE controllerSatellite A.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 764
is as follows:
L
ctrl
=G(t)m (33)
where
G =
1
:B:
2
B [ ] B [ ] (34)
Therefore, by substituting Eqs. (30) and (33) into Eq. (29), the
dynamic equation for the satellite is expressed as
_ o
so
= [(C(q
so
)o
oi
o
)]o
so
I
1
[o
so
]Io
so
I
1
[(IC(q
so
)o
oi
o
)]o
so
I
1
[(C(q
so
)o
oi
o
)]Io
so
I
1
[(C(q
so
)o
oi
o
)]IC(q
so
)o
oi
o
I
1
3o
2
o
C
3
IC
3
I
1
G(t)m (35)
In order to apply the SDRE technique to the dynamic system,
Eq. (35) should be expressed in a pseudo-linear form. Hence, the
gravity-gradient torque term in Eq. (35) can be rewritten as
follows:
I
1
3o
2
o
C
3
IC
3
=3o
2
o
I
1
2(I
2
I
3
)q
4
C
33
0 2(I
2
I
3
)q
2
C
33
2(I
3
I
1
)q
3
C
33
2(I
3
I
1
)q
4
C
33
0
0 2(I
1
I
2
)q
4
C
32
2(I
1
I
2
)q
1
C
32
2
6
4
3
7
5
q
1
q
2
q
3
2
6
4
3
7
5
(36)
where I
1
, I
2
, and I
3
are components on the diagonal of the
principal moment of inertia I.
The angular velocity vector and the quaternion are used as the
systems states. Note that the vector part of the quaternion is
chosen as a feedback and the assumption of q
4
Z0 is taken here,
since the attitude quaternion has duality and its elements are
correlated [36]
x =
o
so
q
so
" #
(37)
Combining Eqs. (35) and (26), the state-space equations for
attitude motion of the satellite in the orbital-reference coordinate
are obtained:
_ o
so
_ q
so
" #
=
A
11
A
12
A
21
A
22
" #
o
so
q
so
" #

I
1
G(t)
0
33
" #
u (38)
where
A
11
=o
o
[C
2
]I
1
[o
so
]Io
o
I
1
[(IC
2
)] o
o
I
1
[C
2
]I (39)
A
12
=o
2
o
I
1
[C
2
]I
2q
2
0 2q
4
q
1
(2r) q
2
r q
3
(2r)
2q
4
2q
3
0
2
6
4
3
7
5
3o
2
o
I
1
2(I
2
I
3
)q
4
C
33
0 2(I
2
I
3
)q
2
C
33
2(I
3
I
1
)q
3
C
33
2(I
3
I
1
)q
4
C
33
0
0 2(I
1
I
2
)q
4
C
32
2(I
1
I
2
)q
1
C
32
2
6
4
3
7
5
(40)
A
21
=
1
2
q
4
U
33
(41)
0 5 10 15 20 25 30
-5
0
5
x 10
-4
T
y

(
N
.
m
)
0 5 10 15 20 25 30
-10
-5
0
5
x 10
-4

T
x

(
N
.
m
)
0 5 10 15 20 25 30
-5
0
5
x 10
-4
Orbit
T
z

(
N
.
m
)
Fig. 10. Control torque history based on SDRE controllerSatellite A.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 765
A
22
=
1
2
o
so

(42)
while u =m and r 1=(1q
2
4
):
Here C
2
is the second column of the direction cosine matrix,
and U
33
is a 33 unit matrix.
Most of the previous researches that have dealt with the
problem at hand, focused on the periodic property of the
magnetic eld to simplify the solution of the resulting Riccati
equation which in turn becomes periodic. Also, these models are
based upon linearized dynamic and kinematic equations. So,
before the simulation section, a linearized model for satellite
dynamics and kinematics shall be introduced in addition to a
dipole approximation of the Earths magnetic eld. For attitude
control of a satellite with magnetic torquers, Wisniewski [29] and
Psiaki [30] have used linearized equations of motion for a satellite
with magnetic torquer in control design. Eq. (43) represents the
linearized state-space model of the satellite. The rst part is
according to Psiaki linearized model but the Euler angles are
replaced by the quaternion to represent the satellite kinematics.
The input matrix in the second part of Eq. (43) is typical to
Wisniewski, i.e. Eq. (34). These modications are only considered
for the sake of comparison with SDRE controller which is based on
the model in Eq.(38)
d _ o
so
d_ q
so
" #
=
~
A
11
~
A
12
~
A
21
0
33
" #
do
so
dq
so
" #

I
1
G
0
0
33
" #
du (43)
where
~
A
11
=
0 0 o
o
(1s
1
)
0 0 0
o
o
(1s
3
) 0 0
2
6
4
3
7
5 (44)
~
A
12
=
8o
2
o
s
1
0 0
0 6o
2
o
s
2
0
0 0 2o
2
o
s
3
2
6
4
3
7
5 (45)
~
A
21
=
1
2
U
33
(46)
where s
1
(I
2
I
3
)/I
1
, s
2
(I
3
I
1
)/I
2
, and s
3
(I
1
I
2
)/I
3
.
If there is no Earth rotation and orbit precession, a dipole
model of the Earths magnetic eld can be approximated to the
following periodic magnetic eld model [30]:
~
B =
~
b
1
(t)
~
b
2
(t)
~
b
3
(t)
2
6
6
4
3
7
7
5
=
m
f
a
3
coso
o
tUsini
m
cosi
m
2sino
o
tUsini
m
2
6
4
3
7
5 =
~
B(t T) (47)
where i
m
is the inclination of the satellites orbit with respect to the
magnetic equator, a is the orbits semi-major axis, m
f
is the elds
dipole strength (m
f
=7.910
15
Wb m) and T=2 p/o
o
is the orbital
period. Therefore, G
0
is dened as the average of the matrix G in
Eq.(32) over one period and can be expressed as follows:
G
0
=
m
2
f
4R
6
CM
2(cos2i
m
3) 0 0
0 5(cos2i
m
3) 0
0 0 cos2i
m
3
2
6
4
3
7
5 (48)
Now, if the steady-state solution of the Riccati equation is
considered as in Eq. (49), then the control problem is treated as a
simple LQR
PAA
T
PPGR
1
G
T
PQ =0 (49)
0 5 10
-50
0
50
100
R
o
l
l

(
d
e
g
)
10 15 20
-5
0
5
10
20 25 30
-0.5
0
0.5
1
0 5 10
0
50
100
P
i
t
c
h

(
d
e
g
)
10 15 20
-10
0
10
20
20 25 30
0
1
2
3
0 5 10
-50
0
50
100
Y
a
w

(
d
e
g
)
10 15 20
-20
0
20
Orbit
20 25 30
-1
-0.5
0
0.5
Fig. 11. Pointing error based on SDRE controllerSatellite B.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 766
0 5 10
-0.05
0
0.05
W
x

(
d
e
g
/
s
)
10 15 20
-0.01
0
0.01
20 25 30
-1
0
1
x 10
-3
0 5 10
-0.05
0
0.05
W
y

(
d
e
g
/
s
)
10 15 20
-0.01
0
0.01
20 25 30
-5
0
5
x 10
-4
0 5 10
-0.05
0
0.05
W
z

(
d
e
g
/
s
)
10 15 20
-0.01
0
0.01
Orbit
20 25 30
-1
0
1
x 10
-3
Fig. 12. Angular velocity error based on SDRE controllerSatellite B.
0 5 10 15 20 25 30
-0.5
0
0.5
m
x

(
A
.
m
2
)
0 5 10 15 20 25 30
-0.5
0
0.5
m
y

(
A
.
m
2
)
0 5 10 15 20 25 30
-0.5
0
0.5
Orbit
m
z

(
A
.
m
2
)
Fig. 13. Dipole moment history based on SDRE controllerSatellite B.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 767
4. Numerical simulations and discussions
Based on the rationale for the SDRE technique, the attitude
control method for non-linear equations of motion of the satellite
is simulated by using only magnetic torque rods in this section.
Before the simulation of the SDRE technique for the non-linear
system, the system controlled by the LQR method is simulated
under the non-linear attitude dynamic environment. The LQR
method produces the constant gain controller (CGC) to regulate
the non-linear system and its corresponding simulation results
are used as a reference for the analysis of the SDRE technique. The
SDRE technique produces the variable gain controller (VGC)
having the form of Eq. (12) and regulates the non-linear systems
equations within the required accuracy. The results of these
simulations are compared with those by the LQR method. To
show the potential of the SDRE technique as an alternative non-
linear control scheme for spacecraft attitude control.
These simulation procedures are applied to the magnetic-
torquer-based attitude control system. In general, this control
system is comparatively cheap and lightweight and has low
power efciency. However, using the Earths magnetic elds, they
can usually be applied for satellites with a near-polar orbit [29].
Therefore, two small satellites with high-inclined low-Earth-orbit
are selected in these simulation studies. First, the moments
of inertia about principal axes of two satellites (A and B) are
I
A
=diag( 250 250 150)kgm
2
and I
B
=diag( 8 10 6) kgm
2
,
respectively. Satellite A is assumed to be a rectangular parallele-
piped satellite with a large gravity-gradient torque for special
activity such as Techsat 21 [27], while satellite B is assumed to
be a small satellite of scientic purposes with cubic shape. The
control method based on only the magnetic torque rods can be
applied to these two satellites. The candidate orbit charac-
teristics are a circular orbit with a 600 km altitude and a 901
inclination. Also, the orbital angular rate o
o
of satellite is about
0.0010831 rad/s and the orbital period T is about 5801 s.
The weighting matrices (Q and R) have been tuned to meet the
required pointing accuracy (51 for each axis), and the actuators
dipole moments (710 A m
2
and 70.5 A m
2
) for both satellites,
respectively. For Satellite A, Q is a diagonal matrix of diag( 6:0
10
8
8:0 10
8
6:0 10
8
10
5
10
5
10
5
) and R is a diagonal matrix
of diag( 10
10
10
10
10
10
). For Satellite B, Q is a diagonal matrix
of diag( 10
10
5:0 10
11
10
10
10
7
10
7
10
7
) and R is a diag-
onal matrix of diag( 10
8
10
8
10
8
). While the same weighting
matrices are also applied to the LQR method, the magnetic eld
model IGRF 2005 is used for magnetic eld vector simulation in the
SDRE controller and the dipole model of the Earths magnetic eld in
Eq. (47) is used in LQR controller for the sake of consistency with the
approaches presented in [29,30]. The initial attitude for both cases is
301 and rate of 0.051/s in each axis. The integration algorithmused for
the modied Riccati equation is RungeKutta ode4 along a time span
of 30 orbits with 4 s as a sample time.
The pointing and angular velocity errors of the LQR-controlled
dynamic system are shown in Figs. 36. For Satellite A, roll and
pitch errors are controlled within a broader range than the
required boundary, and the yaw error cannot be properly con-
trolled as shown in Fig. 3. In the case of Satellite B, roll and pitch
errors can be coarsely controlled by the CGC while the yaw error
takes upto 20 orbits to converge to the required pointing
accuracy Fig. 5. It means that the satellite using linearized system
is only controllable in two axes and the full controllability must
0 5 10 15 20 25 30
-5
0
5
x 10
-5
T
x

(
N
.
m
)
0 5 10 15 20 25 30
-5
0
5
x 10
-5
T
y

(
N
.
m
)
0 5 10 15 20 25 30
-5
0
5
x 10
-5
Orbit
T
z

(
N
.
m
)
Fig. 14. Control torque history based on SDRE controllerSatellite B.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 768
be achieved by applying other additional control torque. In order
to achieve a good capability, therefore, the additional control
system such as PID controller must be used in attitude control
system with the LQR controller for non-linear system. To solve
this problem, the performance of the attitude control via the VGC
obtained by applying the SDRE technique is evaluated.
The simulation results for the control performance of the non-
linear system using the SDRE technique are given in Figs. 714. For
satellite A, the pointing error converges to the required accuracy
after 5 orbits while the angular velocity error approaches less than
0.003 rad/s as shown in Figs. 7 and 8. Compared with the results of
the LQR controller, the SDRE based controller successfully controlled
the satellite in the three-axes within a reasonable period and
achieved the mission requirements.
Simulations for satellite B in Figs. 11 and 12 show the pointing
and angular velocity errors, respectively. It clearly, can be shown
that the pointing errors take upto 12 orbits to reach the assigned
accuracy which is almost the half time needed when utilizing the
LQR controller. Actually, when the initial conditions for the
angular velocity (0.051/s) are applied to the LQR, it failed to
converge, so a very small value has been chosen for the LQR of
0.0011/s while the initial attitude errors have been kept for both
controllers. The applied dipole moments and the corresponding
control torques for both satellites are shown in Figs. 9, 10 and 13,
14, respectively. The results of satellite A show many spikes in the
dipole moments of roll and yaw axes, and hence similar spikes in
the pitch torque component. According to the simple magnetic
control torque relation in Eq. (31), and due to the nature of the
polar orbit where the magnetic eld vector almost lies in the orbit
plane, the y-axis component is very small. If we expand the vector
Eq. (31) to the scalar form, it can be shown that to apply the
required control torques in x- and z-axis, very high dipole
moments in the same axes should be generated. The same
argument can be applied to satellite B, in addition to the small
dipole moment strength which was originally sized according to
the detumbling mode and not for stabilization mode. The stability
of the SDRE approach is investigated rst through the stability
indices in Eqs. (24) and (25) for the design point, i.e., the pro-
posed initial conditions. Then the stability regions of SDRE
controller (Figs. 15 and 16), is estimated by generating uniformly
random distributed initial conditions Y and O on the intervals
150 150

1 and 4 4

1=s for each axis, respectively, and
the conditions in Eqs. (24) and (25) are investigated. Figs. 17 and
18 have been generated via 200 points of different initial condi-
tions representing the off-design conditions. Each point satised
the stability indices over a time span of 30 orbits.
5. Conclusions
A non-linear control technique for satellite attitude control
using only magnetic torquer is developed based on the SDRE
approach. The proposed controller is derived through the solution
of a modied Riccati equation which takes into account the
dependency of the state matrix on the states, and hence generates
an optimal control signal rather than the standard SDRE which
neglects this dependency or utilizes purely linearized systems
and hence generates a suboptimal controller. The governing
equations of motion of the satellite are factorized and represented
in a pseudo-linear form for the application of the SRDE. Also,
0 5 10 15 20 25 30
-5
0
5
10
15
20
x 10
4
L
y
a
p
u
n
o
v

F
u
n
c
t
i
o
n
0 5 10 15 20 25 30
-30
-20
-10
0
Orbit
M
a
x

E
i
g
e
n
v
a
l
u
e

1
/
2
(
A
'
+
A
)
V
dV/dt
Fig. 15. Stability indices of the SDRE controllerSatellite A.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 769
a linearized model is introduced and used for the LQR controller.
The simulation results for the non-linear system are presented
and analyzed through the comparison with the control perfor-
mance by the LQR method. These results show that the SDRE
technique is capable of controlling the satellite in the three-axes
and achieving the required pointing accuracy within a reasonable
time. Also, the SDRE controller shows a good capability to
stabilize the satellite in case of relatively moderate initial angular
velocities while the LQR controller is only suitable for very small
values of initial angular velocities. The SDRE controller is proved
to be stable for a wide range of pointing and angular velocity
errors through extensive simulations.
Acknowledgments
This work is supported by the Korean Science and Engineering
Foundation (KOSEF) through the National Research Laboratory
Program funded by the Ministry of Science and Technology (no.
M10600000282-06J0000-28210).
References
[1] E.B. Erdem, Analysis and real-time implementation of state-dependent
Riccati equation controlled systems, Ph.D. Dissertation, University of Illinois
at Urbana-Champaign, Illinois, 2001, pp. 535.
[2] J.R. Cloutier, C.N. DSouza, C.P. Mracek, Nonlinear regulation and nonlinear
HN control via the state-dependent Riccati equation technique; Part 1,
Theory; Part 2, Examples, in: Proceedings of the 1st International Conference
on Nonlinear Problems in Aviation and Aerospace, Available through Uni-
versity Press, Embry-riddle Aeronautical University, Daytona Beach, FL, May
1996.
[3] J.R. Cloutier, D.T. Stansbery, The capabilities and art of state-dependent
Riccati equation-based design, in: Proceedings of the American Control
Conference, Anchorage, Alaska, May 2002.
[4] J.R. Cloutier, D.T. Stansbery, Nonlinear, hybrid bank-to-turn/skid-to-turn
autopilot design, in: Proceedings of the Guidance, Navigation, and Control
Conference, Montreal, Canada, August 2001.
[5] J.R. Cloutier, P.H. Zipfel, Hypersonic guidance via the state-dependent Riccati
equation control method, in: Proceedings of the IEEE Conference on Control
Applications, Hawaii, August 1999.
[6] P.K. Menon, V.R. Iragavarapu, Integrated design of agile missile guidance and
control systems, in Naval Surface Warfare Center, SBIR Report, Dahlgren,
Virginia, October 1997.
[7] D. Haessing, B. Friendland, A method for simultaneous state and parameter
estimation in nonlinear systems, in: Proceedings of the American Control
Conference, Albuquerque, NM, June 1997.
[8] J.R. Cloutier, D.T. Stansbery, Control of a continuously stirred tank reactor
using an asymmetric solution of the state-dependent Riccati equation, in:
Proceedings of the IEEE Conference on Control Applications, Hawaii, August
1999.
[9] E.B. Erdem, A.G. Alleyne, Experimental real-time SDRE control of an under-
actuated robot, in: Proceedings of the American Control Conference, San
Diego, CA, June 1999.
[10] E.B. Erdem, A.G. Alleyne, Globally stabilizing second-order nonlinear systems
by SDRE control, in: Proceedings of the American Control Conference, San
Diego, CA, June 1999.
[11] B. Friendland, Feedback control of systems with parasitic effects, in: Proceed-
ings of the American Control Conference, Albuquerque, NM, June 1997.
[12] D.K. Parrish, D.B. Ridgely, Control of an articial human pancreas using the
SDRE method, in: Proceedings of the American Control Conference, Albu-
querque, NM, June 1997.
-4
-4
-4
-
4
-4
-
4
-
4
-4
-4
-4
-4
-4
-4
-4
-
4
-2
-2
-2
-2
-2
-2
-
2
-
2
-
2
-
2
-2
-2
-
2
-
2
-2
-2
-
2
-2
-2
-2
-2
-
2
-
2
-2
0
0 0 0 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2 2
2 2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
4
4
4
4 4
4
4
4
W
x
(deg/s)
W
z

(
d
e
g
/
s
)


-4 -3 -2 -1 0 1 2 3 4
-4
-3
-2
-1
0
1
2
3
4
W
y
(deg/s)
-5
-4
-3
-2
-1
0
1
2
3
4
5
Fig. 18. Angular velocity stability regimes.
0 5 10 15 20 25 30
-5
0
5
10
15
x 10
6
L
y
a
p
u
n
o
v

F
u
n
c
t
i
o
n
0 5 10 15 20 25 30
-1
-0.5
0
Orbit
M
a
x

E
i
g
e
n
v
a
l
u
e

1
/
2
(
A
'
+
A
)
V
dV/dt
Fig. 16. Stability indices of the SDRE controllerSatellite B.
-80
-
6
0
-
6
0
-
6
0
-6
0
-6
0
-6
0
-6
0
-
4
0
-40
-
4
0
-4
0
-4
0
-4
0
-4
0
-4
0 -2
0
-2
0
-
2
0
-2
0
-2
0
-2
0
-
2
0
-
-
2
0
-
2
0
2
0
4
0
4
0
-2
0
0
0
0
0
0
0
0
0 0
0
2
0
2
0
2
0
20
20
2
0
2
0
2
0
2
0
2
0
4
0
40
4
0
40
40
4
0
4
0
40
4
0
4
4
0
0
4
0
60
60
6
0
6
0
6
0
6
0
6
0
6
0
6
0
6
0
Roll (deg)
Y
a
w

(
d
e
g
)

-150 -100 -50 0 50 100 150
-150
-100
-50
0
50
100
150
Pitch (deg)
-80
-60
-40
-20
0
20
40
60
Fig. 17. Euler angles stability regimes.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 770
[13] S.P. Banks, S.K. Al-Jurani, Pseudo-linear systems, Lie algebras, and stability,
IMA Journal of Mathematical Control & Information 13 (1996) 385401.
[14] M. Sznaier, J.R. Cloutier, R.A. Hull, D. Jacques, C.P. Mracek, Receding horizon
control Lyapunov function approach to suboptimal regulation of non-
linear systems, Journal of Guidance, Control, and Dynamics 23 (3) (2000)
399405.
[15] C.P. Mracek, J.R. Cloutier, Control design for the nonlinear benchmark
problem via the state-dependent Riccati equation method, International
Journal of Robust and Nonlinear Control 8 (45) (1998) 401433.
[16] C.P. Mracek, J.R. Cloutier, D.A. DSouza, A new technique for nonlinear
estimation, in: Proceedings of the IEEE Conference on Control Applications,
Dearborn, MI, September 1996.
[17] R.R. Harman, I.Y. Bar-Itzhack, Pseudolinear and state-dependent Riccati
equation lters for angular rate estimation, Journal of Guidance, Control,
and Dynamics 22 (5) (1999) 723725.
[18] W. Luo, Y.C. Chu, Attitude control using the SDRE technique, in: Seventh
International Conference on Control, Automation, Robotics and Vision,
Singapore, December 2002.
[19] D.B. French, Hybrid control strategies for rapid, large angle satellite slew
maneuvers, M.S. Thesis, Air Force Institute of Technology, Wright-Patterson
AFB, Ohio, March 2003, AFIT/GA/ENY/03-2.
[20] D.T. Stansbery, J.R. Cloutier Position and attitude control of a spacecraft using
the state-dependent Riccati equation technique, in: Proceedings of the
American Control Conference, Chicago, IL, June 2000.
[21] D.K. Parrish, D.B. Ridgely, Attitude control of a satellite using the SDRE
method, in: Proceedings of the American Control Conference, Albuquerque,
NM, June 1997.
[22] M. Xin, S.N. Balakrishnan, State dependent Riccati equation based spacecraft
attitude control, in: AIAA Aerospace Sciences Meeting and Exhibit, 40th,
Reno, NV, January 2002.
[23] D.J. Irvin, A study of linear vs. nonlinear control techniques for the recon-
guration of satellite formations, M.S. Thesis, Air Force Institute of Technology,
Wright-Patterson AFB, Ohio, March 2001, AFIT/GA/ENY/01M-02.
[24] I. Chang, S.-Y. Park, K.-H. Choi, Decentralized attitude control for satellite
formation ying via the state-dependent Riccati equation technique, Inter-
national Journal of Non-Linear Mechanics 44 (8) (2009) 891904.
[25] I. Chang, S.-Y. Park, K.-H. Choi, Nonlinear attitude control of a tether-
connected multi-satellite in three-dimensional space, IEEE Transactions on
Aerospace and Electronic Systems 46 (4) (2010) 19501968.
[26] R. Wisniewski, Satellite attitude control using only electromagnetic actuation,
Ph.D. Dissertation, Aalborg University, Aalborg, Denmark, December 1996.
[27] A. Das, R. Cobb, M. Stallard, TechSat 21: a revolutionary concept in
distributed space based sensing, in: Proceedings of the AIAA Defense and
Civil Space Programs Conference and Exhibit, Huntsville, AL, October 1998,
AIAA 99-4570.
[28] J.R. Wertz, W.J. Larson, Space Mission Analysis and Design, 3rd ed, Micro-
cosm Press, El Segundo, CA, 1999, pp. 359364.
[29] R. Wisniewski, Linear time varying approach to satellite attitude control
using only electromagnetic actuation, Journal of Guidance, Control, and
Dynamics 23 (4) (2000) 640647.
[30] M.L. Psiaki, Magnetic torquer attitude control via asymptotic periodic linear
quadratic regulation, Journal of Guidance, Control, and Dynamics 24 (2)
(2001) 386394.
[31] M. Lovera, E.D. Marchi, S. Bittanti, Periodic attitude control techniques for
small satellites with magnetic actuators, IEEE Transactions on Control
Systems Technology 10 (1) (2002) 9095.
[32] E. Silani, M. Lovera, Magnetic spacecraft attitude control: a survey and some
new results, Control Engineering Practice 13 (2005) 357371.
[33] H. Yan, I.M. Ross, K.T. Alfriend, Pseudospectral feedback control for three-axis
magnetic attitude stabilization in elliptic orbits, Journal of Guidance, Control,
and Dynamics 30 (4) (2007) 11071115.
[34] B. Wie, Space Vehicle Dynamics and Control, AIAA Educational Series, Reston,
VA, 1998, pp. 307380.
[35] P.C. Hughes, Spacecraft Attitude Dynamics, Dover, Mineola, NY, 1986
pp. 232280.
[36] B. Wang, K. Gong, D. Yang, J. Li, Fine attitude control by reaction wheels using
variable-structure controller, Acta Astronautica 52 (2003) 613618.
M. Abdelrahman et al. / International Journal of Non-Linear Mechanics 46 (2011) 758771 771