Information Fusion 82 (2022) 19–27

Contents lists available at ScienceDirect

Information Fusion
journal homepage: www.elsevier.com/locate/inffus

Multi-sensor multi-rate fusion estimation for networked systems: Advances

and perspectives✩
Yuxuan Shen a,b , Zidong Wang c , Hongli Dong a,b , Hongjian Liu d,e ,∗
a
Artificial Intelligence Energy Research Institute, Northeast Petroleum University, Daqing 163318, China
b
Heilongjiang Provincial Key Laboratory of Networking and Intelligent Control, Northeast Petroleum University, Daqing 163318, China
c
Department of Computer Science, Brunel University London, Uxbridge, Middlesex, UB8 3PH, United Kingdom
d
Key Laboratory of Advanced Perception and Intelligent Control of High-end Equipment, Ministry of Education, Anhui Polytechnic University, Wuhu 241000, China
e
School of Mathematics and Physics, Anhui Polytechnic University, Wuhu 241000, China

ARTICLE INFO ABSTRACT

Keywords: In industrial systems, the multi-rate sampling strategy has been widely used due to the advantage in balancing
Multi-rate systems cost and performance as well as the psychical characteristics of the hardware. Accordingly, the analysis and
Networked systems synthesis problems of the multi-rate systems (MRSs) have received considerable research attentions owing
H ∞ state estimation
to the significant engineering background. Among others, the state estimation problem, aims to estimate the
Kalman filtering
system state based on the contaminated measurement signals, is one of the most important topics in the area
Multi-sensor information fusion
of signal processing. In the past decades, plenty of research results have been obtained on the state estimation
problems for MRSs. The intent of this survey is to provide a timely and systematic review with respect to
the available state estimation algorithms for networked MRSs and the corresponding fusion methods. First, a
general state–space model of the MRSs is given and the methods that transform the MRSs into single-rate ones
are introduced. Then, the recent advances on the state estimation as well as fusion estimation problems for
MRSs are discussed based on the performance indices used. Finally, some future research topics are given in
the MRS state estimation problems.

1. Introduction systems. Consequently, the problem of state estimation for networked

systems has been attracting ever-increasing research interest [8–13].
With the rapid development of the wireless communication tech- Due to the large scale, it is often the case that the networked system
nology and the digital technology, the networked systems have been contains numerous spatially distributed system components (including
widely applied in the practical applications such as process monitor- multiple sensors). As mentioned before, the continuous signals from the
ing, power grids, industrial control systems, traffic systems, and etc. system components are first sampled into discrete signals with fixed
[1–7]. In the networked systems, the signals generated by the sys- sampling periods before being transmitted. Traditionally, to simplify
tem components, including the sensors, the underlying plant, and the the design of the state estimation algorithms, it is assumed that a
controllers, are first sampled and then transmitted through the commu- unified sampling period is chosen for different signals [14–18]. Un-
nication networks. Compared to the traditional systems, the networked fortunately, such an assumption is not realistic in real practice. Since
systems provide the advantages of low cost, easy maintenance, and different system components own different physical characteristics, it
high reliability. Nevertheless, the introduction of the communication is quite difficult (if not impossible) to unify the sampling periods for
network largely increases the complexity of the networked systems, and different components. Furthermore, setting different sampling periods
hence brings major challenges to the state estimation of the networked for different system components (i.e., multi-rate sampling) according

✩ This work was supported in part by the National Natural Science Foundation of China under Grants 61933007, U21A2019, 62103095, 61873058 and
61873148, the AHPU Youth Top-notch Talent Support Program of China under Grant 2018BJRC009, the Natural Science Foundation of Anhui Province of China
under Grant 2108085MA07, the Natural Science Foundation of Heilongjiang Province of China under Grant LH2021F005, the Royal Society of the UK, and the
Alexander von Humboldt Foundation of Germany.
∗ Corresponding author at: Key Laboratory of Advanced Perception and Intelligent Control of High-end Equipment, Ministry of Education, Anhui Polytechnic
University, Wuhu 241000, China.
E-mail addresses: [email protected] (Y. Shen), [email protected] (Z. Wang), [email protected] (H. Dong), [email protected]
(H. Liu).

https://doi.org/10.1016/j.inffus.2021.12.005
Received 6 October 2021; Received in revised form 20 November 2021; Accepted 9 December 2021
Available online 1 January 2022
1566-2535/© 2021 Published by Elsevier B.V.
Y. Shen et al. Information Fusion 82 (2022) 19–27

Fig. 1. The architecture of the centralized fusion process.

Fig. 2. The architecture of the distributed fusion process.

to the importance of their signals is preferable in engineering practice

[19–22]. As such, the state estimation problems for networked multi- Generally speaking, in the existing literature, there are mainly two
rate systems (MRSs), which reflect the real situation in the practical ways to deal with the network-induced challenges in the state esti-
engineering, have stirred attentions form researchers. mation problems. The first one is to passively design state estimation
In the state estimation problems for MRSs, due to the existence of algorithms that are robust to the network-induced phenomena. That
the multiple time sequences, it is quite difficult to directly design state is, with the existence of the network-induced phenomena, designing
estimators for the considered MRSs. In other words, the state estimation state estimation algorithms such that the desired performance require-
methods developed for single-rate systems (SRSs) cannot be directly ment is achieved, see for examples [53–56]. The other way is to
applied to MRSs. Therefore, much research effort has been devoted actively introduce communication protocols into the communication
to solving state estimation problems for MRSs [23–32]. In literature, networks to alleviate/avoid the network-induced phenomena, and then
a widely accepted way to deal with the state estimation problems develop state estimation algorithms whilst considering the influence
for MRSs is that we first transmit the MRSs into equivalent SRSs, of the introduced communication protocols. Some of the widely used
and then design state estimators for the equivalent SRSs by using the communication protocols are the static event-triggered protocol [57–
renowned state estimation methods such as the 𝐻∞ estimation method, 60], the dynamic event-triggered protocol [61–64], the self-triggered
the Kalman filtering method, and the set-membership filtering method. protocol [65,66], the Round-Robin protocol [67,68], the weighted
Once the estimate of the state of the transmitted SRSs is obtained, the Try-Once-Discard protocol [69,70], and the random access protocol
estimate of the state of the MRSs can be easily obtained by some simple [71–74].
matrix operations. In this paper, we aim to provide a systematic review of the existing
The multiple sensors in the networked systems provide comple- results on the state estimation problems for networked MRSs and the
mentary or redundant information about the plant of interest. To applied multi-sensor information fusion methods. The remainder of
achieve an accurate estimation performance, it is suggested to utilize this paper is outlined as follows. In Section 2, a general state–space
the multi-sensor information fusion methods to fuse the information model of the MRSs is presented and various methods that transform
from the multiple sensors. In general, the multi-sensor information the MRSs into SRSs are surveyed. In Section 3, research results are
fusion methods can be categorized into two types: the centralized fusion discussed on the state estimation problems for MRSs where the methods
methods and the distributed fusion methods [33–37]. In the central- in Section 2 are used to transmit the MRSs. Moreover, the multi-
ized fusion methods, the measurements from the multiple sensors are sensor information fusion methods applied are also reviewed. Section 4
transmitted to a fusion center where the measurements are fused to a provides the conclusion and the future research topics.
new measurement. Then, the system state is estimated based on the
obtained new measurement. In the distributed fusion methods, each 2. Multi-rate systems
local estimator first gives a local estimate based on its measurement and
then transmits the estimate to a fusion center where the local estimates In this section, we first introduce a general state–space model of the
are fused to an optimal/suboptimal global estimate. The architectures discrete-time MRSs. Then, some effective methods that transform the
of the centralized fusion process and the distributed fusion process are MRSs into SRSs are presented. Moreover, the corresponding estimators
shown in Figs. 1 and 2, respectively. are designed for the transformed SRSs.
On the other hand, due to the limited communication resources
of the communication networks, some network-induced phenomena 2.1. State–space model of the MRSs
arise inevitably in the networked systems. Such network-induced chal-
lenges include, but are not limited to, the missing measurements, the Consider a discrete-time multi-sensor MRS described as follows:
{
communication delays, the channel fading, and the packet disorders. 𝑥(𝑠𝑘+1 ) = 𝐴(𝑠𝑘 )𝑥(𝑠𝑘 ) + 𝐵(𝑠𝑘 )𝑤(𝑠𝑘 ),
If not properly handled, these network-induced phenomena would (1)
𝑦𝑖 (𝑡𝑖𝑘 ) = 𝐶𝑖 (𝑡𝑖𝑘 )𝑥(𝑡𝑖𝑘 ) + 𝐷𝑖 (𝑡𝑖𝑘 )𝑣𝑖 (𝑡𝑖𝑘 )
deteriorate the estimation performance in the state estimation of the
networked systems. Therefore, it is of great importance to find effective where 𝑥(𝑠𝑘 ) ∈ R𝑛𝑥 and 𝑦𝑖 (𝑡𝑖𝑘 ) ∈ R𝑛𝑦 (𝑖 ∈ [1, 𝑁]) are the system state and
solutions to handle the above mentioned network-induced problems. the measurement output from the 𝑖th sensor, respectively. 𝑤(𝑠𝑘 ) and
Up to now, plenty of results have been obtained for state estimation 𝑣𝑖 (𝑡𝑖𝑘 ) are the process noise and the measurement noise on the 𝑖th sensor,
problems for networked systems under network-induced phenomena, respectively. 𝐴(𝑠𝑘 ), 𝐵(𝑠𝑘 ), 𝐶𝑖 (𝑡𝑖𝑘 ), and 𝐷𝑖 (𝑡𝑖𝑘 ) are constant matrices with
see [38–41] for missing measurements, see [42–45] for communication compatible dimensions.
delays, see [46–49] for channel fading, and see [50–52] for packet The state update period of the system (1) is ℎ ≜ 𝑠𝑘+1 − 𝑠𝑘 and
disorders. the sampling period of the 𝑖th sensor is 𝑏𝑖 ℎ ≜ 𝑡𝑖𝑘+1 − 𝑡𝑖𝑘 where 𝑏𝑖 ≥ 1

20
Y. Shen et al. Information Fusion 82 (2022) 19–27

𝐵̄ 𝑖 (𝑡𝑖𝑘 ) ≜ col{𝐵̃ 1𝑖 (𝑡𝑖𝑘 ), 𝐵̃ 2𝑖 (𝑡𝑖𝑘 ), … , 𝐵̃ 𝑏𝑖 (𝑡𝑖𝑘 )},

𝑖
[ ]
𝐶̄𝑖 (𝑡𝑖𝑘 ) ≜ 0 ⋯ 0 𝐶𝑖 (𝑡𝑖𝑘 ) .

Following the similar process, the MRS (1) is transformed into 𝑁

SRSs with uniform periods 𝑏𝑖 ℎ (𝑖 = 1, 2, … , 𝑁). For the 𝑖th SRS, the 𝑖th
local estimator is designed as

𝑥̂ 𝑖 (𝑡𝑖𝑘+1 ) = 𝐴̄ 𝑖 (𝑡𝑖𝑘 )𝑥̂ 𝑖 (𝑡𝑖𝑘 ) + 𝐾𝑖 (𝑡𝑖𝑘 )(𝑦𝑖 (𝑡𝑖𝑘 ) − 𝐶̄𝑖 (𝑡𝑖𝑘 )𝑥̂ 𝑖 (𝑡𝑖𝑘 ))

where 𝑥̂ 𝑖 (𝑡𝑖𝑘 ) is the estimate of 𝑥̄ 𝑖 (𝑡𝑖𝑘 ) and 𝐾𝑖 (𝑡𝑖𝑘 ) is the estimator gain to
be designed.
From the above process, we can see that the lifting technique is
able to transform the linear MRSs to equivalent linear SRSs. Neverthe-
less, due to the augmentation of the states, the designed estimation
algorithm will have a high computational cost. On the other hand,
when applying the lifting technique to the nonlinear MRSs, the trans-
formation will be complicated due to the iteration of the nonlinear
Fig. 3. An illustration of the multi-rate sampling scheme with 𝑏1 = 2 and 𝑏2 = 3.
function.

2.2.2. Iterating the state equation

(𝑖 ∈ [1, 𝑁]) are allowed to be different. That is, in the MRS (1), the
By simply iterating the state equation in (1), we have the following
state update period of the system is allowed to be different with the
new state equation with a state update period 𝑏𝑖 ℎ:
sampling period of the sensor, while the sampling periods of different
sensors are also allowed to be different. An illustration of the multi-rate 𝑥(𝑡𝑖𝑘+1 ) = 𝐴̃ 𝑖𝑏 (𝑡𝑖𝑘 )𝑥(𝑡𝑖𝑘 ) + 𝐵̃ 𝑏𝑖 (𝑡𝑖𝑘 )𝑤̃ 𝑖 (𝑡𝑖𝑘 ).
𝑖 𝑖
sampling scheme is given in Fig. 3. Moreover, the MRS (1) can be seen
as a combination of 𝑁 multi-rate subsystems where the 𝑖th subsystem Accordingly, we have the following SRS:
{ 𝑖
consists of the state equation and the measurement equation of the 𝑖th 𝑥(𝑡 ) = 𝐴̃ 𝑖 (𝑡𝑖 )𝑥(𝑡𝑖 ) + 𝐵̃ 𝑖 (𝑡𝑖 )𝑤̃ 𝑖 (𝑡𝑖 ),
𝑘+1 𝑏𝑖 𝑘 𝑘 𝑏𝑖 𝑘 𝑘
sensor. (3)
𝑦𝑖 (𝑡𝑖𝑘 ) = 𝐶𝑖 (𝑡𝑖𝑘 )𝑥𝑖 (𝑡𝑖𝑘 ) + 𝐷𝑖 (𝑡𝑖𝑘 )𝑣𝑖 (𝑡𝑖𝑘 ).
2.2. Transformation from MRSs to SRSs For the SRS (3), the 𝑖th local estimator is designed as

𝑥̂ 𝑖 (𝑡𝑖𝑘+1 ) = 𝐴̃ 𝑖𝑏 (𝑡𝑖𝑘 )𝑥̂ 𝑖 (𝑡𝑖𝑘 ) + 𝐾𝑖 (𝑡𝑖𝑘 )(𝑦𝑖 (𝑡𝑖𝑘 ) − 𝐶𝑖 (𝑡𝑖𝑘 )𝑥̂ 𝑖 (𝑡𝑖𝑘 ))
After presenting the state–space model of the MRSs, in this subsec- 𝑖

tion, we are going to introduce some effective methods that convert the where 𝑥̂ 𝑖 (𝑡𝑖𝑘 ) is the estimate of 𝑥(𝑡𝑖𝑘 ) and 𝐾𝑖 (𝑡𝑖𝑘 ) is the estimator gain to
MRSs into single-rate ones. be designed.
By iterating the state equation, the 𝑖th multi-rate subsystem of (1)
2.2.1. Using the lifting technique is transformed into the SRS (3). Compared to the SRS (2) obtained by
The lifting technique proposed in [75] is one of the popular methods using the lifting technique, the computational cost of the estimation
that are used to transform the MRSs. The main idea of the lifting algorithm designed for the SRS (3) is low. Nevertheless, the designed
technique is to obtain equivalent SRSs by increasing the state update estimation algorithm only estimates the states at the measurement
period. Taking the 𝑖th multi-rate subsystem of (1) as an example, the sampling instants (i.e., 𝑡𝑖𝑘 (𝑘 = 0, 1, 2, …)) and other states are not
states 𝑥(𝑡𝑖𝑘−1 + ℎ), … , 𝑥(𝑡𝑖𝑘 − ℎ), 𝑥(𝑡𝑖𝑘 ) in the interval (𝑡𝑖𝑘−1 , 𝑡𝑖𝑘 ] are first estimated.
augmented into a vector 𝑥̄ 𝑖 (𝑡𝑖𝑘 ). Then, a new state equation with the
state update period 𝑏𝑖 ℎ is obtained by the aid of the original state
2.2.3. Compensating measurements with zero
equation. Accordingly, the MRSs are transformed into SRSs. In the
In the above two approaches, the MRSs are transformed into SRSs
following, the detailed process is presented.
by increasing the state update period of the state equation. Intuitively,
Denote 𝑥̄ 𝑖 (𝑡𝑖𝑘 ) ≜ col{𝑥(𝑡𝑖𝑘−1 + ℎ), … , 𝑥(𝑡𝑖𝑘 − ℎ), 𝑥(𝑡𝑖𝑘 )}. By recurring to
we can also complete the transformation by decreasing the sampling
(1), it is derived that
period of the measurement equation. A method to achieve this goal is
𝑥(𝑡𝑖𝑘 + 𝑚ℎ) = 𝐴̃ 𝑖𝑚 (𝑡𝑖𝑘 )𝑥(𝑡𝑖𝑘 ) + 𝐵̃ 𝑚𝑖 (𝑡𝑖𝑘 )𝑤̃ 𝑖 (𝑡𝑖𝑘 ) compensating the measurements at non-sampling instants with zero.
By introducing a variable
where {
∏
𝑚 1, if mod(𝑠𝑘 , 𝑏𝑖 ℎ) = 0;
𝜆𝑖 (𝑠𝑘 ) =
𝐴̃ 𝑖𝑚 (𝑡𝑖𝑘 ) ≜ 𝐴(𝑡𝑖𝑘 + (𝑚 − 𝑙)ℎ), 𝐵̃ 𝑚𝑖 (𝑡𝑖𝑘 ) ≜ 𝐹̃𝑚𝑖 (𝑡𝑖𝑘 )𝐵̌ 𝑖 (𝑡𝑖𝑘 ), 0, otherwise
𝑙=1
[ ] with mod(𝑥, 𝑦) being the unique nonnegative remainder on division of
𝐹̃𝑚𝑖 (𝑡𝑖𝑘 ) ≜ 𝐴̃ 𝑖𝑚−1 (𝑡𝑖𝑘 ) 𝐴̃ 𝑖𝑚−2 (𝑡𝑖𝑘 ) ⋯ 𝐴𝑖1 (𝑡𝑖𝑘 ) 0 ⋯ 0 ,
𝑥 by 𝑦, the compensated measurement is formulated as
𝐵̌ 𝑖 (𝑡𝑖𝑘 ) ≜ diag{𝐵(𝑡𝑖𝑘 ), 𝐵(𝑡𝑖𝑘 + ℎ), … , 𝐵(𝑡𝑖𝑘+1 − ℎ)},
𝑦𝑖 (𝑠𝑘 ) = 𝜆𝑖 (𝑠𝑘 )𝐶𝑖 (𝑠𝑘 )𝑥(𝑠𝑘 ) + 𝜆𝑖 (𝑠𝑘 )𝐷𝑖 (𝑠𝑘 )𝑣𝑖 (𝑠𝑘 ).
𝑤̃ 𝑖 (𝑡𝑖𝑘 ) ≜ col{𝑤(𝑡𝑖𝑘 ), 𝑤(𝑡𝑖𝑘 + ℎ), … , 𝑤(𝑡𝑖𝑘+1 − ℎ)}.
Then, a SRS is derived as follows:
Then, it is obtained that {
{ 𝑖 𝑥(𝑠𝑘+1 ) =𝐴(𝑠𝑘 )𝑥(𝑠𝑘 ) + 𝐵(𝑠𝑘 )𝑤(𝑠𝑘 ),
𝑥̄ 𝑖 (𝑡𝑘+1 ) = 𝐴̄ 𝑖 (𝑡𝑖𝑘 )𝑥̄ 𝑖 (𝑡𝑖𝑘 ) + 𝐵̄ 𝑖 (𝑡𝑖𝑘 )𝑤̃ 𝑖 (𝑡𝑖𝑘 ), (4)
(2) 𝑦𝑖 (𝑠𝑘 ) =𝜆𝑖 (𝑠𝑘 )𝐶𝑖 (𝑠𝑘 )𝑥(𝑠𝑘 ) + 𝜆𝑖 (𝑠𝑘 )𝐷𝑖 (𝑠𝑘 )𝑣𝑖 (𝑠𝑘 ).
𝑦𝑖 (𝑡𝑖𝑘 ) = 𝐶̄𝑖 (𝑡𝑖𝑘 )𝑥̄ 𝑖 (𝑡𝑖𝑘 ) + 𝐷𝑖 (𝑡𝑖𝑘 )𝑣𝑖 (𝑡𝑖𝑘 )
For the SRS (4), the 𝑖th local estimator is designed as
where
⎡0 ⋯ 0 𝐴̃ 𝑖1 (𝑡𝑖𝑘 ) ⎤ 𝑥̂ 𝑖 (𝑠𝑘+1 ) =𝐴(𝑠𝑘 )𝑥̂ 𝑖 (𝑠𝑘 )
⎢ ⎥
⎢0 ⋯ 0 𝐴̃ 𝑖2 (𝑡𝑖𝑘 ) ⎥ + 𝐾𝑖 (𝑠𝑘 )(𝑦𝑖 (𝑠𝑘 ) − 𝜆𝑖 (𝑠𝑘 )𝐶𝑖 (𝑠𝑘 )𝑥̂ 𝑖 (𝑠𝑘 ))
𝐴̄ 𝑖 (𝑡𝑖𝑘 ) ≜ ⎢ ⎥,
⎢⋮ ⋮ ⋮ ⋮ ⎥ where 𝑥̂ 𝑖 (𝑠𝑘 ) is the estimate of 𝑥(𝑠𝑘 ) and 𝐾𝑖 (𝑠𝑘 ) is the estimator gain to
⎢0 ⋯ 0 𝐴̃ 𝑖𝑏 (𝑡𝑖𝑘 )⎥⎦ be designed.
⎣ 𝑖

21
Y. Shen et al. Information Fusion 82 (2022) 19–27

The advantages of such an approach are that the computational 3.1. 𝐻∞ estimation approach
cost is low and the designed estimation algorithm can estimate the
state at all state update instants. Unfortunately, the estimation ac- In the past few decades, the 𝐻∞ estimation approach has received
curacy may be low since the estimate is obtained by prediction at particular research interest since it is able to attenuate the influence
non-measurement-sampling instants. of the energy-bounded external disturbances on the estimation perfor-
mance. The core idea of the 𝐻∞ estimation is to design an estimator
2.2.4. Compensating measurements with the zero-order holder
such that a predefined disturbance attenuation level (i.e. the 𝐻∞ per-
Suppose that a zero-order holder is used to compensate the mea-
formance index) is achieved. Based on the methods like the linear
surements at the non-sampling instants. After the compensation, the
measurement available is matrix inequality, the Hamilton–Jacobi inequality as well as the Riccati
equation, the 𝐻∞ state estimation problems have been concerned for
𝑦̄𝑖 (𝑠𝑘 ) ≜ 𝑦𝑖 (𝑡𝑖𝑚 ), 𝑡𝑖𝑚 ≤ 𝑠𝑘 < 𝑡𝑖𝑚+1 . various systems [93,94].
Inspired by the input delay approach which has been used to In the MRS state estimation problems, the 𝐻∞ estimation approach
transform the discrete-time control input into delayed control input has been widely used and a large body of literature has been available
[76], we define a variable 𝜌𝑖𝑠 as follows: [95–100]. In [95], the 𝐻2 and 𝐻∞ filtering problems have been studied
𝑘
for multi-rate linear time-invariant systems with a state update period
𝜌𝑖𝑠 ≜ 𝑠𝑘 − 𝑡𝑖𝑚 , 𝑡𝑖𝑚 ≤ 𝑠𝑘 < 𝑡𝑖𝑚+1 .
𝑘 ℎ and a sampling period 𝑛ℎ. First, a standard filter with a periodic
Noting that 𝑡𝑖𝑚 = 𝑠𝑘 − (𝑠𝑘 − 𝑡𝑖𝑚 ) = 𝑠𝑘 − 𝜌𝑖𝑠 holds for 𝑡𝑖𝑚 ≤ 𝑠𝑘 < 𝑡𝑖𝑚+1 , the filter gain and an update period ℎ has been proposed to cater for the
𝑘
measurement 𝑦̄𝑖 (𝑠𝑘 ) is reformulated as multi-rate sampling. Then, the state update period, the sampling period
and the estimate update period have been uniformed by iterating the
𝑦̄𝑖 (𝑠𝑘 ) = 𝑦𝑖 (𝑠𝑘 − 𝜌𝑖𝑠 )
𝑘 system state equation and the filter equation. Finally, the 𝐻2 and 𝐻∞
= 𝐶𝑖 (𝑠𝑘 − 𝜌𝑖𝑠 )𝑥(𝑠𝑘 − 𝜌𝑖𝑠 ) filters have been designed by solving certain linear matrix inequalities
𝑘 𝑘

+ 𝐷𝑖 (𝑠𝑘 − 𝜌𝑖𝑠 )𝑣𝑖 (𝑠𝑘 − 𝜌𝑖𝑠 ). where nonconvex constraints have been introduced due to the multi-
𝑘 𝑘
rate sampling. Subsequently, the authors of [95] has extended the
Therefore, the 𝑖th multi-rate subsystem of (1) is transformed into a SRS results to multi-rate linear time-invariant systems with packet dropouts
in [96] where the state update period, the sampling period as well as
⎧𝑥(𝑠𝑘+1 ) = 𝐴(𝑠𝑘 )𝑥(𝑠𝑘 ) + 𝐵(𝑠𝑘 )𝑤(𝑠𝑘 ), the estimate update period are different. A sufficient condition has been
⎪ 𝑖 𝑖 provided on the existence of a stable filter. In [97], the 𝐻∞ estimation
⎨ 𝑦̄𝑖 (𝑠𝑘 ) = 𝐶𝑖 (𝑠𝑘 − 𝜌𝑠𝑘 )𝑥(𝑠𝑘 − 𝜌𝑠𝑘 ) (5)
⎪ problem has been concerned for multi-sensor multi-rate linear time-
⎩ + 𝐷𝑖 (𝑠𝑘 − 𝜌𝑖𝑠 )𝑣𝑖 (𝑠𝑘 − 𝜌𝑖𝑠 ).
𝑘 𝑘 invariant systems where the estimate update period is integer multiple
For the SRS (5), the 𝑖th local estimator is designed as of the sampling period/state update period. The measurements of the
𝑥̂ 𝑖 (𝑠𝑘+1 ) = 𝐴(𝑠𝑘 )𝑥̂ 𝑖 (𝑠𝑘 ) sensors have been transmitted to a fusion center in a competitive
way. By fusing the measurements received in one estimate update
+ 𝐾𝑖 (𝑠𝑘 )(𝑦̄𝑖 (𝑠𝑘 ) − 𝐶𝑖 (𝑠𝑘 − 𝜌𝑖𝑠 )𝑥(𝑠
̂ 𝑘 − 𝜌𝑖𝑠 ))
𝑘 𝑘 period, the estimator in the fusion center has estimated the system
where 𝑥̂ 𝑖 (𝑠𝑘 ) is the estimate of 𝑥(𝑠𝑘 ) and 𝐾𝑖 (𝑠𝑘 ) is the estimator gain to state. A sufficient condition has been obtained on the mean-square
be designed. stability as well as the 𝐻∞ performance and the 𝐻∞ estimator has been
characterized.
Remark 1. In this section, for simplification of the formulas, we only For time-varying MRSs, the quantized finite-horizon 𝐻∞ filtering
consider the case that the system state update period and the sampling problem has been studied in [98] under the stochastic communica-
period are different. Note that, the transformation methods provided in tion protocol. The lifting technique has been first applied to han-
this section can be easily extended to the case where the system state dle the MRSs, and then the desired 𝐻∞ filter has been designed
update period, the sampling period and the estimate update period are by solving certain Riccati different equations. Similar to [98], the
mutually different.
variance-constrained finite-horizon 𝐻∞ state estimation problem has
3. Estimation problems for multi-rate systems been considered in [99] for time-varying MRSs. In [100], the sequential
fusion 𝐻∞ filtering problem has been investigated for time-varying
Due to the practical engineering significance of the multi-rate sam- MRSs where 𝑝 asynchronous sensors are used to measure the system
pling, the state estimation problem for MRSs has become an active and send the measurements to a fusion center asynchronously. In the
research topic. Up to now, a number of estimation approaches have fusion center, the estimator updates the estimate once a measurement is
been applied on MRS state estimation problems. According to the received. With such a multi-rate sampling strategy, a sequential fusion
performance indices used, the applied estimation approaches can be 𝐻∞ filter has been designed based on the Krein-space approach.
categorized into the 𝐻∞ estimation approach [77–81], the Kalman The fault detection problem is a long-standing research topic that
filtering approach and its variants [82–84], the set-membership filter- has been widely studied [101–105]. In the MRSs, due to reasons like
ing approach [85–87], the moving horizon estimation approach [88] sensor aging, random sensor failure, or harsh environment, the fault
and etc. On the other hand, in the state estimation problems for may occur which largely degrades the system performance. There-
MRSs, the multi-sensor information fusion methods are widely used fore, it is of great importance to design fault detection algorithm
that include the weighted by matrix fusion method, the weighted by
to detect the fault. Recently, the 𝐻∞ fault detection problems for
scalar fusion method, the covariance intersection fusion method, the
MRSs have received initial research interest. In [106], the 𝐻∞ fault
sequential covariance intersection fusion method and so on [37,89–
detection problem has been investigated for MRSs with asynchronous
92]. The weighted by matrix fusion method and the weighted by
state update rate and sampling rate. By iterating the state equation,
scalar fusion method can obtain the optimal fused estimate but need
the cross-covariances among local estimation errors. The covariance the equivalent SRSs have been derived, and then an observer-based
intersection fusion method and the sequential covariance intersection fault detection filter has been designed. With the help of the linear
fusion method can avoid the calculation of the cross-covariances but matrix inequality method, sufficient conditions have been developed
cannot obtain optimal fused estimate. such that the 𝐻∞ norm from the noises and faults to the fault estimation
In the following, the available results on state estimation problems error is less that a given attenuation level. In [107], the intermittent
for MRSs are reviewed and the corresponding fusion methods are also fault detection problem has been concerned for a class of nonuniformly
discussed. sampled MRSs. The nonuniform sampling interval is governed by a

22
Y. Shen et al. Information Fusion 82 (2022) 19–27

Markov process with partly unknown and uncertain transition proba- process at each time instant. The distributed fusion estimation problem
bilities. Due to the existence of the Markov process, the MRSs has been has been considered in [126] for multi-rate linear systems where the
transformed into SRSs with Markovian jumping parameters. Sampling- estimation process and the fusion process are similar to those in [123].
interval-dependent fault detection filters have been designed such that The fault detection problems for MRSs based on the Kalman filtering
the residual estimation error satisfies the 𝐻− and 𝐻∞ performances approach have received little research attentions. In [127], the Kalman
simultaneously. filtering problem has been investigated for a class of non-uniformly
sampled MRSs. After transforming the MRSs into SRSs with the lifting
3.2. Kalman filtering approach and its variants technique, a filter has been designed for the obtained SRSs and the
stability as well as the convergence of the filtering error has been
The classic Kalman filtering approach is one of the most celebrated analyzed. Then, the fault detection and isolation problems have been
filtering approaches. For linear systems with Gaussian noises whose concerned for the underlying non-uniformly sampled MRSs.
statistics are exactly known, the classic Kalman filtering is an optimal Other than the classic Kalman filtering approach, the modified
filtering approach under which the filtering error covariance is min- Kalman filtering approaches have also been widely used in the state
imized at each time instant. Nevertheless, when applied to nonlinear estimation problems for MRSs. In [128], the extended Kalman filtering
systems or systems with non-Gaussian noises, the classic Kalman filter- problem has been discussed for nonlinear systems where the availabil-
ing is no longer applicable. Therefore, some modified Kalman filtering ity of the primary and the secondary measurements are asynchronous.
approaches have been developed that includes the extended Kalman Several methods that effective for the asynchronous measurements
approach [108,109], the unscented Kalman approach [110,111], the have been discussed and modified in the extended Kalman filtering
cubature Kalman approach [112,113] and so on. framework. In [129], the state estimation problem has been dealt
For state estimation problems concerning the MRSs, the classic with for nonlinear multi-sensor MRSs by using the modified sigma
Kalman filtering approach is another popular approach that has been point Kalman filtering approach. A modified unscented Kalman filter
widely studied [114–120]. In [115], the fusion estimation problem has been proposed in [130] for the multi-rate INS/GPS integrated
has been studied for multi-sensor systems by using the traditional navigation systems. The state estimation problem has been investigated
Kalman filtering approach where the sampling periods of the sensors in [131] for nonlinear systems with a normal measurement and an
are asynchronous. By using the lifting technique, the multi-sensor infrequent integral measurement. The systems have been first refor-
MRS has been transformed into synchronous single-rate single-sensor mulated to equivalent variable dimension systems. Then, a variable
systems. Local Kalman filters has been first designed to obtain the local dimension unscented Kalman filter has been designed. In [132], a
estimates which have been then fused to a global estimate with the joint-unscented Kalman filtering algorithm has been proposed for a
help of the fusion method in [121]. In [119], the problem in [115] has continuous stirred-tank reactor system with asynchronous sensors.
been reconsidered and a novel fusion estimation algorithm has been
proposed which fuses the local estimates in a recursive form and is 3.3. Other estimation approaches
optimal in the linear minimum variance sense. In [120], the result
in [115] has been extended to multi-sensor systems with arbitrary In the state estimation problems for MRSs, the moving horizon
number of sensors and arbitrary sampling rates. Both the centralized estimation approach, the particle filtering approach, the 𝑙2 -𝑙∞ filtering
asynchronous fusion estimation algorithm and the distributed asyn- approach and the set-membership filtering approach have been taken
chronous fusion estimation algorithm have been developed based on into consideration either [133–138]. In the moving horizon estimation,
the Kalman filtering. It has been verified that the estimation perfor- the estimate is generated by solving a predefined optimization problem
mances of both fusion estimation algorithms are equivalent under the based on the measurements in a moving time interval with fixed
full-rate communication assumption, while the centralized one outper- length. Due to its efficiency in handling nonlinear systems, the moving
forms the distributed one when the communications are constrained. In horizon estimation has received ever-increasing research interest in
[122], the federated Kalman filtering problem has been investigated for the past decade [135,136,139]. In [135], the state estimation problem
asynchronous multi-sensor MRSs with random missing measurements. has been solved for mobile robot systems with asynchronous sensors
The Kalman fusion estimation algorithm has been developed in according to the moving horizon estimation. The sampling rates of the
[123] for MRSs over the sensor networks. The system state update sensors have been uniformed by compensating the measurements from
period, the measurement transmission period, and the estimate update the slow-rate sensor with a prediction value. Then, a moving horizon
period have been allowed to be different. The local Kalman filters have estimation algorithm has been developed by solving a regularized least-
been first designed based on the SRSs derived by the lifting technique. squares problem. The moving horizon estimation problem has been
Then, the local estimates have been fused with the weighted by matrix concerned in [136] for linear systems with multi-rate measurements
fusion method. Due to the asynchronism of the local estimates, at and correlated noises. By introducing a switching variable, the multi-
a certain time instant, only the available estimates have been fused. rate measurements have been combined in a new measurement model.
In [124], the hierarchical fusion estimation problem has been coped Based on the obtained SRSs, the desired moving horizon estimator has
with for MRSs over sensor networks. The underlying sensor network been designed.
consists of 𝑁 sensor clusters and each sensor cluster contains multiple The particle filtering approach, roots on the sampling-based ap-
sensors and a cluster head. First, the estimator in the cluster head proximation techniques, is an effective filtering approach for nonlinear
has generated a local estimate based on the fused measurement which systems or systems with non-Gaussian noises. In [140], both the particle
has been obtained by fusing the available measurements from the filtering and extended Kalman filtering have been applied to an in-
sensors in this cluster with the sequential fusion method. Then, the tensified chemical process subject to asynchronous measurements. The
cluster head has communicated with other cluster heads to generate 𝑙2 -𝑙∞ filtering problem has been studied in [141] for MRSs. The lifting
the fused estimate. In the estimate fusion process, only the available technique has been used to handle the MRSs and the 𝑙2 -𝑙∞ filter has
local estimates have been used also due to the asynchronism of the been designed according to the solution of linear matrix inequalities
local estimates. In [125], the Kalman fusion estimation problem has with a nonconvex constraint. The set-membership filtering problem
been also investigated for MRSs over sensor networks where the state has been concerned in [142] for MRSs over sensor networks. A set
update period and the sampling period are different. Different from of local filters have been designed such that the filtering errors have
[123], the estimates at the non-sampling instants have been obtained been constrained in a given ellipsoid. In [143], the zonotopes-based
through prediction based on the state equation. Therefore, the estimates distributed set-membership filtering problem has been investigated for
from all local estimators have been available in the estimation fusion MRSs where the state estimates belong to the computed sets.

23
Y. Shen et al. Information Fusion 82 (2022) 19–27

3.4. Handling the network-induced challenges (2) The complex networks can characterize plenty of real-world dy-
namical systems and have received consistent research interest
In the networked systems, due to the limited resources of the in the past decades [156,157]. Unfortunately, the studies on the
communication networks, the network-induced phenomena inevitably state estimation problems for complex networks with multi-rate
occur [144,145]. Up to now, considerable research attentions have sampling strategy are quite few. Therefore, the state estimation
been paid on the state estimation problems for networked MRSs sub- problem for multi-rate complex networks will be an attractive
ject to network-induced phenomena and plenty of results have been area.
available. In the following, some typical network-induced phenomena (3) Other than the communication protocols discussed in this sur-
are introduced and the corresponding state estimation problems for vey, there are also some effective communication protocols,
networked MRSs are summarized. e.g. the FlexRay protocol, introduced in the networked systems
The packet dropout is one of the most frequently occurred network- and the corresponding state estimation algorithms have been
induced phenomena which may be caused by many reasons such as developed. Nevertheless, how to modify the existing results to
intermittent sensor failures, network congestion and so on. In [146], make them applicable to MRSs or how to develop novel es-
the linear-minimum-mean-square-error observer design problem has timation algorithms suitable for MRSs with above-mentioned
been considered for multi-sensor MRSs with multiple packet dropouts. protocols remain open and challenging.
The observer has been designed by minimizing the estimation error (4) In the estimate fusion process, due to the asynchronism of the
covariances. In [147], the almost surely state estimation problem has local estimates, only the estimates available at the current time
been studied for MRSs subject to both Markovian packet dropouts and instant are fused. As such, the valuable estimate information
random packet dropouts characterized by the Bernoulli distributed ran- are not fully used which would lead to deterioration of the
dom variables. An estimator has been designed such that the estimation estimation accuracy. To this end, a trend for future research
error is almost surely exponentially stable. The non-fragile distributed is to propose novel fusion strategies that make full use of the
𝐻∞ filtering problem has been concerned in [148] for MRSs with asynchronous estimates.
packet dropouts characterized by the Gilbert–Elliott model. A sufficient (5) Another promising research topic is to study the fault detec-
condition has been derived on the exponential stability and the 𝐻∞ tion, diagnosis, and isolation problems for MRSs. Although some
performance. initial results have been obtained, the corresponding theories
The signal quantization and saturation are two ubiquitous network- for MRSs are far from mature. Hence, it is of great importance
induced phenomena due to the inherent nature of the digital transmis- to develop fault detection, diagnosis, and isolation theories for
sion and the physical constraints of the hardware, respectively [149– MRSs.
151]. In [152], the variance-constrained 𝐻∞ state estimation problem
has been studied for MRSs subject to measurement quantization. The Declaration of competing interest
quantization effect has been characterized by a logarithmic quantizer
and transformed to sector-bounded uncertainties. The desired 𝐻∞ filter The authors declare that they have no known competing finan-
has been designed by resorting to the stochastic analysis approach and cial interests or personal relationships that could have appeared to
the Lyapunov theory. In [153], the multi-objective filtering problem influence the work reported in this paper.
has been concerned for MRSs with random sensor saturations. The
saturation function has been rewritten as a combination of a linear term References
and a nonlinear function satisfying the sector condition.
In the above results, the state estimation algorithms have been [1] J. Mao, Y. Sun, X. Yi, H. Liu, D. Ding, Recursive filtering of networked nonlinear
systems: A survey, Internat. J. Systems Sci. 52 (6) (2021) 1110–1128.
developed such that, with the existence of the network-induced phe-
[2] J. Hu, H. Zhang, H. Liu, X. Yu, A survey on sliding mode control for networked
nomena, the estimation performance still satisfies given performance control systems, Internat. J. Systems Sci. 52 (6) (2021) 1129–1147.
requirements. In another way, we can also introduce certain commu- [3] D. Ding, Q.-L. Han, X. Ge, J. Wang, Secure state estimation and control of
nication protocols to reduce the occurrence of the network-induced cyber–physical systems: A survey, IEEE Trans. Syst. Man Cybern.: Syst. 51 (1)
phenomena. In [154], the event-triggered protocol has been introduced (2021) 176–190.
[4] Y. Shi, E. Tian, S. Shen, X. Zhao, Adaptive memory-event-triggered 𝐻∞ control
in the fusion estimation problem for MRSs with sensor degradations.
for network-based T-S fuzzy systems with asynchronous premise constraints,
In [98], the stochastic communication protocol has been considered in IET Control Theory Appl. 15 (4) (2021) 534–544.
𝐻∞ state estimation for MRSs. The round-robin protocol has been dis- [5] N. Xu, Y. Chen, A. Xue, Q. Yu, A new approach to stability analysis of switched
cussed in [68] and [142] where the recursive state estimation problem systems composed of two subsystems with average dwell time switching,
Internat. J. Systems Sci. 51 (10) (2020) 1862–1872.
and the distributed set-membership filtering problem have been coped
[6] H. Geng, Y. Liang, Y. Cheng, Target state and Markovian jump ionospheric
with for MRSs, respectively. The weighted try-once-discard protocol has height bias estimation for OTHR tracking systems, IEEE Trans. Syst. Man
been introduced in [155] for recursive filtering problem of MRSs. Cybern.: Syst. 50 (7) (2020) 2599–2611.
[7] X. Tian, G. Wei, L. Wang, J. Zhou, Wireless-sensor-network-based target local-
4. Conclusions and future work ization: A semidefinite relaxation approach with adaptive threshold correction,
Neurocomputing 405 (2020) 229–238.
[8] Y. Chen, J. Ren, X. Zhao, A. Xue, State estimation of Markov jump neural
In this survey, the state estimation approaches developed for MRSs networks with random delays by redundant channels, Neurocomputing 453
and the corresponding multi-sensor information fusion methods have (2021) 493–501.
been discussed and reviewed. First, a general state–space model has [9] Q. Liu, Z. Wang, J. Zhang, D.H. Zhou, Necessary and sufficient conditions
been given to characterize the MRSs. The methods that transform the for fault diagnosability of linear open- and closed-loop stochastic systems
under sensor and actuator faults, IEEE Trans. Automat. Control (2021) http:
MRSs into SRSs have been summarized. Then, the recent advances on
//dx.doi.org/10.1109/TAC.2021.3108587, (in press).
the fusion estimation problems for MRSs have been reviewed based on [10] C. Han, H. Zhang, M. Fu, Optimal filtering for networked systems with
the applied estimation approach. According to the literature review, Markovian communication delays, Automatica 49 (10) (2013) 3097–3104.
some future research topics are given as follows: [11] X.-M. Zhang, Q.-L. Han, X. Yu, Survey on recent advances in networked control
systems, IEEE Trans. Ind. Inf. 12 (5) (2016) 1740–1752.
(1) In the existing literature, the state estimation problems are [12] X.-M. Zhang, Q.-L. Han, X. Ge, D. Ding, L. Ding, D. Yue, C. Peng, Networked
mostly investigated for linear MRSs or nonlinear MRSs with control systems: A survey of trends and techniques, IEEE/CAA J. Autom. Sin.
7 (1) (2020) 1–17.
strict assumption on the nonlinear function. It would be an inter- [13] R. Caballero-Águila, A. Hermoso-Carazo, J. Linares-Pérez, Optimal state estima-
esting and challenging topic to investigate the state estimation tion for networked systems with random parameter matrices, correlated noises
problems for MRSs with general nonlinearities. and delayed measurements, Int. J. Gen. Syst. 44 (2) (2015) 142–154.

24
Y. Shen et al. Information Fusion 82 (2022) 19–27

[14] Y. Cui, Y. Liu, W. Zhang, F.E. Alsaadi, Sampled-based consensus for nonlinear [42] Q. Song, Y. Chen, Z. Zhao, Y. Liu, F.E. Alsaadi, Robust stability of fractional-
multiagent systems with deception attacks: The decoupled method, IEEE Trans. order quaternion-valued neural networks with neutral delays and parameter
Syst. Man Cybern.: Syst. 51 (1) (2021) 561–573. uncertainties, Neurocomputing 420 (2021) 70–81.
[15] B. Qu, N. Li, Y. Liu, F.E. Alsaadi, Estimation for power quality disturbances with [43] Q. Wu, Q. Song, Z. Zhao, Y. Liu, F.E. Alsaadi, Stabilization of T-S fuzzy
multiplicative noises and correlated noises: A recursive estimation approach, fractional rectangular descriptor time-delay system, Internat. J. Systems Sci.
Internat. J. Systems Sci. 51 (7) (2020) 1200–1217. 52 (11) (2021) 2268–2282.
[16] X. Li, F. Han, N. Hou, H. Dong, H. Liu, Set-membership filtering for piece- [44] Y. Chen, Z. Chen, Z. Chen, A. Xue, Observer-based passive control of non-
wise linear systems with censored measurements under Round-Robin protocol, homogeneous Markov jump systems with random communication delays,
Internat. J. Systems Sci. 51 (9) (2020) 1578–1588. Internat. J. Systems Sci. 51 (6) (2020) 1133–1147.
[17] H. Tan, B. Shen, K. Peng, H. Liu, Robust recursive filtering for uncertain [45] J. Li, H. Dong, Z. Wang, X. Bu, Partial-neurons-based passivity-guaranteed
stochastic systems with amplify-and-forward relays, Internat. J. Systems Sci. state estimation for neural networks with randomly occurring time-delays, IEEE
51 (7) (2020) 1188–1199. Trans. Neural Netw. Learn. Syst. 31 (9) (2020) 3747–3753.
[18] W. Qian, Y. Li, Y. Zhao, Y. Chen, New optimal method for 𝐿2 -𝐿∞ state [46] Y. Yuan, Y. Wang, L. Guo, Sliding-mode-observer-based time-varying formation
estimation of delayed neural networks, Neurocomputing 415 (2020) 258–265. tracking for multi-spacecrafts subjected to switching topologies and time-delays,
[19] T. Bonargent, T. Ménard, O. Gehan, E. Pigeon, Adaptive observer design for a IEEE Trans. Automat. Control 66 (8) (2021) 3848–3855.
class of Lipschitz nonlinear systems with multirate outputs and uncertainties:
[47] X. Gu, T. Jia, Y. Niu, Consensus tracking for multi-agent systems subject to
Application to attitude estimation with gyro bias, Int. Robust Nonlinear Control
channel fading: A sliding mode control method, Internat. J. Systems Sci. 51
31 (8) (2021) 3137–3162.
(14) (2020) 2703–2711.
[20] Y. Salehi, B. Huang, Offline and online parameter learning for switching
[48] L. Liu, L. Ma, J. Zhang, Y. Bo, Distributed non-fragile set-membership filtering
multirate processes with varying delays and integrated measurements, IEEE
for nonlinear systems under fading channels and bias injection attacks, Internat.
Trans. Ind. Electron. (2021) http://dx.doi.org/10.1109/TIE.2021.3095807, (in
J. Systems Sci. 52 (6) (2021) 1192–1205.
press).
[49] X. Guan, J. Hu, J. Qi, D. Chen, F. Zhang, G. Yang, Observer-based 𝐻∞ sliding
[21] J. Salt, J. Alcaina, Á. Cuenca, A. Baños, Multirate control strategies for avoiding
mode control for networked systems subject to communication channel fading
sample losses. Application to UGV path tracking, ISA Trans. 101 (2020)
and randomly varying nonlinearities, Neurocomputing 437 (2021) 312–324.
130–146.
[22] M. Moarref, L. Rodrigues, Observer design for linear models of multi-rate [50] D. Liu, Z. Wang, Y. Liu, F.E. Alsaadi, F.E. Alsaadi, Recursive state estimation
asynchronous aerospace systems, Aerosp. Syst. 3 (2020) 127–137. for stochastic complex networks under Round-Robin communication protocol:
[23] A. Smyth, M. Wu, Multi-rate Kalman filtering for the data fusion of displacement Handling packet disorders, IEEE Trans. Netw. Sci. Eng. 8 (3) (2021) 2455–2468.
and acceleration response measurements in dynamic system monitoring, Mech. [51] Y. Cui, L. Xu, Y. Shen, 𝐻2 /𝐻∞ Filtering for networked systems with data
Syst. Signal Process. 21 (2) (2007) 706–723. transmission time-varying delay, data packet dropout and sequence disorder,
[24] R. Cristi, M. Tummala, Multirate, multiresolution, recursive Kalman filter, Signal J. Franklin Inst. B 354 (13) (2017) 5443–5462.
Process. 80 (9) (2000) 1945–1958. [52] M.B.G. Cloosterman, L. Hetel, N. van de Wouw, W.P.M.H. Heemels, J. Daafouz,
[25] S. Safari, F. Shabani, D. Simon, Multirate multisensor data fusion for linear H. Nijmeijer, Controller synthesis for networked control systems, Automatica 46
systems using Kalman filters and a neural network, Aerosp. Sci. Technol. 39 (10) (2010) 1584–1594.
(2014) 465–471. [53] W. Chen, J. Hu, Z. Wu, X. Yu, D. Chen, Finite-time memory fault detection
[26] D. Andrisani, C.-F. Gau, Estimation using a multirate filter, IEEE Trans. filter design for nonlinear discrete systems with deception attacks, Internat. J.
Automat. Control 32 (7) (1987) 653–656. Systems Sci. 51 (8) (2020) 1464–1481.
[27] R.G. Shenoy, D. Burnside, T.W. Parks, Linear periodic systems and multirate [54] J. Song, D. Ding, H. Liu, X. Wang, Non-fragile distributed state estimation over
filter design, IEEE Trans. Signal Process. 42 (9) (1994) 2242–2256. sensor networks subject to DoS attacks: The almost sure stability, Internat. J.
[28] X. Chi, X. Jia, F. Cheng, M. Fan, Networked 𝐻∞ filtering for Takagi–Sugeno Systems Sci. 51 (6) (2020) 1119–1132.
fuzzy systems under multi-output multi-rate sampling, J. Franklin Inst. B 356 [55] Q. Li, J. Liang, H. Qu, 𝐻∞ Estimation for stochastic semi-Markovian switching
(6) (2019) 3661–3691. CVNNs with missing measurements and mode-dependent delays, Neural Netw.
[29] H. Lin, S. Sun, An overview of multirate multisensor systems: Modelling and 141 (2021) 281–293.
estimation, Inf. Fusion 52 (2019) 335–343. [56] Y. Ju, Y. Liu, X. He, B. Zhang, Finite-horizon 𝐻∞ filtering and fault isolation
[30] Y. Zhang, Z. Wang, L. Zou, H. Fang, Event-based finite-time filtering for for a class of time-varying systems with sensor saturation, Internat. J. Systems
multirate systems with fading measurements, IEEE Trans. Aerosp. Electron. Syst. Sci. 52 (2) (2021) 321–333.
53 (3) (2017) 1431–1441. [57] G. Wei, L. Liu, L. Wang, D. Ding, Event-triggered control for discrete-time
[31] B. Qu, Z. Wang, B. Shen, Fusion estimation for a class of multi-rate power systems with unknown nonlinearities: An interval observer-based approach,
systems with randomly occurring SCADA measurement delays, Automatica 125 Internat. J. Systems Sci. 51 (6) (2020) 1019–1031.
(2021) 109408. [58] S. Wen, T. Huang, X. Yu, M.Z.Q. Chen, Z. Zeng, Aperiodic sampled-data
[32] H. Tan, B. Shen, Y. Liu, A. Alsaedi, B. Ahmad, Event-triggered multi-rate sliding-mode control of fuzzy systems with communication delays via the
fusion estimation for uncertain system with stochastic nonlinearities and colored event-triggered method, IEEE Trans. Fuzzy Syst. 24 (5) (2016) 1048–1057.
measurement noises, Inf. Fusion 36 (2017) 313–320. [59] X. Zhou, Y. Chen, Q. Wang, K. Zhang, A. Xue, Event-triggered finite-time 𝐻∞
[33] H. Lin, S. Sun, Distributed fusion estimation for multi-sensor asynchronous control of networked state-saturated switched systems, Internat. J. Systems Sci.
sampling systems with correlated noises, Internat. J. Systems Sci. 48 (5) (2017) 51 (10) (2020) 1744–1758.
952–960.
[60] J. Suo, N. Li, Q. Li, Event-triggered 𝐻∞ state estimation for discrete-time de-
[34] Z. Deng, P. Zhang, W. Qi, J. Liu, Y. Gao, Sequential covariance intersection
layed switched stochastic neural networks with persistent dwell-time switching
fusion Kalman filter, Inform. Sci. 189 (2012) 293–309.
regularities and sensor saturations, Neurocomputing 455 (2021) 297–307.
[35] S. Sun, H. Lin, J. Ma, X. Li, Multi-sensor distributed fusion estimation with
[61] Y. Niu, L. Sheng, M. Gao, D. Zhou, Dynamic event-triggered state estimation
applications in networked systems: A review paper, Inf. Fusion 38 (2017)
for continuous-time polynomial nonlinear systems with external disturbances,
122–134.
IEEE Trans. Ind. Inf. 17 (6) (2020) 3962–3970.
[36] R. Caballero-Águila, A. Hermoso-Carazo, J. Linares-Pérez, Fusion estimation
[62] Z. Zhang, S.-F. Su, Y. Niu, Dynamic event-triggered control for interval type-
using measured outputs with random parameter matrices subject to random
2 fuzzy systems under fading channel, IEEE Trans. Cybern. 51 (11) (2021)
delays and packet dropouts, Signal Process. 127 (2016) 12–23.
5342–5351.
[37] R. Caballero-Águila, A. Hermoso-Carazo, J. Linares-Pérez, Networked fusion
estimation with multiple uncertainties and time-correlated channel noise, Inf. [63] Z. Cao, Y. Niu, H.-K. Lam, J. Zhao, Sliding mode control of Markovian jump
Fusion 54 (2020) 161–171. fuzzy systems: A dynamic event-triggered method, IEEE Trans. Fuzzy Syst. 29
[38] M. Sathishkumar, R. Sakthivel, C. Wang, B. Kaviarasan, S.M. Anthoni, (10) (2021) 2902–2915.
Non-fragile filtering for singular Markovian jump systems with missing [64] P. Zhang, Y. Yuan, L. Guo, Fault-tolerant optimal control for discrete-time
measurements, Signal Process. 142 (2018) 125–136. nonlinear system subjected to input saturation: A dynamic event-triggered
[39] J. Mao, D. Ding, Y. Song, Y. Liu, F.E. Alsaadi, Event-based recursive filtering for approach, IEEE Trans. Cybern. 51 (6) (2021) 2956–2968.
time-delayed stochastic nonlinear systems with missing measurements, Signal [65] W. Xu, D.W.C. Ho, J. Zhong, B. Chen, Event/self-triggered control for leader-
Process 134 (2017) 158–165. following consensus over unreliable network with DoS attacks, IEEE Trans.
[40] J. Hu, H. Zhang, X. Yu, H. Liu, D. Chen, Design of sliding-mode-based control Neural Netw. Learn. Syst. 30 (10) (2019) 3137–3149.
for nonlinear systems with mixed-delays and packet losses under uncertain [66] W. Xu, S. Yang, J. Cao, Fully distributed self-triggered control for second-order
missing probability, IEEE Trans. Syst. Man Cybern.: Syst. 51 (5) (2021) consensus of multiagent systems, IEEE Trans. Syst. Man Cybern.: Syst. 51 (6)
3217–3228. (2021) 3541–3551.
[41] T. Wen, L. Zou, J. Liang, C. Roberts, Recursive filtering for communication- [67] M. Zhu, Y. Chen, Y. Kong, C. Chen, J. Bai, Distributed filtering for Markov
based train control systems with packet dropouts, Neurocomputing 275 (2018) jump systems with randomly occurring one-sided lipschitz nonlinearities under
948–957. round-robin scheduling, Neurocomputing 417 (2020) 396–405.

25
Y. Shen et al. Information Fusion 82 (2022) 19–27

[68] Y. Shen, Z. Wang, B. Shen, Q.-L. Han, Recursive state estimation for networked [96] Y. Liang, T. Chen, Q. Pan, Multi-rate stochastic 𝐻∞ filtering for networked
multi-rate multi-sensor systems with distributed time-delays under the Round- multi-sensor fusion, Automatica 46 (2) (2010) 437–444.
Robin protocol, IEEE Trans. Cybern. (2021) http://dx.doi.org/10.1109/TCYB. [97] H. Song, L. Yu, W.-A. Zhang, Multi-sensor-based 𝐻∞ estimation in heteroge-
2020.3021350, (in press). neous sensor networks with stochastic competitive transmission and random
[69] Y. Ju, G. Wei, D. Ding, S. Liu, A novel fault detection method under weighted sensor failures, IET Control Theory Appl. 8 (3) (2014) 202–210.
try-once-discard scheduling over sensor networks, IEEE Trans. Control Netw. [98] S. Liu, Z. Wang, L. Wang, G. Wei, On quantized 𝐻∞ filtering for multi-rate
Syst. 7 (3) (2020) 1489–1499. systems under stochastic communication protocols: The finite-horizon case,
[70] Z. Zhang, Y. Niu, H.-K. Lam, Sliding-mode control of T-S fuzzy systems Inform. Sci. 459 (2018) 211–223.
under weighted try-once-discard protocol, IEEE Trans. Cybern. 50 (12) (2020) [99] L. Wang, Z. Wang, G. Wei, F.E. Alsaadi, Variance-constrained 𝐻∞ state
4972–4982. estimation for time-varying multi-rate systems with redundant channels: The
[71] D. Li, J. Liang, F. Wang, X. Ren, Observer-based 𝐻∞ control of two-dimensional finite-horizon case, Inform. Sci. 501 (2019) 222–235.
delayed networks under the random access protocol, Neurocomputing 401 [100] X. Feng, C. Wen, J.H. Park, Sequential fusion 𝐻∞ filtering for multi-rate multi-
(2020) 353–363. sensor time-varying systems–A Krein-space approach, IET Control Theory Appl.
[72] Z. Zhang, Y. Niu, Z. Cao, J. Song, Security sliding mode control of interval 11 (3) (2017) 369–381.
type-2 fuzzy systems subject to cyber attacks: The stochastic communication [101] K. Huang, S. Wu, F. Li, C. Yang, W. Gui, Fault diagnosis of hydraulic systems
protocol case, IEEE Trans. Fuzzy Syst. 29 (2) (2020) 240–251. based on deep learning model with multirate data samples, IEEE Trans. Neural
[73] K. Zhu, J. Hu, Y. Liu, N.D. Alotaibi, F.E. Alsaadi, On 𝓁2 -𝓁∞ output-feedback Netw. Learn. Syst. (2021) http://dx.doi.org/10.1109/TNNLS.2021.3083401, (in
control scheduled by stochastic communication protocol for two-dimensional press).
switched systems, Internat. J. Systems Sci. 52 (14) (2021) 2961–2976. [102] Y. Zheng, W. Zhou, W. Yang, L. Liu, Y. Liu, Y. Zhang, Multivariate/minor
[74] Y. Dong, Y. Song, G. Wei, Efficient model-predictive control for networked fault diagnosis with severity level based on Bayesian decision theory and
interval type-2 T-S fuzzy system with stochastic communication protocol, IEEE multidimensional RBC, J. Process Control 101 (2021) 68–77.
Trans. Fuzzy Syst. 29 (2) (2021) 286–297. [103] J. Zhang, Q. Zhang, X. He, G. Sun, D. Zhou, Compound-fault diagnosis of
[75] D.G. Meyer, A new class of shift-varying operators, their shift-invariant equiva- rotating machinery: A fused imbalance learning method, IEEE Trans. Control
lents, and multirate digital systems, IEEE Trans. Automat. Control 35 (4) (1990) Syst. Technol. 29 (4) (2021) 1462–1474.
429–433. [104] Y. Zhao, X. He, J. Zhang, H. Ji, D. Zhou, M.G. Pecht, Detection of intermittent
[76] E. Fridman, A. Seuret, J.-P. Richard, Robust sampled-data stabilization of linear faults based on an optimally weighted moving average 𝑇 2 control chart with
systems: An input delay approach, Automatica 40 (8) (2004) 1441–1446. stationary observations, Automatica 123 (2021) 109298.
[77] J.-Y. Li, B. Zhang, R. Lu, Y. Xu, T. Huang, Distributed 𝐻∞ state estimator design [105] L. Ke, Y. Zhang, B. Yang, Z. Luo, Z. Liu, Fault diagnosis with synchrosqueezing
for time-delay periodic systems over scheduling sensor networks, IEEE Trans. transform and optimized deep convolutional neural network: An application in
Cybern. 51 (1) (2021) 462–472. modular multilevel converters, Neurocomputing 430 (2021) 24–33.
[78] J.-Y. Li, B. Zhang, R. Lu, Y. Xu, Robust distributed 𝐻∞ state estimation for [106] Y. Zhang, Z. Wang, L. Zou, Z. Liu, Fault detection filter design for networked
stochastic periodic systems over constraint sensor networks, IEEE Trans. Syst. multi-rate systems with fading measurements and randomly occurring faults,
Man Cybern.: Syst. 50 (11) (2020) 4396–4407. IET Control Theory Appl. 10 (5) (2016) 573–581.
[79] H. Huang, T. Huang, X. Chen, Guaranteed 𝐻∞ performance state estimation [107] Y. Zhang, Z. Wang, F.E. Alsaadi, Detection of intermittent faults for nonuni-
of delayed static neural networks, IEEE Trans. Circuits Syst. II 60 (6) (2013) formly sampled multi-rate systems with dynamic quantisation and missing
371–375. measurements, Internat. J. Control 93 (4) (2020) 898–909.
[80] H. Huang, T. Huang, X. Chen, Further result on guaranteed 𝐻∞ performance [108] J. Wang, Y. Alipouri, B. Huang, Dual neural extended Kalman filtering approach
state estimation of delayed static neural networks, IEEE Trans. Neural Netw. for multirate sensor data fusion, IEEE Trans. Instrum. Meas. 70 (2021) 6502109.
Learn. Syst. 26 (6) (2014) 1335–1341. [109] L. Li, T. Wang, Y. Xia, N. Zhou, Trajectory tracking control for wheeled mobile
[81] Y. Liu, B. Shen, H. Shu, Finite-time resilient 𝐻∞ state estimation for discrete- robots based on nonlinear disturbance observer with extended Kalman filter, J.
time delayed neural networks under dynamic event-triggered mechanism, Franklin Inst. B 357 (13) (2020) 8491–8507.
Neural Netw. 121 (2020) 356–365. [110] W. Li, G. Wei, F. Han, Y. Liu, Weighted average consensus-based unscented
[82] G. Hu, B. Gao, Y. Zhong, C. Gu, Unscented Kalman filter with process noise Kalman filtering, IEEE Trans. Cybern. 46 (2) (2015) 558–567.
covariance estimation for vehicular ins/gps integration system, Inf. Fusion 64 [111] W. Song, J. Wang, S. Zhao, J. Shan, Event-triggered cooperative unscented
(2020) 194–204. Kalman filtering and its application in multi-UAV systems, Automatica 105
[83] C. Yang, Z. Yang, Z. Deng, Robust weighted state fusion Kalman estimators for (2019) 264–273.
networked systems with mixed uncertainties, Inf. Fusion 45 (2019) 246–265. [112] Q. Chen, C. Yin, J. Zhou, Y. Wang, X. Wang, C. Chen, Hybrid consensus-based
[84] G. Wei, W. Li, D. Ding, Y. Liu, Stability analysis of covariance intersection- cubature Kalman filtering for distributed state estimation in sensor setworks,
based Kalman consensus filtering for time-varying systems, IEEE Trans. Syst. IEEE Sens. J. 18 (11) (2018) 4561–4569.
Man Cybern.: Syst. 50 (11) (2020) 4611–4622. [113] M.V. Kulikova, G.Y. Kulikov, SVD-based factored-form cubature Kalman filtering
[85] L. Liu, L. Ma, J. Guo, J. Zhang, Y. Bo, Distributed set-membership filtering for continuous-time stochastic systems with discrete measurements, Automatica
for time-varying systems: A coding-decoding-based approach, Automatica 129 120 (2020) 109110.
(2021) 109684. [114] M.S. Mahmoud, M.F. Emzir, State estimation with asynchronous multi-rate
[86] Z. Zhao, Z. Wang, L. Zou, J. Guo, Set-membership filtering for time-varying multi-smart sensors, Inform. Sci. 196 (2012) 15–27.
complex networks with uniform quantisations over randomly delayed redundant [115] L.P. Yan, B.S. Liu, D.H. Zhou, The modeling and estimation of asynchronous
channels, Internat. J. Systems Sci. 51 (16) (2020) 3364–3377. multirate multisensor dynamic systems, Aerosp. Sci. Technol. 10 (1) (2006)
[87] S. Chen, L. Ma, Y. Ma, Distributed set-membership filtering for nonlinear 63–71.
systems subject to round-robin protocol and stochastic communication protocol [116] M. Ghosal, V. Rao, Fusion of multirate measurements for nonlinear dynamic
over sensor networks, Neurocomputing 385 (2020) 13–21. state estimation of the power systems, IEEE Trans. Smart Grid 10 (1) (2019)
[88] X. Yin, J. Liu, Event-triggered state estimation of linear systems using moving 216–226.
horizon estimation, IEEE Trans. Control Syst. Technol. 29 (2) (2020) 901–909. [117] N. Davari, A. Gholami, Variational Bayesian adaptive Kalman filter for asyn-
[89] Q. Guan, G. Wei, L. Wang, Y. Song, A novel feature points tracking algorithm chronous multirate multi-sensor integrated navigation system, Ocean Eng. 174
in terms of IMU-aided information fusion, IEEE Trans. Ind. Inf. 17 (8) (2021) (2019) 108–116.
5304–5313. [118] M. Kordestani, M. Dehghani, B. Moshiri, M. Saif, A new fusion estimation
[90] S. Sun, Distributed optimal linear fusion estimators, Inf. Fusion 63 (2020) method for multi-rate multi-sensor systems with missing measurements, IEEE
56–73. Access 8 (2020) 47522–47532.
[91] W.-Q. Liu, X.-M. Wang, Z.-L. Deng, Robust centralized and weighted measure- [119] L.P. Yan, B.S. Liu, D.H. Zhou, Asynchronous multirate multisensor information
ment fusion Kalman estimators for uncertain multisensor systems with linearly fusion algorithm, IEEE Trans. Aerosp. Electron. Syst. 43 (3) (2007) 1135–1146.
correlated white noises, Inf. Fusion 35 (2017) 11–25. [120] Y. Hu, Z. Duan, D.H. Zhou, Estimation fusion with general asynchronous
[92] H. Geng, Y. Liang, Y. Liu, F.E. Alsaadi, Bias estimation for asynchronous multi-rate sensors, IEEE Trans. Aerosp. Electron. Syst. 46 (4) (2010) 2090–2102.
multi-rate multi-sensor fusion with unknown inputs, Inf. Fusion 39 (2018) [121] N.A. Carlson, Federated square root filter for decentralized parallel processors,
139–153. IEEE Trans. Aerosp. Electron. Syst. 26 (3) (1990) 517–525.
[93] F. Han, Z. Wang, H. Dong, H. Liu, Partial-nodes-based scalable 𝐻∞ -consensus [122] L.P. Yan, D.H. Zhou, M.Y. Fu, Y.Q. Xia, State estimation for asynchronous
filtering with censored measurements over sensor networks, IEEE Trans. Syst. multirate multisensor dynamic systems with missing measurements, IET Signal
Man Cybern.: Syst. 51 (3) (2021) 1892–1903. Process. 4 (6) (2010) 728–739.
[94] J. Li, G. Wei, D. Ding, E. Tian, Protocol-based 𝐻∞ filtering for piecewise linear [123] W.-A. Zhang, G. Feng, L. Yu, Multi-rate distributed fusion estimation for sensor
systems: A measurement-dependent equivalent reduction approach, Internat. J. networks with packet losses, Automatica 48 (9) (2012) 2016–2028.
Robust Nonlinear Control 31 (8) (2021) 3163–3178. [124] W.-A. Zhang, B. Chen, M.Z.Q. Chen, Hierarchical fusion estimation for clustered
[95] J. Sheng, T. Chen, S.L. Shah, Optimal filtering for multirate systems, IEEE Trans. asynchronous sensor networks, IEEE Trans. Automat. Control 61 (10) (2016)
Circuits Syst. II 52 (4) (2005) 228–232. 3064–3069.

26
Y. Shen et al. Information Fusion 82 (2022) 19–27

[125] F. Peng, S. Sun, Distributed fusion estimation for multisensor multirate systems [142] S. Liu, Z. Wang, G. Wei, M. Li, Distributed set-membership filtering for multirate
with stochastic observation multiplicative noises, Math. Probl. Eng. 2014 (2014) systems under the Round-Robin scheduling over sensor networks, IEEE Trans.
373270. Cybern. 50 (5) (2020) 1910–1920.
[126] L. Jiang, L.P. Yan, Y.Q. Xia, M.Y. Fu, Optimal distributed fusion for state [143] L. Orihuela, S. Roshany-Yamchi, R.A. García, P. Millán, Distributed set-
estimation in multirate wireless sensor networks, in: Processdings of the 34th membership observers for interconnected multi-rate systems, Automatica 85
Chinese Control Conference, Hangzhou, China, Jul. 28-30, 2015. (2017) 221–226.
[127] W. Li, S.L. Shah, D. Xiao, Kalman filters in non-uniformly sampled multirate [144] J. Hu, G.-P. Liu, H. Zhang, H. Liu, On state estimation for nonlinear dynamical
systems: For FDI and beyond, Automatica 44 (1) (2008) 199–208. networks with random sensor delays and coupling strength under event-based
[128] A. Gopalakrishnan, N.S. Kaisare, S. Narasimhan, Incorporating delayed and communication mechanism, Inform. Sci. 511 (2020) 265–283.
infrequent measurements in Extended Kalman Filter based nonlinear state [145] B. Qu, B. Shen, Y. Shen, Q. Li, Dynamic state estimation for islanded microgrids
estimation, J. Process Control 21 (1) (2011) 119–129. with multiple fading measurements, Neurocomputing 406 (2020) 196–203.
[129] L.P. Yan, B. Xiao, Y.Q. Xia, M.Y. Fu, State estimation for asynchronous multirate [146] H. Geng, Y. Liang, X. Zhang, Linear-minimum-mean-square-error observer for
multisensor nonlinear dynamic systems with missing measurements, Internat. J. multi-rate sensor fusion with missing measurements, IET Control Theory Appl.
Adapt. Control Signal Process. 26 (6) (2012) 516–529. 8 (14) (2014) 1375–1383.
[130] M. Enkhtur, S.Y. Cho, K.-H. Kim, Modified unscented Kalman filter for [147] X. Zheng, H. Zhang, Z. Wang, H. Yan, Almost surely state estimation for multi-
a multirate INS/GPS integrated navigation system, ETRI J. 35 (5) (2013) rate networked systems under random and malicious packet losses, J. Franklin
943–946. Inst. B 356 (17) (2019) 10593–10607.
[131] Y. Guo, B. Huang, State estimation incorporating infrequent, delayed and [148] Y. Shen, Z. Wang, B. Shen, F.E. Alsaadi, Nonfragile 𝐻∞ filtering for dis-
integral measurements, Automatica 58 (2015) 32–38. crete multirate time-delayed systems over sensor networks characterized by
[132] M. Mosalanejad, M.M. Arefi, UKF-based soft sensor design for joint estimation Gilbert-Elliott models, Internat. J. Robust Nonlinear Control 30 (8) (2020)
of chemical processes with multi-sensor information fusion and infrequent 3194–3214.
measurements, IET Sci. Meas. Technol. 12 (6) (2018) 755–763. [149] W. Qian, Y. Li, Y. Chen, W. Liu, 𝐿2 -𝐿∞ Filtering for stochastic delayed
[133] L. Armesto, J. Tornero, M. Vincze, On multi-rate fusion for non-linear sampled- systems with randomly occurring nonlinearities and sensor saturation, Internat.
data systems: Application to a 6D tracking system, Robot. Auton. Syst. 56 (8) J. Systems Sci. 51 (13) (2020) 2360–2377.
(2008) 706–715. [150] Y. Chen, Z. Wang, B. Shen, Q.-L. Han, Local stabilization for multiple input-
[134] R. López-Negrete, L.T. Biegler, A moving horizon estimator for processes with delay systems subject to saturating actuators: The continuous-time case, IEEE
multi-rate measurements: A nonlinear programming sensitivity approach, J. Trans. Automat. Control (2021) http://dx.doi.org/10.1109/TAC.2021.3092556,
Process Control 22 (4) (2012) 677–688. (in press).
[135] A. Liu, W.-A. Zhang, M.Z.Q. Chen, L. Yu, Moving horizon estimation for [151] L. Chen, Y. Chen, N. Zhang, Synchronization control for chaotic neural networks
mobile robots with multirate sampling, IEEE Trans. Ind. Electron. 64 (2) (2017) with mixed delays under input saturations, Neural Process. Lett. 53 (2021)
1457–1467. 3735–3755.
[136] L. Zou, Z. Wang, H. Dong, Q.-L. Han, Moving horizon estimation with multirate [152] Y. Zhang, Z. Wang, L. Ma, Variance-constrained state estimation for networked
measurements and correlated noises, Internat. J. Robust Nonlinear Control 30 multi-rate systems with measurement quantization and probabilistic sensor
(17) (2020) 7429–7445. failures, Internat. J. Robust Nonlinear Control 26 (16) (2016) 3507–3523.
[137] M. Elsheikh, R. Hille, A. T.-Codrean, S. Krämer, A comparative review of multi- [153] H. Li, M. Lyu, B. Du, Event-based multi-objective filtering for multi-rate time-
rate moving horizon estimation schemes for bioprocess applications, Comput. varying systems with random sensor saturation, Kybernetika 56 (1) (2020)
Chem. Eng. 146 (2021) 107219. 81–106.
[138] J. Bae, Y. Kim, J.M. Lee, Multirate moving horizon estimation combined with [154] C. Huang, B. Shen, Event-based fusion estimation for multi-rate systems subject
parameter subset selection, Comput. Chem. Eng. 147 (2021) 107253. to sensor degradations, J. Franklin Inst. B 358 (16) (2021) 8754–8771.
[139] Y. Gu, Y. Chou, J. Liu, Y. Ji, Moving horizon estimation for multirate systems [155] Y. Shen, Z. Wang, B. Shen, H. Dong, Outlier-resistant recursive filtering for
with time-varying time-delays, J. Franklin Inst. B 356 (4) (2019) 2325–2345. multi-sensor multi-rate networked systems under weighted Try-Once-Discard
[140] D. Haßkerl, M. Arshad, R. Hashemi, S. Subramanian, S. Engell, Simulation study protocol, IEEE Trans. Cybern. 51 (10) (2021) 4897–4908.
of the particle filter and the EKF for state estimation of a large-scale DAE-system [156] N. Li, Q. Li, J. Suo, Dynamic event-triggered 𝐻∞ state estimation for delayed
with multi-rate sampling, IFAC-PapersOnLine 49 (7) (2016) 490–495. complex networks with randomly occurring nonlinearities, Neurocomputing 421
[141] B. Yu, Y. Shi, H.N. Huang, 𝑙2 -𝑙∞ Filtering for multirate systems based on lifted (2021) 97–104.
models, Circuits Systems Signal Process. 27 (5) (2008) 699–711. [157] Q. Li, Z. Wang, J. Hu, W. Sheng, Simultaneous state and unknown input
estimation for complex networks with redundant channels under dynamic
event-triggered mechanisms, IEEE Trans. Neural Netw. Learn. Syst. (2021)
http://dx.doi.org/10.1109/TNNLS.2021.3070797, (in press).

27