Robust learning observer-based actuator fault
reconstruction for satellite attitude control systems
Qingxian Jia
∗
, Huayi Li
∗†
, Xueqin Chen
∗†
, Yingchun Zhang
∗‡
∗
Research Center of Satellite Technology
Harbin Institute of Technology Harbin, China 150001
Email: jqxhit@gmail.com
†
National & Local United Engineering Research Center
of Small Satellite Technology Changchun, China 130033
Email: lihuayi@hit.edu.cn, cxqhit@hit.edu.cn
‡
Shenzhen Aerospace Dongfanghong HIT
Satellite Company Ltd Shenzhen, China 518057
Email: zhang@hit.edu.cn
Abstract—In this paper, robust actuator fault reconstruction
for satellite attitude control systems (ACSs) in the presence
of space environmental disturbances and actuator faults is
investigated via a learning observer (LO). First, a nonlinear
mathematical model of the satellite ACS satisfying Lipschitz
constraint is established; then an LO is constructed to achieve
robust reconstruction of satellite attitude angles, attitude angler
velocities and actuator faults at the same time. Further, sufficient
conditions for the robust stability of the LO are explicitly
provided in this paper. The design problem of the proposed
LO is formulated into a convex optimization problem in terms
of linear matrix inequalities (LMIs) that can be directly solved
using the LMI toolbox of MATLAB. At last, simulation studies
on a satellite example are performed to illustrate the effectiveness
and applicability of the presented LO-based fault-reconstructing
strategy.
I. INTRODUCTION
An attitude control system (ACS), as an indispensable
subsystem in satellites, becomes increasingly sophisticated due
to the diversity of space missions and the growing requirements
on high reliability and performance. During on-orbit mission
operations, this subsystem unavoidably manifests various cat-
egories of of faults despite the fact that the satellites with
high quality and reliability has been designed before launched.
On the other hand, satellites run in a harsh outer space
environment with hazardous radiation and huge temperature
variations, and it also increases the possibility of faults in ACSs
to a certain extent. As is known, the accuracy and reliability of
the ACS make a direct contribution to the success of satellite
missions [1]. The faults in the ACS may dramatically degrade
control performance and even result in loss of control of the
satellite. As a response to the high requirements on safety and
reliability of the satellite, fault diagnosis for satellite ACS has
become an attractive research subject of many investigations
and several recent excellent results can be found in [1]–[6].
Generally speaking, the basic fault-diagnostic module is
divided into three steps: fault detection, isolation and re-
construction. Compared with the extensive research on fault
detection and isolation (FDI) strategies for satellite ACSs,
fault reconstruction for this subsystem in the satellites has
received little attention during the past two decades. As an
essential component in active fault-tolerant control system for
satellite ACSs, fault reconstruction for this system has been
attracting much attention of researchers in recent years, and
great progresses have been published. Among them, many fault
reconstruction observers have been developed for satellite AC-
Ss, sliding mode observers [5], [7], adaptive observers (AOs)
[8], [9], descriptor observers [10], [11], learning observers
(LOs) [6], [12], [13], just to name only a few.
Inspired with our previous work on LOs [12], [13], LO-
based robust actuator fault reconstruction for satellite ACSs
is investigated in this paper. First, considering space environ-
mental disturbances and actuator faults, a nonlinear mathe-
matical model of satellite ACS with Lipschitz constraint is
established; then a nonlinear LO is constructed to achieve
robust reconstruction of satellite attitude angles, attitude angler
velocities and actuator faults at the same time. Further, the
robust stability of the proposed LO is proved using Lyapunov
indirect method together with 𝐻
∞
techniques. In addition,
observer gain matrices can be conveniently computed using
the linear matrix inequality (LMI) toolbox of Matlab. Finally,
simulation results on a satellite ACS demonstrate the proposed
LO can not only accurately reconstruct actuator faults, but
also is superior to the conventional AO involving the pure
integration [15].
II. M
ATHEMATICAL MODELLING OF SATELLITE ATTITUDE
CONTROL SYSTEMS
In this work, the rigid-body satellite is considered and its
nonlinear attitude dynamics is characterized as
𝐽
𝑠
˙𝜔 + 𝜔
×
𝐽
𝑠
𝜔 = 𝑢
𝑐
+ 𝑢
𝑑
(1)
where 𝐽
𝑠
= diag{𝐽
𝑥
,𝐽
𝑦
,𝐽
𝑧
}, 𝐽
𝑥
, 𝐽
𝑦
and 𝐽
𝑧
are the in-
ertia moments of the satellite along principal axes; 𝜔 =
[
𝜔
𝑥
𝜔
𝑦
𝜔
𝑧
]
𝑇
, 𝜔
𝑥
, 𝜔
𝑦
and 𝜔
𝑧
represent the satellite an-
gular velocities; 𝑢
𝑐
is the control torques along principal axes
and they are provided by the actuators mounted on the satellite.
Parameter 𝑢
𝑑
is space disturbance torque. The term 𝜔
×
is a
556
2014 International Conference on Mechatronics and Control (ICMC)
July 3 - 5, 2014, Jinzhou, China
978-1-4799-2538-4/14/$31.00 ©2014 IEEE