Published in IET Signal Processing
Received on 6th October 2009
Revised on 12th June 2010
doi: 10.1049/iet-spr.2009.0244
ISSN 1751-9675
Robust set-membership filtering for systems
with missing measurement: a linear matrix
inequality approach
F. Yang
1,2
Y. Li
3
1
School of Information Science and Engineering, East China University of Science and Technology, Shanghai 200237,
People’s Republic of China
2
Centre for Intelligent and Networked Systems, Central Queensland University, Rockhampton, QLD 4702, Australia
3
Department of Information Systems and Computing, Brunel University, Uxbridge, Middlesex UB8 3PH, UK
E-mail: fwyang@ecust.edu.cn
Abstract: This study addresses the robust set-membership finite-horizon filtering problem for a class of discrete time-varying
systems with missing measurement and polytopic uncertainties in the presence of unknown-but-bounded process and
measurement noises. A robust set-membership filter is developed and a recursive algorithm is derived for computing the state
estimate ellipsoid that is guaranteed to contain the true state. An optimal possible estimate set is computed recursively by
solving the semi-definite programming problem. Simulation results are provided to demonstrate the effectiveness of the
proposed method.
1 Introduction
Filtering plays an important role in signal processing [1].Itis
well-known that the Kalman filter requires the process and
measurement noises to be white Gaussian processes [2].
However, the Kalman filter may lead to poor performanc e
for non-Gaussian noises [3]. Recently, the H
1
filtering
method has been proposed, which provides an energy
bounded gain for the worst-case estimation error without
the need for knowledge of noise statistics [4]. In this
filtering, process and measurement noises are assumed to be
arbitrary rather than Gaussian processes. However, there is
no provision in H
1
filtering to ensure that the variance of
the state estimation error lies within acceptable bounds [5].
In this respect, it is natural to consider process and
measurement noises as unknown-but-bounded which belong
to given sets in appropriate vector spaces [6, 7]. All
possible state estimates can be characterised by the set of
state estimates consistent with both the measurements
received and the constraints on the unknown process and
measurement noises whose norms are less than the
prescribed scalars, and the true state is contained in this set
of state estimates. The actual estimate thus is a set in state
space rather than a single vector. This estimation problem
has been referred to as a set-membership (set-valued)
filtering problem [6, 8 –12].
The set-membership filtering problem was first considered
by Witsenhausen [13]. The set of all possible values of
the states compatible with the measurement of outputs is
completely characterised by their support functions. An
ellipsoidal approximation algorithm was provided by
Schweppe [7]. In this algorithm, the measurements are used
to calculate recursively a bounding ellipsoid to the set of
possible states, under the assumption that the sets
containing the initial condition and the process and
measurement noises are, or can be approximated, by
ellipsoids. The solution to a set-membership filtering
problem with the instantaneous constraints was determined
by using the results derived for an energy constraint in [8].
The resulting estimator is similar to that proposed by
Schweppe [7]; however it has an important advantage that
the gain matrix does not depend on the particular output
measurements and is therefore precomputable. Recently,
many researchers have attempted to deal with the set-
membership filtering problems for various uncertain
systems. For example, a convex optimisation approach has
been applied to the case of norm-bounded uncertainty in
the system matrices to provide a set of state estimates in
[14]. A combinational ellipsoidal constraint of the uncertain
system matrix and uncertain process noise was introduced
for set-membership filtering in [15]. A recursive scheme for
constructing an ellipsoidal state estimation set of all states
consistent with the measured output and the given noise
and unstructured uncertainty described by a sum quadratic
constraint was presented in [16 –18]. In the existing
literature concerning set-membership filtering techniques, it
is implicitly assumed that the measurements always contain
consecutive useful signals; see for example [19– 21].In
many real-world applications, however, the measurements
are not consecutive but contain missing measurements. For
example, the missing measurements exist in signal shading
and blocking, intermittent sensor failure, network
IET Signal Process., 2012, Vol. 6, Iss. 4, pp. 341 – 347 341
doi: 10.1049/iet-spr.2009.0244
&
The Institution of Engineering and Technology 2012
www.ietdl.org