Published in IET Signal Processing
Received on 21st June 2013
Revised on 26th December 2013
Accepted on 8th January 2014
doi: 10.1049/iet-spr.2013.0431
ISSN 1751-9675
Linear minimum-mean-square error estimation of
Markovian jump linear systems with randomly
delayed measurements
Yanbo Yang
1,2
, Yan Liang
1,2
, Feng Yang
1,2
, Yuemei Qin
1,2
, Quan Pan
1,2
1
School of Automation, Northwestern Polytechnical University, Xi’an, People’s Republic of China
2
Key Laboratory of Information Fusion Technology, Ministry of Education, People’s Republic of China
E-mail: liangyan@nwpu.edu.cn
Abstract: This study presents the state estimation problem of discrete-time Markovian jump linear systems with randomly
delayed measurements. Here, the delay is modelled as the combination of different number of binary stochastic variables
according to the different possible delay steps. In the actually delayed measurement equation, multiple adjacent step
measurement noises are correlated. Owing to the stochastic property from the measurement delay, the estimation model is
rewritten as a discrete-time system with stochastic parameters and augmented state reconstructed from all modes with their
mode uncertainties. For this system, a novel linear minimum-mean-square error (LMMSE, renamed as LMRDE) estimator for
the augmented state is derived in a recursive structure according to the orthogonality principle under a generalised framework.
Since the correlation among multiple adjacent step noises in the measurement equation, the measurement noises and related
second moment matrices of corresponding previous instants in each current step are also needed to be estimated or calculated.
A numerical example with possibly delayed measurements is simulated to testify the proposed method.
1 Introduction
Considerable research has been undertaken in the field of
estimation theory in relation to the discrete-time Markovian
jump systems, such as non-linear filtering [1–5], target
tracking [6, 7] and fault-tolerant robust filtering [8].
By the fact that the recursive computation of the first two
moments of the interested vector sometimes is sufficient in
practice, the linear minimum-mean-square error (LMMSE)
estimator for Markovian jump linear systems (MJLSs) has
been paid much attention. In [9], the LMMSE estimator
was derived from geometric augments through estimating
x
k
1
Q
k
=i
{}
instead of estimating directly the original state x
k
,
where 1
{·}
is a Dirac function and Θ
k
is a discrete-time
Markov chain. One desirable advantage is that the gain
matrices of the resultant LMMSE estimator are not
data-dependent, and hence can be calculated off-line. In
[10], it was shown that under the assumption of mean
square stability of the MJLS and the ergodicity of the
associated Markov chain, the error covariance matrix will
converge to the unique positive-semi-definite solution of an
Nn-dimensional algebraic Riccati equation, where n is the
dimension of the state vector and N is the number of states
of the Markov chain. Furthermore, a time-invariant
LMMSE estimator was derived. By the fact that roundoff
errors in solving the above Riccati equation can cause the
loss of the covariance of the state prediction error, an array
algorithm with the better numerical robustness was
developed [11]. Moreover, the information filter and its
corresponding array algorithm [12] were proposed to deal
with multi-sensor measurements sequentially. In general, no
measurement delay is considered in the above estimators.
In networked control systems, the sensors may be far away
from the estimator/controller and random measurement delay
because of data transmission and relay or missing is definitely
inevitable. Many important results have been obtained
including estimation with missing data [13–16] and robust
control [17, 18]. In [19], a stochastic extended Kalman filter
algorithm is presented to provide optimal estimates of
interconnected network states for systems with one-step
randomly delayed measurements, where some or all
measurements are delayed. In [20], the discrete-time system
with the measurements delayed less than l + 1 steps was
considered and the derived estimator involves solving l +1
distinct Kalman filters with the same dimension as the
original system. In [21], the least squares filtering problem
was investigated for a class of non-linear discrete-time
stochastic systems using observations with stochastic
one-step delays modelled by a Bernoulli sequence, and the
extended and unscented Kalman filters were then proposed.
In [22], through presenting Gaussian approximation about
the one-step posterior predictive probability density
functions of the state and delayed measurement, a novel
Gaussian approximation filter was derived for a class of
non-linear stochastic systems with one-step random
Bernoulli delay of measurements. An unscented filtering
algorithm was derived in [23] for a class of non-linear
discrete-time stochastic systems using noisy observations
www.ietdl.org
658
&
The Institution of Engineering and Technology 2014
IET Signal Process., 2014, Vol. 8, Iss. 6, pp. 658–667
doi: 10.1049/iet-spr.2013.0431