Chinese Journal of Aeronautics, (2018), 31(8): 1786–1796

Chinese Society of Aeronautics and Astronautics

& Beihang University
Chinese Journal of Aeronautics
cja@buaa.edu.cn
www.sciencedirect.com

Observer-based control for the platform of a

tethered space robot
Guang ZHAI a,*, Heming ZHENG a, Bo ZHANG b

a
Aerospace Engineering School, Beijing Institute of Technology. Beijing 100081, China
b
Aerospace System Engineering Shanghai, Shanghai 201109, China

Received 13 July 2017; revised 23 August 2017; accepted 13 September 2017

Available online 3 February 2018

KEYWORDS
Adaptive control;
Attitude dynamics;
Disturbance observer;
Sliding mode control;
Tethered space robot
Abstract This paper addresses the attitude control problem of a space tethered robot platform in
the presence of unknown external disturbance caused by a connecting elastic tether. The tether-
generated unknown disturbance leads to tremendous challenges for attitude control of the platform.
In this work, the perturbed attitude dynamics of the platform are derived with a consideration of
the libration of the elastic tether, and then with the purpose of compensating the unknown distur-
bance, major attention is dedicated to develop a nonlinear disturbance observer based on gyros
measurements, after which, an adaptive attitude scheme is proposed by combining the disturbance
observer with a sliding mode controller. Finally, benefits from the observer based on an adaptive
controller are validated by series of numerical simulations.
Ó 2018 Chinese Society of Aeronautics and Astronautics. Production and hosting by Elsevier Ltd. This is
an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction work in this field is focused on space robots with manipula-

Recently, the fast increment of debris in Earth’s orbit has
received significant attentions. Evaluation on the orbital envi-
ronment indicates that large space debris, such as non-
functional satellites and upper stages, are ensured to impose
the most serious impact to the orbit environment in the near
future. With the purpose of suppressing the rapid growth of
the debris population, different types of space robot systems
are proposed for debris capture and removal. Lots of previous
* Corresponding author.
E-mail address: gzhai@bit.edu.cn (G. ZHAI).
Peer review under responsibility of Editorial Committee of CJA.
Production and hosting by Elsevier
tors,1–5 while other researchers6,7 have proposed tethered space
robots (TSRs) for large-size debris capture and removal. A
TSR system is commonly composed of a maneuverable plat-
form and a tether-connected net/gripper. Compared with a
manipulator-based space robot system, a TSR achieves poten-
tial advantages due to its enhanced error tolerance, increasing
capturing distance, and reliable safety.
There are many challenges arise from TSR practical imple-
mentation, one of which is the control problem for the
post-capture system. Generally, the dynamic behaviors of a
post-capture TSR are very similar to those of a typical tethered
satellite system. Tethered satellites have been extensively stud-
ied in the past decades under the mass point and inelastic
tether assumption. The dumbbell model is widely used to
describe the tether libration within a local vertical local
horizontal frame. For example, based on the dumbbell model,
Williams8 investigated the dynamics and control problem for a

https://doi.org/10.1016/j.cja.2018.01.012
1000-9361 Ó 2018 Chinese Society of Aeronautics and Astronautics. Production and hosting by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Platform of a tethered space robot 1787

tethered system in an elliptical orbit, and his work indicates Wu22 proposed a robust composited attitude control scheme
that certain combinations of parameters may lead to a tether for flexible spacecraft stabilization. Lee23 addressed the prob-
slack. Motivated by the potential application for orbital trans- lem of relative attitude control between two spacecraft sub-
portation, Lorenzini9 studied the problem of target rendezvous jected to an unknown disturbance, and a disturbance
with a tether tip. To facilitate the relative motion analysis observer was also used to enhance the control performance.
between the tether tip and a target, Takeichi et al.10 extended Liu et al.24 also proposed a composite controller by combining
his work to find a periodic solution of the tether librational a disturbance observer with a PD controller for attitude stabi-
motion in an elliptic orbit. Wen et al.11–13 also studied the lization and vibration reduction of flexible spacecraft, and sim-
deployment and retrieval control problem based on tether ten- ilar contributions are also available in the literature.25,26
sion regulation. Liu et al.14 also extended the previous work to However, in the above mentioned works, an unknown distur-
a tether satellite in a halo orbit, and a nonlinear output track- bance is supposed to be slow time-varying, and sometimes
ing control scheme based on the h
D technique was proposed approximately treated as a constant; therefore, for the case
for system station-keeping control. of disturbance with components of high frequency, perhaps
Due to the weakness of the dumbbell model, many the control performance will be greatly degraded.
researchers have tried to develop the detailed dynamics of a In this paper, in order to enhance the attitude stabilization
TSR by taking into account the tether elasticity. However, performance for a post-capture TSR platform, an adaptive
under the tether elasticity consideration, the longitudinal controller is developed by combining a disturbance observer
vibration of high frequency on a connecting tether makes the with a sliding mode controller to reject a disturbance from
control problem even more challenging. With the efforts to the elastic tether libration. Major attentions are dedicated to
guarantee TSR safeties, many researchers have dedicated their develop a nonlinear disturbance observer which only depends
work to control techniques for the post-capture system. In on gyros measurements. In comparison with the works in liter-
order to avoid possible collisions between captured debris ature23,24, the proposed disturbance observer in this paper also
and the platform, Okazaki and Ohtsuka15 developed a switch- works robustly in case of arbitrary time-varying disturbance,
ing control scheme, and the safety insurance conditions for the and meanwhile, the sliding mode controller in the control loop
post-capture TSR were analytically derived; finally, a satisfac- accounts for the residual disturbance after feed forward com-
tory performance of the safety control scheme was validated pensation. Finally, the benefits from the adaptive controller
by numerical simulations. By using Lagrangian theorem, Asla- are analyzed and validated by numerical simulations.
nov and Yudintsev16 studied the dynamics of a post-capture
TSR before a space tug, in which the influences from debris 2. Perturbed attitude dynamics of the platform of a TSR
initial conditions and dynamic parameters, such as debris tum-
bling, tether elasticity, and thruster accelerations, were investi- 2.1. System description
gated and numerically demonstrated. With considerations of
system uncertainties, Cleary and William17 developed a
As shown in Fig. 1, the TSR system under consideration is
wave-based control method to cope with the post-capture sys-
composed of a platform and debris of mass m connected by
tem, and this control scheme is easy to implement in practice
a straight massless tether, which is only capable of exerting
due to its less relative measurement requirement. Huang
force along the straight line connecting the debris. The debris
et al.18–20 presented a robust adaptive back-stepping controller
is considered as a passive object without any active control.
to stabilize a tether-connected debris-gripper combination, and
The tether’s unstrained length and longitudinal stiffness are l
based on utilization of the auxiliary design and the optimal
and EA, respectively. When the system operates in its orbit,
control scheme, the proposed controller effectively reduced
the elastic tether is always kept taut due to the stiffening effects
possibilities of thruster saturation.
As well as guaranteeing system safety, platform attitude
stabilization is also an important aspect for the post-capture
system. In particular, the post-capture TSR is required to per-
form orbital maneuvers for debris removal. However, during
the maneuvers, since attitude misalignment evidently leads to
a remarkable increment on fuel consumptions, the attitude
of the platform should be stabilized precisely with respect to
the given reference. However, the tether-generated distur-
bance, which includes both long-term libration and elastic
vibration, greatly increases the difficulties of platform stabi-
lization. In the early 1990s, optimal control based on an inte-
gral linear quadratic regulator21 (LQR) was proposed to
suppress the resonances on flexible spacecraft, but LQR appli-
cations are very limited due to its dependency on accurate
dynamic models. To cope with unknown external distur-
bances, robust attitude control techniques have attracted much
attention in recent years. One of the robust attitude control
methodologies is the observer-based control technique, which
estimates external unknown disturbances based on necessary
measurements for feed forward compensation. By incorporat-
ing a disturbance observer into the control loop, Yan and Fig. 1 Illustration of a Post-capture TSR system.
1788 G. ZHAI et al.

from the gravitational gradient and centrifugal force, and the _ cos a cos b
nlða_ þ XÞ sin a cos b
m r ¼ ½nl
stiffness-caused longitudinal elongation is characterized by n,

nlb_ cos a sin b^ _ sin a cos b
nlða_ þ XÞ
x þ ½nl
which is defined as
l þ Dl  cos a cos b
nlb_ sin a sin b^ _ sin b
y þ ½nl
n¼ >1 ð1Þ
l þ nlb_ cos b^
z ð4Þ
where Dl denotes the tension-caused elongation. where X ¼ ½0;
X; 0T is the orbit velocity vector. Further, the
Furthermore, for simplicities, the following assumptions absolute velocity of the debris with respect to the inertial frame
are also made during system dynamics development: (1) the can be calculated as
system moves in a well-known non-eccentric orbit; (2) the pas-
sive debris is treated as a mass point and its attitude motion is v ¼ vo þ vr ð5Þ
beyond the scope of this work; (3) the mass of the debris is where vo denotes the orbital velocity, and in case of a circular
much smaller than that of the platform, and thus, the orbit orbit, it can be computed as
center of the whole system can be assumed to coincide with
the mass center of the platform; (4) the distance from the vo ¼ X  R ð6Þ
tether’s attach point to the mass center of the platform is much where R is the orbital position vector of the system and
shorter than the length of the tether, (5) environmental pertur- kRk ¼ R. Then, the kinetic energy term can be written as
bations, such as atmosphere resistance, earth oblateness, and
1
sun pressure, are all neglected. T ¼ M0 ðvo þ vr Þ  ðvo þ vr Þ ð7Þ
The reference frames used for dynamics modeling are also 2
shown in Fig. 1. The Earth inertial frame, represented by where M0 is the total mass of the system. Then by substituting
Oxyz, is a non-rotating frame with the origin point fixed at Eq. (4) and Eq. (6) into Eq. (7), one can obtain the explicit
the mass center of the Earth. The local vertical local horizontal expression of the kinetic energy. To save the length of the
(LVLH) frame, represented by Oxoyozo, is a rotating frame, paper, the kinetic energy expression will not be represented
and its origin is located at the mass center of the platform. in detail. Next, the gravitational potential energy of the system
xo-axis is aligned with the orbital velocity, zo-axis is aligned can be formulated as
with the vector from the Earth’s center to the mass center of 1
the spacecraft, yo-axis is finally completed by following the Vg ¼
le M0 ð8Þ
jR þ qj
right-hand principal. The body-fixed frame, denoted as Oxbyb-
zb, is the frame to describe the attitude of the platform with where le is the constant gravitational coefficient. With Eq. (3),
1
respect to the LVLH frame. expanding jRþqj into a Taylor series and neglecting the high-
order terms lead to
2.2. Tether libration  
1 1 nl n2 l2
¼ 1 þ cos a þ 2 ð3 cos a
1Þ
2
ð9Þ
jR þ qj R R 2R
In this section, elastic tether liberation equations will be devel-
oped within the LVLH frame, and the in- and out-of-plan Then substitute Eq. (9) into Eq. (8), and the expression of
librations and the tether longitudinal oscillation will also be the gravitational potential energy can also be achieved as
described by the developed equations. Generally, to derive  
l M0 nl n2 l2
the equations of tether librations, the most effective approach Vg ¼
e 1 þ cos a þ 2 ð3 cos2 a
1Þ ð10Þ
seems to be the Lagrangian technique. The following equation R R 2R
represents Lagrange’s formula which provides an energy Additionally, consider the tether elastic elongation, the
method based on finding expressions for the kinetic energy elastic potential energy terms can be written as
and generalized forces:

 1
d @T @T @V Ve ¼ EAlðn
1Þ2 ð11Þ

þ ¼ Qi ð2Þ 2
dt @ q_ j @qj @qj
By combining Eq. (10) and Eq. (11), the total potential
where qj and Qi denote the generalized coordinate and corre- energy can be expressed as
sponding generalized force, while T and V represent the kinetic Vtotal ¼ Vg þ Ve ð12Þ
energy and potential energy, respectively. Seeing the geometric
relations in Fig. 1 and considering the 4th assumption, the Finally, applying Lagrangian equations with the expres-
position vector of the debris in the LVLH frame, denoted by sions of the kinematic energy and the gravitational potential,
q, can be approximately expressed as the nonlinear equations for the tether libration can be formu-
lated as
q ¼ ðnl cos a cos bÞ^
x þ ðnl sin a cos bÞ^
y þ ðnl sin bÞ^
z ð3Þ !
n_ _
where x^, y^, and z^ are the unit vector components of the LVLH €
a
2 _ þ 3X2 cos a sin a cos b ¼ Fa

b tan b ðX þ aÞ ð13Þ
n
frame, and a and b are the in- and out-of-plane libration
angles. Then, by taking the first order of derivative with Eq. " #
(3), the velocity vector of the debris within the LVLH frame n_
€ þ 2 b_ þ ðX þ aÞ
b _ þ 3X cos a cos b sin b ¼ Fb
2 2 2
ð14Þ
can be obtained as n
Platform of a tethered space robot 1789

M0 l€n þ cln_
M0 lnðX þ aÞ

_ 2 cos2 b
nb_ 2 þ EAlðn
1Þ Jx_ ¼
x Jx þ uc þ ud ð20Þ
þ M0 lX nð1
3 cos a cos bÞ ¼ Fl
2 2 2
ð15Þ where J 2 R 33
denotes the positive-definite inertia matrix,
where c denotes the damping ratio of the tether, Fa and Fb are x ¼ ½x1 ; x2 ; x3 T indicates the angular velocity of the plat-
the non-dimensional forces for the in- and out-of-plane libra- form with respect to the inertial frame, and x denotes the
tions, while Fl represents the external force acted along the skew-symmetric matrix of x which can be calculated as
2 3
tether. Eqs. (13)–(15) clearly indicate that the dynamics of 0
x3 x2
the elastic tether are seriously nonlinear and coupled, so any 6 7
x ¼ 4 x3 0
x1 5 ð21Þ
perturbation acted on the system evidently leads to a longitu-

x2 x1 0
dinal oscillation, and, as a result, generates an oscillated ten-
sional force. Under the assumption of d  l, then the tether where uc 2 R3 is the active control torque, while ud 2 R3
tensional force vector, denoted as F, can be written within denotes the disturbance torque caused by the librating tether.
the body-fixed frame as Here, it is worthy to note that the gravitational gradient distur-
h i bance torque will not be considered since it is much smaller
F ¼
Fl Cob^l ¼
EAlðn
1Þ
cln_ Cob^l ð16Þ than the tether-generated one. As shown in Fig. 1, the distur-
bance torque can be written as
where Fl is the tether tensional force, which can be calculated
ud ¼ F  d ð22Þ
as
where d is the position vector of the attach point in the body-
Fl ¼
EAlðn
1Þ ð17Þ fix frame.
where Cob is the transformation matrix from the LVLH frame Substitute Eq. (22) into Eq. (20), and then the perturbed
to the body-fixed frame, which can be calculated in terms of attitude dynamics of the platform can be represented as
Euler angles of yaw #, pitch h, and roll w. When the platform x_ ¼
J
1 ðx Jx þ F  d þ uc Þ ð23Þ
is well controlled, the transformation matrix can be approxi- 
where x represents the skew-symmetric matrix of x . Modi-
mated as Cob  I33 , in which I33 2 R33 denotes the identity
fied Rodrigues parameters (MRPs) are used to formulate the
matrix. Meanwhile,
kinematics model of the platform, in which the MRPs repre-
2 3
cos b sin a sent a minimal parameterization with a singularity at 360°,
^l ¼ 6
4 sin b 5
7
ð18Þ avoid the normalization constraint associated with quaternion
parameterization, and allow for rotations of greater than 180°.
cos b cos a The attitude kinematics of the platform can be represented as
where ^l is the unit vector describing the tether direction in the r_ ¼ GðrÞx ð24Þ
LVLH frame. In most of past works7, with the purpose of
where r ¼ ½ r1 r2 r3 T , and GðrÞ is defined as
reducing dynamic complexities, an inelastic tether is consid-
  1 
ered. Basically, by setting n ¼ 1 and n_ ¼ €n ¼ 0, the inelastic 1
GðrÞ ¼ I33 1
rT r þ r þ rrT ð25Þ
dynamics can also be obtained based on Eqs. (13)–(15). Note 4 2
that for the inelastic case, the dynamics only describe the in- where r represents the skew-symmetric matrix of r . The
and out-of-plane librations of the tether, and according to rotation matrix from the LVLH frame to the body-fixed frame
Eq. (15), the inelastic tether tensional force, represented as is
F , should be rewritten as
  4ð1
r2 Þ 8
le GðrÞ ¼ I33
r þ ½r 2 ¼ Cob ð26Þ
 _
F ¼
M0 l b þ ðf þ aÞ
2 _ 2
_ cos b þ 3 ð3 cos a cos b
1Þ
2 2 2 ð1 þ r2 Þ2 ð1 þ r2 Þ2
R
ð19Þ
3. Gyros-based disturbance observer
where f is the true anomaly of the system. Substituting Eq. (19)
into Eq. (16), the tensional force for the inelastic tether can 3.1. Gyros-based observer
also be achieved as a vector within the LVLH frame. Peyman
and Hassan27 proposed an anti-sway control method to damp
As Eq. (22) shows, the attitude of the platform is always sub-
the attitude librations of main spacecraft, and the success of
jected to tether-generated perturbation, which adversely
their method depends on accurate measurements of tether ten-
degrades the attitude stabilization performance after the debris
sion and in-plane libration angles; however, if the measure-
capture. To stabilize the platform precisely, it is significantly
ments are badly corrupted by inherent noises, the control
important to develop a robust controller against the perturba-
performance will be significantly degraded.
tion. Although Eq. (22) reveals that the disturbance torque
could be obtained by measuring the vector of the tether ten-
2.3. Attitude dynamics of the platform sional force, necessary sensors for the measurement, such as
a tension sensor and a camera (measuring the liberation angle),
Next, the perturbed attitude dynamics of the platform will be will greatly complicate the system, and in addition, if measure-
derived by incorporating the tether-generated disturbance. ments are badly corrupted by inherent noises, the control per-
Generally, by considering the platform as a rigid body, its non- formance will be significantly degraded. In this section, a
linear attitude dynamics can be given as nonlinear disturbance observer (DO) will be developed to esti-
1790 G. ZHAI et al.

mate the disturbance acted on the platform, and in the context, turbance observer will be discussed based on Lyapunov stabil-
the proposed observer will be used to compensate for the dis- ity criterion. Define the positive definite matrix as
turbance torque by incorporating it into the control loop. The
1
basic disturbance observer is constructed as follows: P ¼ ðCJ
1 Þ ð36Þ

^_ d ¼
Wð^ud
ud Þ
u ð27Þ In general, it is not difficult to find a positive matrix Mp
that satisfies
where ^ud denotes the estimation of the lumped disturbance,
and W 2 R33 is the diagonal observer gain matrix. By defining P þ PT P Mp ð37Þ
the estimation error vector as e ¼ ^ud
ud , then the error
Then construct the Lyapunov candidate function as
dynamics of the observer can be written as follows:
e_ ¼
We
u_ d ð28Þ VðxÞ ¼ eT PT Pe ð38Þ

As Eq. (28) shows, when the disturbance varies slowly via Since both P and PT P are positive definite, thus the scalar
time, it is easy to find an appropriate W and guarantee the esti- function VðxÞ is positive definite. Taking the first order of
mation error to be bounded. Substituting Eq. (20) into Eq. derivative with respect to time yields
(27), the basic disturbance observer can be expressed as
_
VðxÞ ¼ e_ T PT Pe þ eT PT P_e ð39Þ
u_ d ¼
Wð^ud
Jx_
x Jx þ uc Þ
^ ð29Þ
For the time-varying disturbance, we suppose that the dis-
If x_ and x are both available, the basic disturbance obser- turbance rate is bounded as
ver represented in Eq. (29) could be utilized for practice. Gen-
erally, the angular velocity is accurately acquired by using 0 < ku_ d k 6 s ð40Þ
optical fibre gyros, while for the angular acceleration, there
where s is a constant real number. Eq. (30) is substituted into
is no available sensor that can be used directly for measure-
Eq. (39), and
ment; hence, the basic observer in Eq. (29) should be modified
h iT h i
_ _
before practical implementations. Aiming at eliminating the _
VðxÞ ¼ ^ _
n þ uðxÞ
u_ d PT Pe þ eT PT P ^ _
n þ uðxÞ
u_ d
angular acceleration term which appears in the basic observer,
an auxiliary variable, denoted as ^n, is introduced as ð41Þ

ud ¼ ^n þ uðxÞ
^ ð30Þ Further, with Eqs. (34) and (35), Eq. (41) can be rewritten
as
where uðxÞ is a vector function defined as
h 
iT
uðxÞ ¼ Cx ð31Þ _
VðxÞ ¼
^n þ x Jx
uðxÞ
uc þ x_ WT
u_ Td PT Pe
where C is a positive definite 3  3 matrix to be determined. n h i o
Substituting Eq. (30) into Eq. (29), the basic observer can be þ eT PT P W
^ n þ x Jx
uðxÞ
uc þ x_
u_ d
8 9
reformulated as > >
 < =
_
^ _
n ¼
uðxÞ
W ^n þ uðxÞ
Jx_
x Jx þ uc ð32Þ ¼ ^ n þ uðxÞ
½
x_
x Jx þ uc T WT
u_ Td PT Pe
:|fflfflfflfflffl{zfflfflfflfflffl}
> |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl} >
;
^ud ud
Substituting Eq. (31) into Eq. (32), then we have 8 2 3 9
>
< >
=
_ 6 7
^
n ¼
W^n þ ðWJ
CÞx_ þ Wx Jx
WuðxÞ
Wuc ð33Þ þ eT PT PW
4^ n þ uðxÞ
x_
x Jx þ uc 5
u_ d
: |fflfflfflfflffl{zfflfflfflfflffl}
> |fflfflfflfflfflfflfflfflfflfflffl
ffl {zfflfflfflfflfflfflfflfflfflfflffl
ffl } >
;
^ud ud
By examining the right-side terms, if the following condi- h i
tions hold: ¼
ð^ud
ud ÞT WT þ u_ Td PT Pe
W ¼ CJ
1 ð34Þ h i

eT PT P Wð^ ud
ud ÞT þ u_ Td
then the term associated with the angular acceleration can be T
canceled from the expression in Eq. (33), and the disturbance ¼
ðeWT þ u_ d Þ PT Pe
eT PT PðWe þ u_ d Þ
observer can be reconstructed as  
¼
eT WT PT P þ PT PW e
u_ Td PT Pe
eT PT Pu_ d
_
^
n ¼
W^n þ Wx Jx
WuðxÞ
Wuc ð35Þ ¼
eT ðP þ PT Þe
u_ Td PT Pe
eT PT Pu_ d
Eqs. (35) and (30) provide a type of disturbance observer that ð42Þ
only depends on x, which could be acquired by using gyros
Additionally, from Eq. (37), further yield
with high qualities, and moreover, from Eqs. (30) and (35), it
can be proven that the reduced-order observer has similar
eT ðP þ PT Þe 6
eT Mp e < 0 ð43Þ
error dynamics to those of the basic one.
On the other hand, by employing Rayleigh inequality, it
3.2. Observer stability shows that the Lyapunov candidate function expressed in
Eq. (38) is bounded as
As Eq. (28) shows, the error dynamics are always affected by kmin ðPT PÞkek2 6 VðxÞ 6 kmax ðPT PÞkek2 ð44Þ
the disturbance rate, so in the following, the stability of the dis-
Platform of a tethered space robot 1791

where kmin and kmax are the minimum and maximum eigenval- VðxÞ
pffiffiffiffiffiffiffi kmin ðPT Pkek2 6 VðxÞ ) kek2 6 ) kek
ues of PT P, respectively, and kek ¼ eT e denotes the 2-norm kmin ðPT PÞ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !
of the error vector. Note that
Vðx0 Þ
ð1
cÞkmin ðMp Þ
6 exp t
kmax ðPT PÞ ¼ kPT Pk 6 kPT kkPk ¼ kPk2 ð45Þ kmin ðPT PÞ 2kPk2
where k  k denotes the 2-norm of the matrix. With Eq. (44), 2skPk2
þ ðfor k > 0Þ ð57Þ
then we have kmin ðMp Þ
VðxÞ The inequality Eq. (57) reveals that in case of time-varying
VðxÞ 6 kPk2 kek2 ) kek2 P ð46Þ
kPk2 disturbance, the error converges with an exponential rate of
Applying Rayleigh inequality, the inequality Eq. (43) can
ð1
cÞkmin ðMp Þ=2kPk2 , and the steady error is bounded by
also be rewritten as the term 2skPk2 =kmin ðMp Þ . Considering the special case that
both C and J are diagonal matrices, it is simple to find that

eT ðP þ PT Þe 6
eT Mp e 6
kmin ðMp Þkek2 ð47Þ by increasing the eigenvalues of C, the steady-state boundary
Combining the inequality Eq. (46) with the inequality Eq. of the disturbance estimation will be decreased, and accord-
(47) leads to ingly, a more accurate estimation could be achieved.
kmin ðMp Þ

eT ðP þ PT Þe 6
eT Mp e 6
VðxÞ ð48Þ 4. Composite attitude controller
kPk2
With the disturbance boundary inequality Eq. (40), then 4.1. Structure of the composite attitude controller
applying the Schwartz inequality yields
T With the purpose of stabilizing the attitude libration of the
u_ Td PT Pe ¼ ðeT PT Pu_ d Þ 6 skPk2 kek ð49Þ
platform, a composite controller is proposed in this section.
Then combining Eq. (47) with Eq. (49), we have As shown in Fig. 2, the basic structure of the composite
controller (also named DO-based controller) consists of a
_
VðxÞ 6
kmin ðMp Þkek2 þ 2skPk2 kek ð50Þ feedback control loop and an observer-based feedforward
Define c as compensation loop. The feedback control loop, which involves
a sliding mode controller, works to stabilize the attitude
2skPk2 motion, while the feedforward compensation loop’s task is to
c¼ >0 ð51Þ
kmin ðMp Þkek reject the disturbance torque caused by the librating tether.
Then with Eq. (51), the inequality Eq. (50) can be reformu-
lated as 4.2. Sliding mode controller

_
VðxÞ 6
ð1
cÞkmin ðMp Þkek2 ð52Þ Sliding mode control is a type of variable structure control
Form the expression Eq. (46), if c > 1, method that alters the dynamics of a nonlinear system. Rather
than continuous controllers, a sliding mode controller switches
2skPk2 from one continuous structure to another based on the current
kek < ðif c > 1Þ ð53Þ
kmin ðMp Þ system state, with a discontinuous control force, and the
which means that the error is always bounded, while for the system could be guaranteed to slide within the vicinity of pre-
case of 0 6 c 6 1, scribed switching manifolds. The major advantage of sliding
mode control is low sensitivity to the uncertainties of system
2skPk2 parameters and external disturbances. Although the proposed
kek > ðif 0 6 c 6 1Þ ð54Þ gyros-based disturbance observer has been adopted for closed-
kmin ðMp Þ
loop control, still there is the residual disturbance, which is
With Eq. (46), then we have determined by the disturbance estimation error, that imposes
a negative effect on the control performance. Therefore, by
_ ð1
cÞkmin ðMp Þ
VðxÞ 6
VðxÞ 6 0 ðif 0 6 c 6 1Þ using sliding mode control, the attitude control performance
kPk2 could be further enhanced due to the robustness of a sliding
ð55Þ mode controller against the residual disturbance.
By using the MRPs, the sliding manifolds are given as
From Eq. (53) and Eq. (55), it can be confirmed that the
error is always bounded. Furthermore, from Eq. (55), we can s ¼ x
hðrÞ ð58Þ
also get T
! where s ¼ ½ s1 ; s2 ; s3  , while hðrÞ, the vector function in

ð1
cÞkmin ðMp Þ terms of r, is obtained by using a desired vector field from
VðxÞ 6 Vðx0 Þ exp t ðif 0 6 c 6 1Þ the attitude kinematic relations, which is given by
kPk2
ð56Þ hðrÞ ¼ G
1 ðrÞKðr
rd Þ ð59Þ

where t denotes time. Further, substituting Eq. (56) into Eq. where rd is the desired orientation, K is a diagonal matrix with
(44) and combining with Eq. (54) finally yield negative elements, and G
1 ðrÞ is calculated as follows:
1792 G. ZHAI et al.

Fig. 2 Illustration of a closed-control loop structure.

2  
G
1 ðrÞ ¼ 4ð1 þ rT rÞ ð1
rT rÞI33
2r þ 2rrT ð60Þ 5. Numerical results
Under a perfect control, the attitude of the platform moves
on the surface of s ¼ 033 . To enable decoupled sliding In this section, the effectiveness of the proposed composite
motions and exponential convergence toward the expected ori- control scheme is validated by series of numerical simulations.
entation, the sliding mode controller is constructed as The numerical model of a TSR after debris capture is con-

 structed within Matlab, and the gyros-based disturbance
dh observer and sliding mode controller are also incorporated
uc ¼
x Jx þ J ½GðrÞhðrÞ
GðrÞs þ Ks satðs; eÞ ð61Þ
dt into the control loop. For the purpose of comparison, the
where Ks is a 3  3 positive definite diagonal matrix, e is a control performance in absence of the disturbance observer,
small positive scalar, while satðs; eÞ represents the saturation so-called non-DO controller, is also investigated. Meanwhile,
used to minimize the chattering in the control input and is due to the disturbance estimation accuracy which significantly
defined by depends on the selected eigenvalues for the observer gain
8 matrix, the sensitivities on the variation of the observer gain
< 1 if si > e
> are also discussed based on numerical results. To facilitate
satðs; eÞ ¼ sei if jsi j > e ð62Þ the analysis on the observer performance, the uncertainties
>
:
0 if si <
e of gyros measurement are not considered. The entire simula-
tion takes one orbit resolution, and the time step is set to be
The attitude stabilization problem requires the final orien- 1s. The parameters associated with the simulations are listed
tation with respect to the LVLH frame to be zero, and then in Table 1.
the final desired orientation can be given as rd ¼ 0 . To sim- Like the longitudinal libration, the post-capture system is
plify the controller construction, the matrix K is assumed to supposed to experience both in- and out-of-plane librations
be given by
K ¼ diagðks ; ks ; ks Þ ð63Þ
Then Eq. (59) can be rewritten as Table 1 Simulation parameters.

2 Parameter Value
hðrÞ ¼ 4ks ð1 þ rT rÞ fð1
rT rÞI33
2½r  þ 2rrT gr ð64Þ
Orbit radius, R 6971 km
Eq. (64) can be further simplified significantly by taking Mass of debris, m 50 kg
some cross product operation, which leads to Tether attach distance, d 0.1 m
Tether length, l 100 m

1 Elastic modulus, E 1 GPa
hðrÞ ¼ 4ks ð1 þ rT rÞ r ð65Þ
Tether cross section, A 1  10
6 m2
Then taking the first order of derivative with Eq. (65) yields Tether damping ratio, c 0:2 m=s
h i Tether initial condition a0 ¼ b0 ¼ 0:1 rad; a_ 0 ¼ b_ 0 ¼ 0 rad=s
dh
1
1
¼ 4ks ð1 þ rT rÞ I33
2ð1 þ rT rÞ rrT ð66Þ Platform initial condition r0 ¼ ½0 0 0T ; x0 ¼ ½0 0 0 rad=s
dt J diag(112, 124, 85) kgm2
Substitute Eq. (66) and Eq. (65) into Eq. (61), and finally Observer initial condition n0 ¼ n_ 0 ¼ 0
the sliding mode controller can be completed. In addition, it Observer gain matrix, C diag(0.2, 0.2, 0.2)
is not difficult to prove the stability28 of the controller by ks
0.02 s
1
selecting a proper Lyapunov function. e 0.01
Platform of a tethered space robot
because of the non-zero relative velocity of the debris at the
capture moment. The initial librations are described by the
tether’s initial conditions listed in Table 1. It should be noted
that, because the out-of-plane libration is unstable, therefore,
during the simulation, a passive out-of-plane libration dam-
per is assumed to void the continuous tumbling of the
post-capture system, and both the libration angles and tether
tensional force are also treated as unknown terms. Only the
attitude angle velocities are measured by using gyros, and
no active control is acted on the captured debris and the
librating tether.
As shown in Fig. 3, the libration trajectory of the captured
debris is represented with respect to the LVLH frame, which
clearly shows that the libration always remains bounded
within the whole simulation, and the maximum libration
amplitudes on both x and y directions are 10 m and 40 m.
The in- and out-of-plane libration angles are also depicted
in Fig. 4.
Due to the dynamic coupling, the in- and out-of-plane
librations naturally lead to a longitudinal libration and gener-
ate a tensional force on the connecting tether. The tensional
forces for elastic and inelastic tethers are both presented in
Fig. 5. For the inelastic tether, the tensional force varies slowly
within a range from 3 N to 4 N, while for the elastic tether, the
tensional force dramatically oscillates around the tensional
values of the inelastic tether, but the oscillation is gradually
suppressed by the passive damping of the tether. The results
Fig. 3 Libration trajectory of the captured debris in the LVLH
frame.
Fig. 4 In- and out-of-plane libration angles.
1793
for the tether tensional force illustrate that by comparing with
the inelastic case, the disturbance generated by the elastic
tether becomes more serious regarding to high-frequency lon-
gitudinal libration. Although by suffering a longitudinal oscil-
lation, the tensional force remains positive and reveals that the
tether is always maintained taut during the libration. Fig. 6
shows the attitude motion of an uncontrolled platform; obvi-
ously, the platform becomes unstable under the disturbance
from the librating tether.
To attain the unknown disturbance torque acted on the
controlled platform, the proposed gyros-based observer is
used for disturbance estimation. The estimated disturbance
torque, as well as the modeled ones from Eq. (8), is simulta-
neously plotted in Fig. 7. Since the tether attach point is
located on the yaw axis of the body-fixed frame, the modeled
disturbance torque component along the yaw axis always
remains zero. From the results, one can see that the observer
response, which is depicted with a red dash line, tracks the
modeled disturbance accurately despite the initial short-term
oscillation on the yaw axis. Actually, the short-term oscilla-
tion corresponds with the convergence of the observer from
Fig. 5 Tensional force of the connecting tether.
Fig. 6 Uncontrolled attitude motion of the platform.
1794 G. ZHAI et al.

Fig. 7 History of disturbance estimation. Fig. 9 Control torques of the DO-based controller.

Fig. 8 Errors of disturbance estimation.

Fig. 10 Attitude control error of the platform.

the initial condition to the true value, and it can be regulated
by choosing different gain matrices and initial conditions for
the disturbance observer.
The disturbance estimation errors are shown in Fig. 8. One
can see that the error along the yaw axis nearly falls to zero,
while the ones along the pitch and roll axes gradually converge
to zero, but they are subjected to high-frequency oscillations.
Recalling the results from Fig. 7, in addition to the high-
frequency oscillation, the disturbance component along the
pitch axis also experiences long-term variation, which results
in a faster time-varying rate than that along the roll axis.
Finally, according to Eq. (40), if all the observer parameters
are fixed, the estimation error uniquely depends on the time-
varying rate of the disturbance; hence, the estimation error
along the pitch axis is greater than that along the roll axis.
In order to suppress the induced attitude libration from the
connecting tether, the DO-based controller, which involves the
disturbance observer within the control loop, is applied for
the platform. In fact, the control torques, as shown in Fig. 9,
represent the combination output of the disturbance compen-
sation and the sliding mode controller. From the result, one Fig. 11 Attitude control errors with different observer gain
can see that the control torques nearly equal to the disturbance matrices.
Platform of a tethered space robot
torque shown in Fig. 7, which reveals that the control effort is
mostly dedicated to the disturbance compensation.
The attitude control errors of the platform for both DO-
based control and non-DO-based control cases are represented
in Fig. 10, in which obviously, the results represent the most
important evidence that proves the benefits from the distur-
bance observer can be confirmed by incorporating the distur-
bance observer into the control loop, and finally, the control
precision for attitude stabilization is remarkably enhanced.
The negative effects from the tether-generated disturbance tor-
que are effectively migrated after introducing the disturbance
compensation, which illustrates the performance promotion
of the attitude control system of the platform.
The quality of disturbance estimation significantly affects
the DO-based control performance: the higher accuracy of dis-
turbance estimation is, the better performance can be achieved.
As analyzed earlier in Eq. (40), the eigenvalues of the observer
gain matrix affect the disturbance estimation errors inherently.
To validate the dependence of control accuracy on the
observer gain matrix, the results in Fig. 11 are dedicated to
illustrate the control errors with different observer gain matri-
ces; one can observe that with variations of the eigenvalues of
W from ki ¼ 0:2 to ki ¼ 0:1, the platform suffers degradation
on control accuracy. The results also indicate that for higher
stabilization accuracy expectation, greater eigenvalues should
be selected to construct the observer gain matrix.
6. Conclusions
A spacecraft with tether-captured debris is a typical nonlinear
system that is perturbed by a disturbance torque. Robustness
against the disturbance is an important property required for
platform attitude stabilization. With the purpose of improving
the attitude stabilization accuracy, a gyros-based disturbance
observer is proposed and incorporated into the control loop
to compensate the tether-generated disturbance torque. The
gyros-based disturbance observer only depends on the mea-
surements of attitude angular velocities. Theoretical and
numerical analysis indicates that compared with the case of
non-observer-based control, the disturbance observer-based
controller achieves remarkable performance promotion. More-
over, the dependence of control accuracy on the observer gain
matrix is also validated by numerical simulations.
Acknowledgements
The authors are grateful to the editors and anonymous review-
ers for their many constructive and helpful comments. This
work was supported by the National Natural Science Founda-
tion of China (11102018).
References
1. Zhang L, Jia QX, Chen G, Sun HX. Pre-impact trajectory
planning for minimizing base attitude disturbance in space
manipulator systems for a capture task. Chin J Aeronaut 2015;28
(4):1199–208.
2. Jin MH, Zhou C, Liu YC, Liu H. Analysis of reaction torque-
based control of a redundant free-floating space robot. Chin J
Aeronaut 2017;30(5):1765–76.
1795
3. Zhou C, Jin M, Liu Y, et al. Hybrid task priority-based motion
control of a redundant free-floating space robot. Chin J Aeronaut
2017;30(6):2024–33.
4. Xu WF, Meng DS, Chen YQ, Qian HH. Dynamics modeling and
analysis of a flexible-base space robot for capturing large flexible
spacecraft. Multibody Sys Dyn 2014;32(3):357–401.
5. Chu XY, Zhang JR, Lu S, Zhang Y, Sun Y. Optimised collision
avoidance for an ultra-close Rendezvous with a failed satellite
based on the Gauss pseudospectral method. Acta Astronaut
2016;128:363–76.
6. Mankala KK, Agrawal SK. Dynamic modeling and simulation of
impact in tether net gripper systems. Multibody Sys Dyn 2004;11
(3):235–50.
7. Huang PF, Zhang F, Ma J, Meng ZJ. Dynamics and configuration
control of the maneuvering-net space robot system. Adv Space Res
2015;55(4):1004–14.
8. Williams P. Dynamics and control of spinning tethers for
rendezvous in elliptic orbits. J Vib Control 2006;12(7):737–71.
9. Lorenzini EC. Error-tolerant technique for catching a spacecraft
with a spinning tether. J Vib Control 2004;10(10):1393–413.
10. Takeichi N, Natori MC, Okuizumi N, Higuchi K. Periodic
solutions and controls of tethered systems in elliptic orbits. J Vib
Control 2004;10(10):1473–91.
11. Wen H, Zhu ZH, Jin DP, Hu HY. Space tether deployment
control with explicit tension constraint and saturation function. J
Guid Control Dyn 2016;39(4):1–6.
12. Wen H, Jin DP, Hu HY. Three-dimensional deployment of
electro-dynamic tether via tension and current control with
constraints. Acta Astronaut 2016;129:253–9.
13. Wen H, Zhu ZH, Jin DP, Hu HY. Tension control of space tether
via online quasi-linearization iterations. Adv Space Res
2016;57:754–63.
14. Liu G, Huang J, Ma GF, Li CJ. Nonlinear dynamics and station-
keeping control of a rotating tethered satellite system in halo
orbits. Chin J Aeronaut 2013;26(5):1227–37.
15. Okazaki M, Ohtsuka T. Switching control for guaranteeing the
safety of a tethered satellite. J Guid Control Dyn 2006;29
(4):822–30.
16. Aslanov VS, Yudintsev VV. Dynamics of large debris connected
to space tug by a tether. J Guid Control Dyn 2013;36
(36):1654–60.
17. Cleary S, William JO. Control of space debris using an elastic
tether and wave-based control. J Guid Control Dyn 2016;39
(6):1–15.
18. Huang PF, Wang DK, Meng ZJ, Zhang F, Guo J. Adaptive
postcapture backstepping control for tumbling tethered space
robot-target combination. J Guid Control Dyn 2015;39(1):1–7.
19. Huang PF, Wang DK, Meng ZJ, Zhang F, Liu ZX. Impact
dynamic modelling and adaptive target capturing control for
tethered space robots with uncertainties. IEEE/ASME Trans
Mechatron 2016;21(5):2260–71.
20. Huang PF, Zhang F, Meng ZJ, Liu ZX. Adaptive control for
space debris removal with uncertain kinematics, dynamics and
states. Acta Astronaut 2016;128:416–30.
21. Meirovitch L, Sharony Y. Optimal vibration control of flexible
spacecraft during a minimum-time maneuver. J Opt Theor Appl
1991;69(1):31–54.
22. Yan RD, Wu Z. Attitude stabilization of flexible spacecrafts via
extended disturbance observer based controller. Acta Astronaut
2017;133:73–80.
23. Lee D. Nonlinear disturbance observer-based robust control of
attitude tracking of rigid spacecraft. Nonlinear Dyn Available
2017;88(2):1–12.
24. Liu H, Guo L, Zhang YM. An anti-disturbance PD control
scheme for attitude control and stabilization of flexible space-
crafts. Nonlinear Dyn 2011;67(3):2081–8.
1796
25. Chan LP, Naghdy F, Stirling D. Extended active observer for
force estimation and disturbance rejection of robotic manipula-
tors. Rob Auton Syst 2013;61(12):1277–87.
26. Mohammadi A, Tavakoli M, Marquez HJ, Hashemzadeh. Non-
linear disturbance observer design for robotic manipulators.
Control Eng Pract 2013;21(3):253–67.
G. ZHAI et al.
27. Peyman Y, Hassan S. Anti-sway control of tethered satellite
systems using attitude control of the main satellite. Acta Astronaut
2015;111:300–7.
28. Crassidis JL, Markley FL. Sliding mode control using modified
rodrigues parameters. J Guid Control Dyn 1996;19(6):1381–3.