International Journal of Control, Automation, and Systems (2012) 10(5):1042-1048
DOI 10.1007/s12555-012-0522-2
ISSN:1598-6446
eISSN:2005-4092
http://www.springer.com/12555
A Probabilistic Approach for Model Following of Markovian Jump Linear
Systems Subject to Actuator Saturation
Linpeng Wang, Jin Zhu*, and Junhong Park
Abstract: This paper is concerned with the model following problem of Markovian jump linear sys-
tems (MJLSs), which suffer from stochastic uncertainties and actuator saturation. By applying a proba-
bilistic approach based on particles, a sequence of control inputs is designed to guarantee that the mod-
el following error remains within a desired region in a certain probability, as well as the control cost is
optimal. Motivated by this, the stochastic control problem is represented by chance constrained pro-
gramming, and approximated as a determinate optimization one, which is solved by mixed integer li-
near programming (MILP). Furthermore, an improved particle control approach is proposed to reduce
the computation complexity. The effectiveness of this improved approach is demonstrated by an exam-
ple along with complexity comparison.
Keywords: Actuator saturation, Markovian jump linear systems, model following, particle control
approach.
1. INTRODUCTION
The problems of model following with uncertainties
have received a great deal of attention over the past
decades. Up to date, much work has been done in this
field with lots of achievements. For instance, robust
tracking and model following problem is discussed in [1]
and a linear controller is designed to track dynamic
signals for uncertain linear systems; a linear memoryless
controller is developed in [2] to track reference model for
uncertain linear time-delay systems such that the tracking
error can be made arbitrarily small; [3] considers the
problem of robust tracking and model following for linear
systems with multiple delayed state perturbations, time-
varying uncertain parameters and disturbance; while [4]
proposes a discrete-time integral sliding mode control
scheme to achieve zero-tracking error for robust tracking
and model following of uncertain linear systems. All these
literature helps a deep understanding of model following
problems. However, the practical systems are getting
more and more complex recently and far from ideal
assumptions. Thus current research interests are focusing
on how to model the system dynamics accurately.
The practical systems of model following are affected
by many factors. Firstly, system dynamics, which is
usually driven by only continuous time, will vary with
some random discrete events occurring, resulting the
system states are now driven by both continuous time
and discrete events. Next, system dynamics usually
suffers from uncertainties arising from modeling error,
external disturbances or a combination of these. Such
uncertainties, which are best represented by a stochastic
model instead of a set-bounded one, can be of any
distributions rather than Gaussian. Thirdly, actuator
saturation or input constraint, is another important factor
for system dynamics if physical limitation is considered.
The performance of dynamic systems is degraded by
actuator saturation, even the systems are unstable due to
this phenomenon [5,6]. Considering above factors, the
development of theory along with technology is facing
big challenges for model following problems. Motivated
by this, Markovian jump linear system (MJLS) model
with stochastic uncertainties and actuator saturation is
used in this paper to describe such dynamics, where the
mentioned discrete events are governed by a Markov
chain [7]. Differing from the existed achievements about
model following of MJLS, the distribution of uncertainty
herein is of arbitrary rather than Gaussian such that jump
linear quadratic Gaussian (JLQG) method will be invalid
in this case [8]. All these practical considerations mean
that a new approach should be developed to deal with the
complex situations.
In this paper, a probabilistic approach for model
following of MJLS subject to stochastic uncertainties and
actuator saturation is proposed. This probabilistic ap-
proach, roughly speaking, is to approximate the stochas-
tic problem as a determinate one using particles, and the
possible sample paths can be represented by particles
which are generated according to the distributions of
© ICROS, KIEE and Springer 2012
________
Manuscript received May 22, 2011; revised March 9, 2012;
accepted June 4, 2012. Recommended by Editorial Board membe
Shengyuan Xu under the direction of Editor Myotaeg Lim.
This work is supported by the National Natural Science Foun-
dation of China (Grants no. 60904021) and the Fundamental Re-
search Funds for the Central Universities (Grants no. WK210006
0004).
Linpeng Wang and Jin Zhu are with the Department of Auto-
mation, University of Science and Technology of China, 230027,
Hefei, Anhui, China (e-mails: wlp816@mail.ustc.edu.cn, jinzhu@
ustc.edu.cn).
Junhong Park is with the Department of Mechanical Engineer-
ing, Hanyang University, Seoul, 133-791, Korea (e-mail: parkj@
hanyang.ac.kr).
* Corresponding author.