Digital Signal Processing 46 (2015) 59–67
Contents lists available at ScienceDirect
Digital Signal Processing
www.elsevier.com/locate/dsp
H
∞
filtering for networked linear systems with multiple packet
dropouts and random delays
Xiu-Ying Li, Shu-Li Sun
Department of Automation, Heilongjiang University, Harbin 150080, PR China
a r t i c l e i n f o a b s t r a c t
Article history:
Available
online 4 August 2015
Keywords:
H
∞
filtering
Packet
dropout
Random
delay
Networked
system
Linear
matrix inequality
This paper is concerned with an H
∞
filtering problem for the network-based linear systems with multiple
packet dropouts and random delays. Due to the limited bandwidths of communication channels, the
measured outputs will be delayed or even be lost during the transmission from the sensor to the
remote filter, where the delays are randomly varying but bounded and packet dropouts are possibly
consecutive. Based on a recent developed model that describes the phenomena of packet dropouts
and time delays simultaneously, an H
∞
filter is designed to ensure the filtering error system to be
mean-square exponentially stable and guarantee a prescribed H
∞
filtering performance. Asufficient
condition for the existence of such a filter is provided via a linear matrix inequality (LMI) method. The
effectiveness and applicability of the proposed algorithm is demonstrated by a practical F-404 aircraft
engine system.
© 2015 Elsevier Inc. All rights reserved.
1. Introduction
With the rapid development of computer and communication
technology, networks have been widely used as the medium in
modern engineering systems to connect the spatially distributed
sensors, actuators and controllers or filters. Such systems are the
so-called networked control systems (NCSs) which have many ad-
vantages,
such as low cost, simple installation and maintenance,
convenient system diagnosis and increased system agility [1]. How-
ever,
due to the limited bandwidths of the communication chan-
nels,
the network-induced delays are inevitable during the data
transmission through the networks from the sender to the re-
ceiver [2,3].
A transmission delay may be less or larger than one
sampling period which is the so-called short time-delay or large
time-delay. For example, the TCP/IP communication protocol has
the large communication delays but the UDP/IP has generally the
short communication delays [4]. The network-induced delays have
the random characteristic in nature but are bounded. If the packet
is with a delay longer than a certain pre-determined number, one
possible strategy is to discard the packet and treat it as a packet
dropout, which could deteriorate the system performance or even
lead to the instability. So the random delays and packet dropouts
are the two important issues which have attracted considerable re-
search
attention in the NCSs.
E-mail address: sunsl@hlju.edu.cn (S.-L. Sun).
The filtering problem for networked systems has been a focus
of research due to their important engineering applications such
as target tracking, signal processing and control application [5–8].
It is well known that the Kalman filtering is the classical scheme
to deal with the estimation problem effectively. In a network en-
vironment,
the Kalman filtering should be modified to conduct
the random phenomena [9–12]. However, one primary limitation
of Kalman filtering is that the external disturbances are required
to be Gaussian noises with known statistical property. Such a re-
quirement
is not always satisfied in practical applications. For this
case, the H
∞
method is taken as an alternative method. A great
number of important results for the H
∞
control problems have
been reported [13–16]. As for H
∞
filtering problem, the objec-
tive
is to minimize the highest energy gain of the estimation error
for all initial conditions and noises where the noise signals are
assumed to be arbitrary but with bounded energy or bounded av-
erage
power rather than just Gaussian. Hence, the H
∞
filtering
problem of networked systems has also received extensive research
attention [17–22].
Due
to the random nature of transmission delays and packet
dropouts, they can be described by Bernoulli distributed white se-
quences
or Markov chains [23–25]. As for the first type, a set of
Bernoulli distributed random variables are used to establish the
NCSs with multiple packet dropouts and random transmission de-
lays
simultaneously [26,27]. However, in [26], the model deals with
the networks with retransmission mechanism which may cause
the network congestion. Anew model is established to describe
the phenomena of multiple random delays and packet dropouts
http://dx.doi.org/10.1016/j.dsp.2015.07.008
1051-2004/
© 2015 Elsevier Inc. All rights reserved.