IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO.

1, JANUARY 2018 3

Encoding-Decoding-Based Control and Filtering of

Networked Systems: Insights,
Developments and Opportunities
Zidong Wang, Fellow, IEEE, Licheng Wang, Shuai Liu, and Guoliang Wei

Abstract—In order to make the information transmission more information theory. From then on, constant research effort
efficient and reliable in a digital communication channel with from the researchers of different fields has been put on
limited capacity, various encoding-decoding techniques have been its developments and considerable progress has been made
proposed and widely applied in many branches of the signal
processing including digital communications, data compression, in the aspects of theory and application. Coincidentally, in
information encryption, etc. Recently, due to its promising ap- the same year, another celebrated mathematician N. Wiener
plication potentials in the networked systems (NSs), the analysis published his famous treatise “Cybernetics: Or Control and
and synthesis issues of the NSs under various encoding-decoding Communication in the Animal and the Machine” [133], which
schemes have stirred some research attention. However, because laid down the theoretical foundation for the servomechanisms
of the network-enhanced complexity caused by the limited
network resources, it poses new challenges to the design of (whether electrical, mechanical or hydraulic), automatic navi-
suitable encoding-decoding procedures to meet certain control or gation, analog computing, artificial intelligence, neuroscience
filtering performance for the NSs. In this survey paper, our aim and reliable communications, and made the birth of control
is to present a comprehensive review of the encoding-decoding- theory. Nevertheless, during the past few decades, the infor-
based control and filtering problems for different types of NSs. mation theory and control theory have almost been regarded
First, some basic introduction with respect to the coding-decoding
mechanism is presented in terms of its engineering insights, as two independent disciplines with few common features and
specific properties and theoretical formulations. Then, the recent connections due to their different concerns.
representative research progress in the design of the encoding- Broadly speaking, the information theory mainly focuses
decoding protocols for various control and filtering problems on how to obtain, store, process and transmit the data but
is discussed. Some possible further research topics are finally seldom pays attention to the specific usage of the data itself.
outlined for the encoding-decoding-based NSs.
On the contrary, the control theory concerns how to make use
Index Terms—Encoding-decoding-based communication mech- of the obtained data to achieve the prescribed tasks through
anism, encoding-decoding-based control, encoding-decoding- certain strategies. For general systems, the most basic but
based state estimation, networked systems.
important idea to realize desirable system performances is
the “feedback”, which means the obtained signals (usually the
I. I NTRODUCTION system outputs) should be fed back to the systems’ inputs in
A. Engineering Insights certain manners by designing proper feedback mechanisms.
However, in most cases, such an idea is usually implemented

I T has been well recognized that the publication titled

“A Mathematical Theory of Communication” [112], [113]
by C. E. Shannon in 1948 marked the establishment of the
upon the assumption that the signal transmission between
the components is with infinite precision, namely, the signals
can be perfectly transmitted and utilized to contribute to the
Manuscript received September 16, 2017; accepted November 18, 2017. feedback schemes.
This work was supported in part by the Royal Society of the UK, the National In recent years, owing to the rapid development of the
Natural Science, Foundation of China (61329301, 61374039), the Program for networked communication technology, the networked systems
Capability Construction of Shanghai Provincial Universities (15550502500),
and the Alexander von Humboldt Foundation of Germany. Recommended by (NSs) have emerged as a new yet popular system architecture
Associate Editor Hongyi Li. (Corresponding author: Zidong Wang.) with an extensive utilization in the industry for its low power
Citation: Z. D. Wang, L. C. Wang, S. Liu, and G. L. Wei, “Encoding- cost, simple installation, easy maintenance and high reliability.
Decoding-Based control and filtering of networked systems: insights, de-
velopments and opportunities,” IEEE/CAA J. of Autom. Sinica, vol. 5, no. 1, Different from the non-networked systems where the data
pp. 3−18, Jan. 2018. flows through the components in a point-to-point way, the
Z. Wang is with the Department of Computer Science, Brunel Univer- NSs mean that all the data communications between the com-
sity London, Uxbridge, Middlesex, UB8 3PH, United Kingdom. (e-mail:
[email protected]). ponents are executed via a commonly shared communication
L. Wang and G. Wei are with the Department of Control Sci- network (channel). Although the NSs possess various merits
ence and Engineering, University of Shanghai for Science and Technol- mentioned above, the introduction of the communication net-
ogy, Shanghai 200093, China. (e-mail: [email protected]; guo-
[email protected]). work also brings about some new and noteworthy problems
S. Liu is with the Business School, University of Shanghai for Science and including 1) how the network-based data exchange affects the
Technology, Shanghai 200093, China. (e-mail: [email protected]). system performance, and 2) how to maximize the utilization of
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. limited network bandwidth while ensuring the required system
Digital Object Identifier 10.1109/JAS.2017.7510727 performance.
4 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

H, G(·) and xd (k) be the original signal before being encoded,

the encoding mapping, the encoding alphabet, the decoding
mapping, and the decoded signal, respectively, where k is
the discrete time step and the H , {s1 , s2 , . . . , sm } is a set
whose elements si (i = 1, 2, . . . , m) are certain deterministic
codewords. In most existing literature, the authors’ attempts
have always been made to design a pair of encoder-decoder
appropriately such that at each time instant k, the decoding
error ed (k) , xd (k) − x(k) is possibly small at the minimum
cost of the network bandwidth.
In view of the aforementioned discussion, it is obvious that
there are two important performance indices to evaluate the
designed encoding-decoding procedure, namely, the decoding
Fig. 1. Structure of the general communication system. error and the bandwidth condition. Intuitively, the large net-
work bandwidth would result in a small decoding error. For
If the network environment is ideal, that is, the bandwidth of
the NSs, if the network resource is not the concern (e.g., the
the network is sufficiently large, the assumption of the perfect
bandwidth can be sufficiently large), the decoding error would
data transmission would make sense and it is reasonable to
probably be very small and hence has little impact on the
ignore the influence from the network communication on
system performance. In this case, the coding-decoding-based
the system performance. However, in the situations that the
NSs reduce to the traditional NSs for the analysis and synthesis
network resources are subject to certain constraints (e.g.,
issues. However, in most practical situations, due to the limited
limited bandwidth of the channel), it is no longer proper to
network resources, the network bandwidth serving as a major
entirely neglect the distortion of the transmitted data caused by
concern should be incorporated into the system design. In this
the limited communication capacity of the network. Especially,
sense, for the bandwidth-constrained systems, the concept of
when it comes to the large-scale NSs such as wireless sensor
“minimum bit rate” is proposed to characterize the smallest
networks (WSNs), multi-agent systems (MASs) and smart
needed channel capacity for the codeword transmission, above
grids, the limited network resources impose severe restrictions
which the performance of the controlled (or filtering error) sys-
on the data communication quality and even deteriorate the
tems is explored including asymptotical/exponential stability,
overall system performances. On the other hand, from the
ultimately bounded stability, input-to-state stability (ISS), etc.
standpoint of the network security, the NSs are usually fragile
In the past few years, a rich body of research results
and vulnerable to be attacked, e.g., the data flowing through
with respect to the minimum bit-rate-based control/filtering
the system components are easily stolen by the adversaries for
problem for both linear and nonlinear NSs have appeared.
some hostile purposes, see [15], [39], [76], [143]. Therefore,
Among them, it has been shown in [120] and [91] that for
in order to make the data transmission of the NSs more
linear systems, the minimum
P bit rate R (bits/sample) should
efficient and secure via a resource-constrained communication 1
satisfy R > log(2) Re(λi (A))>0 (λi (A)), where log is the
network, the encoding-decoding technique stands out as an
base-e logarithm, Re(·) denotes the real part of a complex-
ideal candidate for its advantage in data compression and
valued scalar, A is the system matrix, and λi {A} stands for
encryption.
the ith eigenvalue of matrix A. Furthermore, the literature [97]
B. Theoretical Outlines revealed that in order to globally (or semiglobally) stabilize
Normally, as shown in Fig. 1, a classic communication the n−dimensional nonlinear systems with a feedforward
system always contains three basic parts: the information structure, the necessary bit rate should be n + 1 (or n).
source, the information channel and the information sink, Compared with the traditional NSs, the introduction of the
based on which, the encoding-decoding technique can be encoding-decoding communication mechanism would make
accordingly divided into two main categories: the source the dynamics of the NSs more complicated and diverse, which
encoding-decoding and the channel encoding-decoding. In this accordingly imposes certain challenges on the analysis and
paper, our attention will be specifically focused on the source synthesis issues. Moreover, because of the inevitable data
encoding-decoding technique, which aims to transform the distortion resulted from the encoding-decoding process, the
original data into certain codewords. Generally speaking, the traditional approaches cannot be directly generalized to cope
source information encoding technique possesses the following with the performance analysis issues of encoding-decoding-
three prominent features: 1) realizing the data compression, 2) based NSs. In this case, the influence of the encoding-decoding
facilitating the digital data transmission, and 3) enhancing the mechanism on the system performance should be well taken
data security. From the aspect of mathematics, the process of into consideration. In other words, how to choose suitable
encoding can be regarded as a mapping that maps the original theoretical tools to quantitatively evaluate the relationship
data to certain codewords. Correspondingly, the decoding between the system dynamics and the encoding-decoding
process can be viewed as the “inverse” operation which intends communication protocol appears to be particularly important.
to recover the original data as exactly as possible by excavating To date, there are generally three theoretical methodologies to
the information contained in the codewords. To be more spe- deal with the encoding-decoding-based analysis and synthesis
cific, in the networked communication systems, let x(k), F(·), problems for the NSs, namely, the information-theory-based
WANG et al.: ENCODING-DECODING-BASED CONTROL AND FILTERING OF NSS: INSIGHTS, DEVELOPMENTS AND OPPORTUNITIES 5

Fig. 2. Framework of surveyed contents.

methodology, the ISS-based methodology and the hybrid- are presented. The main contents that are reviewed in this
system-based methodology. paper and the framework are shown in Fig. 2.
The theoretical basis of the information theory methodology
is the celebrated Shannon information theory, based on which, II. M AIN ENCODING - DECODING SCHEMES IN NS S
all information including the original data, codewords and de- In the framework of the coding-decoding-based data com-
coded data are described as random variables [23], [24], [92]. munications for NSs, it is of crucial importance to construct
Within such a framework, the main objective is to find general a suitable pair of encoder and decoder which could greatly
necessary conditions for the channel capacity measured by restrain the impacts from the resulted data distortion on the
the concept of “entropy”, under which the original data can system performance. To this end, we are going to give a clear
be reconstructed from the decoded data with the acceptable look on some commonly utilized encoding-decoding schemes
distortion. Then, by resorting to the decoded data, the desired with the following system model
control or filtering performance can be guaranteed. On the
other hand, for most of the effective encoding schemes, a x(t + 1) = Ax(t) + Bu(t) + Dw(t) (1)
basic requirement is that the decoding error ed (k) should be or
bounded or even convergent. Thus, it is natural to view the ẋ(t) = Ax(t) + Bu(t) + Dw(t) (2)
decoded error as the bounded disturbance of the dynamical
system, and the ISS-based approach [69], [98] has been pro- with the measurement output
posed as a theoretical tool to handle the analysis and synthesis
y(t) = Cx(t) + Ev(t) (3)
issues for these systems. For the continuous-time systems with
a digital transmission manner, the received codewords by the where x(t) ∈ Rnx , u(t) ∈ Rnu , and y(t) ∈ Rny are,
decoder at a certain discrete instant can be seen as an impulse respectively, the system state, the system input, and the
sequence that affects the decoder dynamics. Consequently, the measurement output. w(t) ∈ Rnw and v(t) ∈ Rnv are
hybrid-system-based approach [68], [70] is valid for analyzing the process and the measurement noises. A, B, C, D, and
such a kind of systems’ dynamical behaviors. E are the known matrices with compatible dimensions. For
In this paper, our purpose is to offer an up-to-date review of analysis convenience, in this paper, it is assumed that the
existing research results on encoding-decoding-based control system state x(t) is accessible for the stabilization problem,
and filtering problems for the NSs. Various usually adopted while for the filtering issue, it is supposed that only the
design schemes of the encoding-decoding procedure in the measurement output y(t) is observable. Till now, in the
NSs and their individual characteristics are reviewed. Sub- existing literature, two representative design solutions of the
sequently, recent advances on the encoding-decoding-based encoding-decoding procedure have been proposed to deal
control and filtering problems for the NSs are, respectively, with the control and filtering problems in the NSs, namely,
surveyed. Then, the control and filtering issues for distributed the quantization-based encoding-decoding technique and the
systems with encoding-decoding communication scheme are symbolic-based encoding-decoding technique. Several typical
discussed. Finally, conclusions and future research directions encoding-decoding schemes in the NCSs are exhibited in
6 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

Fig. 3. In the following, we are in a position to successively il- of the sub-hyperrectangle Is11 (a) × Is22 (a) × · · · × Isnn (a) can
lustrate these encoding-decoding design approaches mentioned be defined as
above.
ηa (s1 , s2 , . . . , sn ) , [b1 b2 · · · bn ]T (5)

where bi = −a + (2si q−1)a , i = 1, 2, . . . , n.

In what follows, we employ the proposed encoding-
decoding protocol to the continuous-time system (2) for the
stabilization problem. To be more specific, our goal is to
design a pair of encoder-decoder and certain decoded-signal-
based control strategy such that the closed-loop system is
asymptotically stable. For simplicity of the illustration, we
assume that the disturbance w(t) in (2) satisfies w(t) ≡ 0
Fig. 3. The summarization of several typical encoding-decoding schemes in and the main idea for the design of the encoding-decoding-
NCSs. based control strategy can be outlined with the following steps.
To start with, at each encoding time jT (j = 1, 2, . . . and
“T ” is the encoding period), by calculating the error e(jT )
A. Quantization-Based Encoding-Decoding Mechanism between the system state x(jT ) and the decoded state x̂(jT )
which will be defined later, where e(jT ) = x(jT ) − x̂(jT − ),
The quantization-based encoding-decoding method is one we are able to determine the uniform quantization param-
of the most common methods for the NSs with the limited eter “a” and thus partition the hyperrectangle Ba into q n
communication capacity, based on which, a number of coder- equal sub-hyperrectangles. Here, x̂(jT − ) denotes the limit
decoder structures have been presented in the existing litera- of x̂(jT ) at the point of jT from the left, i.e., x̂(jT − ) ,
ture, see [47], [56], [65], [111], [148]. In addition, according lim²>0,²→0 x̂(jT − ²). Subsequently, generate the correspond-
to the case whether the quantizer parameters can be dynam- ing codeword sequence {s1 , s2 , . . . , sn } according to the re-
ically adjusted, the quantizer can be classified into the static gion where the error “e(jT )” belongs to and then transmit the
quantizer and the dynamic one. In the following, in terms codeword sequence {s1 , s2 , . . . , sn } to the decoder to obtain
of the type of the used quantizer, three typical coder-decoder the decoded state x̂(jT ). Finally, design the controller “u(jT )”
structures are listed and their corresponding characteristics are in terms of the decoded state x̂(jT ) to stabilize the original
summarized. system.
1) Dynamic-quantization-based encoding-decoding scheme Based on the above analysis, the detailed form of the
I: Such a method was proposed in [111] and the similar idea encoder and decoder is given as follows:
appeared in the earlier literature [67]. For the sake of making Coder: For e(jT ) ∈ Is11 (a(jT )) × Is22 (a(jT )) × · · · ×
n
this approach more understandable, first, let us briefly recall Isn (a(jT )) ⊂ Ba(jT ) , we have
some basic concepts of the uniform quantization, and then we g(e(jT )) = {s1 , s2 , . . . , sn } (6)
will show how the uniform quantization approach is applied
to the encoder-decoder procedure. where a((j +1)T ) = ϕ(T, q, x(0))a(jT ) for j = 1, 2, . . . with
Let the scaling parameter (the length of the quantization ϕ(T, q, x(0)) being the function of certain form with respect
interval) a > 0 and integer q be given, and the Ba denotes to the encoding period T , the quantization level q and the
the hyperrectangle in Rn centered at 0 with the edge length initial condition x(0).
a, i.e., Ba = {δ ∈ Rn : |δ|∞ ≤ a}, where | · |∞ is the infinite Decoder:

vector norm. Then, we uniformly partition the hyperrectangles  x̂(0) = 0


Ba into q n sub-hyperrectangles Is11 (a) × Is22 (a) × · · · × Isnn (a),  x̂(t)
˙ = Ax̂(t) + Bu(t), ∀t ∈ [jT, (j + 1)T )
where s1 , s2 , . . . , sn ∈ {1, 2, . . . , q} and (7)
½ ¾ 

 x̂(jT ) = x̂(jT − ) + ηa (s1 , s2 , . . . , sn )
2a 
I1i (a) , δi | − a ≤ δi < −a + u(t) = K x̂(t).
q
½ ¾ From structures of the coder (6) and the decoder (7), we
2a 4a
I2i (a) , δi | − a + ≤ δi < −a + (4) have the following observations. 1) The proposed encoding
q q approach is a kind of differential encoding technique. It is
.. obvious that, in (6), only the decode error e(jT ) is encoded at
.
½ ¾ the encoding time jT . In fact, by using the differential quan-
2a
Iqi (a) , δi | a − ≤ δi ≤ a tization encoding technique to encode the signal difference,
q the amount of the information to be sent is greatly reduced,
where δi is the ith element of the vector δ. Here, each sub- see [22]. 2) It can be seen from the iteration relationship
hyperrectangle Is11 (a) × Is22 (a) × · · · × Isnn (a) is labeled with a between a((j + 1)T ) and a(jT ) that the quantization region
sequence {s1 , s2 , . . . , sn }, which represents the specific region dynamically varies at each encoding instant, and such a
of the sub-hyperrectangle and will be sent to the decoder for quantizer is referred to as the dynamic quantizer. 3) In order
the reconstruction of the original data. Moreover, the center to enforce the decoding error be convergent, it is required
WANG et al.: ENCODING-DECODING-BASED CONTROL AND FILTERING OF NSS: INSIGHTS, DEVELOPMENTS AND OPPORTUNITIES 7

that ϕ(T, q, x(0)) < 1, from which, we can see that the the multiple of the system sampling period, see [111], [136],
convergence of the encoding-decoding algorithm is closely [148].
associated with the encoding period, the quantization level and Such a “zooming in/zooming out” quantization-based
the initial condition of the system. encoding-decoding scheme was first proposed in [5], under
Apart from the continuous-time case, for the discrete-time which, a consensus algorithm has been designed for a class of
counterpart, the encoding-decoding communication strategy discrete-time MASs that the information exchange among the
for the NSs has been considered in [148], and the correspond- agents was implemented through a digital communication net-
ing extended control/filtering problems subject to the network- work. Recently, some extensions have emerged, see [54], [57],
induced phenomena have been investigated in [126], [136]. [64], [65], [140], however, most of which are concerned with
2) Dynamic-quantization-based encoding-decoding scheme the discrete-time systems and the results for the continuous-
II: This kind of encoding-decoding scheme is based on the time counterpart are relatively few.
so-called “zooming in/zooming out” quantization technique 3) Static-quantization-based encoding-decoding scheme:
which was proposed in [134]. For the discrete-time system Having presented the above two different kinds of encoding-
(1) with w(t) ≡ 0, the following encoder is given by decoding schemes, now we would like to give a simple
 discussion about another encoding-decoding scheme, namely,

 ξ(0) = 0 the multiple description (MD) encoding-decoding scheme. The


ξ(t) = g(t − 1)s(t) + ξ(t − 1) original idea of the MD encoding-decoding scheme comes
µ ¶ (8)

 1 from the investigation of the audio encoding issue in the tele-

 s(t) = Q (x(t) − ξ(t − 1))
g(t − 1) phone network by Bell Laboratories in the 1970s. As the rapid
development of the network-based (e.g. internet) communica-
where ξ(t) is the encoder’s internal state, s(t) is the generated tion techniques, in recent decades, the MD encoding-decoding
codeword which needs to be sent to the decoder, Q(·) is a technology has gained considerable research attention from
finite-level uniform quantizer, and g(t) is a scaling function many researchers in different disciplines.
which is used to adjust the amplitude of the signal to be The idea of the MD encoding-decoding scheme is based
quantized. In most of the literature, the quantizer Q(·) is on the diversity principle which could effectively enhance the
chosen as the symmetric one with the following form transmission capacity and reliability of the channels, see [18],


 0, 1/2 ≤ z < 1/2 [30], [47], [121], [122]. From the point of the information


 theory view, the diversity principle can be understood as fol-
 i,
 (2i − 1)/2 ≤ z < (2i + 1)/2
lows: encode the information source into multiple descriptions
Q(z) = i = 1, 2, . . . , L − 1 (9)

 with the identical importance and then sent to the decoder

 L, z ≥ (2L − 1)/2

 via the parallel individual channel. The framework of the

− Q(−z), z ≤ −1/2 MD encoding-decoding scheme is built on the following two
assumptions: 1) there are multi-channels between the encoder
where L is the quantizer’s saturation value and the associated
and the decoder and all the channels failing to work simultane-
quantization level is 2L + 1.
ously is a small probability event, and 2) the original signal can
By using the generated codeword s(t), the decoder Ψ is
be reconstructed with an acceptable quality as long as at least
designed as
( one of the descriptions is successfully delivered to the decoder,
x̂(0) = 0 and when more descriptions are received by the decoder, the
(10) signal is reconstructed with a higher quality. Therefore, such
x̂(t) = g(t − 1)s(t) + x̂(t − 1)
an approach is able to improve the transmission reliability by
where x̂(t) is the output of the decoder, namely, the decoded rejecting the channel fluctuations resulted from the fading [37],
value. shadowing, crosstalk, etc.
It can be concluded from the structure of the above encoder- So far, according to the adopted method in the encoding-
decoder that 1) although the saturation value L is fixed a decoding process, the multiple description encoding-decoding
priori, due to the introduction of the scaling function g(t), can be classified into various classes including multiple
the signal to be quantized is enforced to be pre-processed description quantization coding (MDQC), the multiple de-
first by the “zooming out” or “zooming in” technique, and scription transformation coding (MDTC), the forward-error-
hence this type of encoding approach is still of the dynamic- correction-based multiple description coding (FECMDC), etc.
quantization-based encoding, 2) as discussed above, encoding In this paper, our attention is focused on the MDQC where the
the difference between x(t) and ξ(t − 1) with scaling rather static uniform quantization method is applied to the addressed
than the system state x(t) itself possibly reduces the bits which encoding-decoding issue.
are utilized to represent the codeword, 3) the saturation value Loosely speaking, the static uniform quantizer-based multi-
L is closely related to the network bandwidth, that is, a small ple description coding procedure can be divided into two steps.
L results in the requirement of a small network bandwidth. The first step is the index generation process which mainly
Especially, when L = 1, it means that only one bit is necessary employs the scalar quantizer to transform the information
to send the codeword, and 4) the encoding period is consistent source into certain indices. Then, represent the indices by
with the system sampling period, which is different from the multiple descriptions in term of a proper index assignment
former encoding approach that the encoding period can be strategy. For convenience, a brief view on the encoding scheme
8 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

with two descriptions is provided where the measurement (dynamic) encoding technique [51], [70]. Since the dynamic
output (3) is treated as the information source to be encoded. binary encoding technique is more complicated than the static
The diagram of the MDC for the NSs with two descriptions one, in the following, we will concentrate our attention on the
is shown in Fig. 4. dynamic binary encoding-decoding technique.
As we know, such a kind of dynamic encoding scheme was
first proposed in [51], which is highly dependent on the change
of certain system parameters. The core of the encoding rule
can be expressed as follows with the system model (1) and
(3).
Encoder : s(tk ) = sign(y(tk ) − y(tk−1 ))
(13)
Decoder : ȳ(tk ) = g(s(tk ))

Fig. 4. Multiple description encoding-based data communication in NSs.

where sign(·) is the sign function, g(·) is the decoding
function, s(tk ) is the codeword generated at tk , and ȳ(tk )
Encoder: is the decoded value of the measurement output y(tk ). It
( follows from (13) that at each encoding instant tk , only the
θl (t) = f1l (yl (t)), l = 1, · · · , ny
(11) sign of the change between the two successive measurement
ϑl (t) = f2l (yl (t)), l = 1, · · · , ny outputs is transmitted to the decoder, which implies that
Decoder: only one bit is needed for the codeword delivery during
 each encoding period. For the decoder design, the expected
 g1l (θl (t)),
 when λ1 (t) = 1, λ2 (t) = 0 decoding function g(·) is generally consisted of two parts,
ȳl (t) = g2l (ϑl (t)), when λ1 (t) = 0, λ2 (t) = 1 namely, g(s(tk )) = ȳ(tk−1 ) + g0 (s(tk )). The first term is the


gcl (θl (t), ϑl (t)), when λ1 (t) = 1, λ2 (t) = 1 decoded value at the last encoding instant and the second one
(12) is a compensation function which aims to predict the differ-
ence between y(tk ) and y(tk−1 ). It is obvious that such an
where f1l (·) and f2l (·) are two coding functions whose outputs encoding scheme greatly reduces the communication burden
θl (t) and ϑl (t) can be regarded as the individual descriptions and hence becomes a good candidate for the network-based
of the source yl (t) with yl (t) being the lth component of control/filtering systems subject to communication constraints.
y(t). g1l (·) and g2l (·) are two side decoding functions and On the other hand, since only one-bit information could be
gcl (·, ·) is the central decoding function. λi (t) (i = 1, 2) are obtained at each decoding instant, how to design an effective
two independent random variables characterizing the packet prediction function g0 (s(tk )) so as to ensure the convergence
dropout phenomena during the transmissions of the individual of the decoding error appears to be the main challenge in the
description, which satisfy the Bernoulli binary distribution design of the encoder-decoder procedure.
taking values on either 1 or 0 with mathematical expectations
λ̄i and variances σi2 , respectively.
The early research effort on the MDC was mainly devoted III. E NCODING -D ECODING -BASED C ONTROL AND
to its theoretical research. Since the scalar quantization-based F ILTERING P ROBLEMS
MDC method was first proposed in [121], the research direc- In this section, we will go over the recent theoretical
tion has been moved from the pure theoretical investigation to developments of both the encoding-decoding-based control
its practical applications. From then on, the MDC has experi- and filtering problems from various aspects including con-
enced a huge development in the theory and applications, and troller/filter design and the system performance evaluation.
myriad coding schemes have appeared. It should be pointed
out that most results of the MDC have been centered on
A. Encoding-Decoding-Based Stabilization Problems
the applications in distributed storage systems [48], diversity
communication system (antenna diversity) [106], image/au- 1) Linear systems: Among the various system performance
dio/video encoding [2], [4], [9], [137]. However, to the best of indices for the networked control systems (NCSs), stability
the authors’ knowledge, the relevant results of the MDC with is the most basic yet crucial one, and thus it needs to be
respect to the control and filtering problems of NSs are very first considered in the system design. However, it should be
few. noted that the introduction of the encoding-decoding-based
communication scheme inevitably leads to the occurrence of
the signal distortion, which would pose considerable additional
B. Symbolic-Based Encoding-Decoding Mechanism complexities on the stability analysis and controller synthesis
This kind of encoding schemes, also called as the “binary issues. A schematic structure of the NCSs with encoding-
encoding”, mainly takes the sign of the source into account, decoding-based data transmission protocol can be described
and thus only one bit is utilized to encode the signal. For in Fig. 5.
convenience, in the sequel, we call it as the binary encoding In [11], the pioneering work has been reported on the NCSs
instead of the symbolic-based encoding. Usually, the binary with quantization-based state feedback control strategy, which
encoding technique including two versions, that is, the abso- plays an important role in the subsequent fruitful results on
lute (static) encoding technique [27], [28] and the differential the encoding-decoding schemes. In [134], the stabilization
WANG et al.: ENCODING-DECODING-BASED CONTROL AND FILTERING OF NSS: INSIGHTS, DEVELOPMENTS AND OPPORTUNITIES 9

issue for a class of linear time-invariant system subject to Both the detectability and stabilizability problems have been
the bandwidth constraint has been considered, where the investigated in [111] for the continuous-time systems with
encoding-decoding-based communication protocol has been Lipschitz nonlinearities and bounded noises.
introduced. For the quantization-based encoding scheme, a With the help of “zooming” technique, an observer-based
sufficient condition for the existence of the bit rate which encoding-decoding procedure has been put forward to globally
enables to stabilize the system has been derived. Moreover, stabilize the closed-loop system without the knowledge of the
when the system reduces to a scalar one, a tight sufficient bound of the initial condition. Motivated by the approach
and necessary condition has been obtained. Nevertheless, the proposed in [111], the subsequent work has been done in
concrete forms of both the encoder and decoder are not a large body of literature. For example, [148] has applied
explicitly given in their paper. The quadratic stabilization the approach to the discrete-time descriptor systems, and
problem for sampled-data systems has been discussed in [42] the nonlinearities existing in both the system state and the
under a bandwidth-constrained situation, where the memory- measurement equations have been considered in [7]. Recently,
less uniform and logarithmic quantizers have been designed the encoding-scheme-based stabilization issue for the non-
to overcome the circumstances of the limited bit rate of the linear uncertain systems has been studied in [66], and the
communication channel and an upper bound on the bit rate has traditional ISS condition has been relaxed to the differentiable
been acquired to achieve the stabilization purpose. Further- ISS one. Moreover, by proposing a new uncertain nonlinear
more, both the state encoding and output encoding schemes system model, the results obtained in [69] have been further
have been developed in [67] for the continuous-time linear generalized.
systems, and the encoding procedure has been well described
in terms of the uniform quantization method. Meanwhile, the
B. Encoding-Decoding-Based State Estimation Problems
quantitative relationship between the number of the values
taken from the encoder and the open-loop system parameters In order to eliminate the confusion, it should be stated that
has been derived, under which the globally asymptotic stability the state estimation problem considered here is different from
of the closed-loop system has been guaranteed. the channel state estimation issue [81], [146] which belongs to
the scope of information theory and aims to recover the state
of the channel to facilitate the information transmission. In the
category of control and signal processing, the state estimation
(also called filtering) problem is to provide a possible accurate
reconstruction of the state of the target plant, and has long been
a focus of research due to its engineering insights in many
branches, see [36], [38], [41], [55], [62], [63], [79], [80], [82],
[83], [115], [116], [124], [131], [144], [145], [149]. For the
data-rate-constrained NCSs, a typical filter structure with the
encoder-decoder pair is shown in Fig. 6. In this subsection, an
Fig. 5. NCSs with the encoding-decoding-based communication protocol. overview will be given on the progress made in the encoding-
decoding-based filtering/state estimation problems in the past
2) Nonlinear systems: Afterwards, a number of research few years.
results have been extended to the encoding-decoding issues for
the nonlinear NCSs, see [44], [66], [67], [69], [93], [97]−[99],
[111], [148]. In [69], by utilizing the decoded system state,
a feedback control strategy has been proposed to stabilize a
class of continuous-time Lipschitz-type nonlinear systems. The
ISS theory has been employed to characterize the impact of
the decoding error on the closed-loop system state, based on
which, it has been demonstrated that the closed-loop system
can achieve globally asymptotical stability as long as the
encoding frequency, the amount of the information provided
by the encoder and the system parameters satisfy certain Fig. 6. Filtering system model with encoding-decoding-based data transmis-
inequality conditions. In [98], the integral ISS condition, which sion scheme.
is a weaker assumption than the ISS, has been obtained to
ensure the stability of the closed-loop system under the same Some initial attempts devoted to the state estimation prob-
requirement on the channel capacity with [69]. Then, [96] lems with the compressed measurement data can be found
extended the result to any smooth nonlinear systems, in which, in [3], [10], and then a rich body of related results have
it has been revealed that if the original nonlinear systems appeared due to the wide utilization of the prevalent network
can be stabilized by certain control strategy without encoding communication techniques in a large range of the practical
process, it can also be stabilized by the state feedback scheme applications, see [8], [26], [87], [88], [114]. For instance,
with the designed encoding communication protocol provided the state estimation problem has been studied in [87] for the
the channel capacity is larger than a given lower bound. linear discrete-time unstable system with a noisy memoryless
10 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

channel of communication constraints. The estimator has been sense, for the encoding-decoding-based control/filtering prob-
explicitly designed and an equivalent relationship between lem in NCSs, the inherent complexities of the analysis and
estimation performance and the communication capacity of design issues have been typically enhanced and a number of
the channel has been revealed. For the same system model, relevant results have been reported in [19], [21], [31], [43],
its almost sure observability/stabilizability has been discussed [45], [77], [100], [123], [129], [132].
in [88], where the communication channel was noisy and 1) Time-delays: In [100], the minimal data-rate stabiliza-
subject to the limited transmission capacity. It has been proved tion problem has been solved for the nonlinear NSs with trans-
that the capability of the channel to guarantee the almost mission delays, which can be arbitrarily large. By applying
sure stabilizability/observability of the plant corresponds to the tool of the hybrid system theory, the original system has
its capability to deliver information with zero probability of been converted into a couple of impulsive delayed nonlinear
error. The communication-constrained remote state estimation systems, and the semi-global and local stability of the closed-
problem has been addressed in [8], where both the source system has been guaranteed. The stabilization problem has
coding and channel coding have been concerned. Without the been concerned in [21] for a class of distributed linear systems
assumption of the channel feedback existing in the communi- by a remote centralized control scheme, in which the random
cation loop, two encoding approaches have been proposed, and packet dropout and communication delay caused by an erasure
then the analysis results on the performance trade-off between network have been simultaneously considered. Three scenarios
the convergence rate of the decoding error and computational of the availability of the feedback channel have been, respec-
complexity have been derived. tively, discussed and different encoding strategies have been
It is noted that the results on the encoding-decoding-based correspondingly provided to ensure the closed-loop stability in
state estimation problems recalled above are generally based the mean absolute sense. In [31], the channel coding technique
on the information theoretic aspects, while, in recent years, has been adopted to reduce the occurrence of packet dropouts
there also has a surge of research attention on this issue from for the NCSs. However, the redundancy of the channel coding
the point of non-information-theoretic view, see [34], [86], would lead to the time-delays. In order to balance such a trade-
[92], [117], [138], [139]. In [86], the robust set-valued state off, the ON-OFF channel coding technique has been utilized
estimation problem has been addressed for the continuous-time where the coding operation was adaptively implemented ac-
uncertain systems via a digital communication with limited cording to the expectation value of the tracking error. The data-
bit rate, where two encoding-decoding-based state estimation rate-constrained stabilization problem has been investigated in
schemes have been presented from the aspects of the estima- [123] for a class of continuous-time linear system. Due to the
tion precision and the algorithm implementation, respectively. limited communication capacity of the channel between the
The concept of “maximin information functional” has been sensor and controller, the sampled data has been encoded to
established in [92] under a nonstochastic information theory certain codewords by the spherical polar coordinates encoding
framework, by which, it has been verified that the largest scheme. Dependent on whether the control inputs were needed
maximin information rate through a memoryless and error- to be sent to the decoder, two coding methods have been
prone channel reflects the zero-error capacity of the channel developed to achieve the stability of the controlled system.
and the result has been applied to the state estimation issue for 2) Packet-dropouts: For the randomly occurring message
a linear time-invariant system. By applying the linear temporal loss during the data transmission, a subband coding technique
coding technique, the optimal state estimation problem has has been employed in [43] to deal with the controller and
been investigated in [34] for a class of discrete-time system encoder-decoder pair design issues. To facilitate the analysis
with an unreliable channel. Both the orthogonal projection of the encoding-decoding procedure, the addressed control
principle and the innovation sequence method have been problem has been tackled in the frequency domain, and
employed to derive the minimum mean square estimation al- the stochastic stability and the H∞ performance have been
gorithm. The extended work has been done in [117], where the ensured in terms of the solution to certain linear matrix
state estimation problem for the linear discrete-time stochastic inequalities (LMIs). In [45], the attention has been focused on
systems subject to the packet dropout has been considered. In the stabilization problem for the discrete-time linear NCSs,
order to compensate the negative effect of the packet loss on where the data transmission has been implemented via a
the estimation performance, a linear coding temporal scheme lossy channel with finite data rate. For the aim of the data
has been employed to the raw measurements, by which some compression, the control signal has been encoded to finite
necessary and sufficient conditions for the estimation error values by the uniform quantization method and a buffer has
system have been established in the mean square sense. been designed in order to compensate the packet-dropout
effect.
For the sake of improving the data transmission reliability
C. Encoding-Decoding-Based Control / State Estimation Sub- via the erasure channel and obtaining a relatively satisfac-
ject to Network-Enhanced Complexities tory control/estimation performance, the multiple description
In the remote networked control/estimation systems, be- encoding technique has been adopted in the NCSs in recent
cause of the power restriction, network congestion or error, years. The MD encoding approach has been first introduced
it is unavoidable that the data transmission via the digital in the NCSs in [47], in which the Kalman filtering problem
channel suffers from the communication imperfections, e.g., has been investigated for linear stochastic systems. Due to the
transmission delay, message loss, channel noise, etc. In this packet dropout existing in the data packet transmission via
WANG et al.: ENCODING-DECODING-BASED CONTROL AND FILTERING OF NSS: INSIGHTS, DEVELOPMENTS AND OPPORTUNITIES 11

the communication network, the MD encoding scheme has limited capacity has been taken into account. Based on the
counteracted the effects of the data loss on the filtering perfor- introduced concept “topological entropy”, sufficient and neces-
mance. It has been demonstrated that compared with the single sary conditions, which are related to the bit rate of the channel
description encoding scheme, the MD one could significantly and the topological entropy of the open-loop system, have
enhance the filtering performance. In [94], the state estimation been established for the solvability of the addressed optimal
problem has been addressed in the sensor networks together control problem. In [1], the control problem for a class of
with the power control mechanism. The sensors were required nonlinear systems subject to the communication constraints
to transmit the measurements to the gateway via a lossy has been investigated. A performance index characterizing the
channel subject to random packet dropouts, and then a novel difference between the quantized and unquantized systems
MD quantizer has been applied to improve the estimation has been defined. By optimizing such an index, the optimal
robustness. For more details with respect to the encoding- dynamic quantizer in a closed form has been obtained. [147]
decoding-based filtering issues with lossy/noisy channels, we has considered the distributed estimation issue through a sen-
refer the readers to [72], [95], [103], [118] and the references sor network with the limited transmission capacity, in which
therein. an optimal-quantization-based encoding-decoding algorithm
3) Event-triggering mechanism: It is recognized that the has been put forward to minimize the estimation error. The
signal delivery through the communication channel, especially tracking problem has been addressed in [25] via a signal-
for the wireless network, would account for a major ratio in noise-ratio-constrained communication network. Two practical
the total power consumption of the NCSs. Therefore, for the network architectures have been illustrated to model the trans-
purpose of keeping the use of the energy resource as small mission channel. Moreover, the optimal encoder-decoder pair
as possible and alleviating the network collision, the event- has been found, by which the tracking error variance has been
triggering data transmission mechanism has been successfully minimized with minimum information transmissions.
applied in the NCSs of digital communications [40], [58], 2) Minimum data-rate: During the past fifteen years, the
[61], [70], [75], [78], [119], [128]. In [52], the early attempt minimum data-rate problem has gained a great deal of research
has been made on the combination of the quantization-based attention in the network-based control and signal processing
control approach and the event-triggering approach to address areas, see e.g. [35], [90], [91], [104], [109], [120]. Some
the stabilization problem for linear continuous-time systems. initial effort has been made in [90], in which the stabiliza-
Based on the quantization method utilized in [67], an encoder- tion problem has been considered for the linear discrete-
decoder pair has been proposed to reduce the amount of time systems with the limited communication capacity. By
the information to be sent via a bit-data-constrained channel, resorting to the asymptotic quantization theory, a sufficient and
meanwhile, the event-triggered scheme has been employed necessary condition has been obtained for the existence of the
to decrease the communication frequency. The boundedness minimum lower bound of the bit rate, above which the closed-
of the closed-loop system state and a lower upper of the loop system can be stabilized in the mth moment sense. The
minimum time interval between two successive events have exponential stabilization problem with limited feedback data
been explicitly studied. Very recently, for a linear scalar system rates has been studied in [91], and moreover, the minimum
with the communication delay, an event-triggered control data rate has been given which can achieve the system stability.
problem combining with the quantization-based encoding- In [35], the minimum bit rate has been obtained to stabilize
decoding scheme has been investigated in [49]. The additional the linear time-invariant systems for both the continuous-
information of the event-triggering time has been further time and the discrete-time cases. By applying the differential
excavated, and it has been shown that for the sufficiently small pulse code modulation techniques, the comparison results
time delay, the timing information carried by the triggered have been provided with respect to the derived minimum
events is large enough to stabilize the closed-loop system bit rate for the fix-step quantization and the variable-step
with any small information transmission rate. Subsequently, quantization approaches. Very recently, a joint optimization
the related work has been extended to the vector systems in problem has been discussed in [104]. For a NCS with the rate-
[50] and the binary differential coding approach in [70]. limited communication channel, a general performance index
including both the control cost and the communication cost
has been constructed, and the corresponding optimal control
D. Optimal Encoding-Decoding Schemes
and encoding problem has been solved.
In the context of bit-rate-constrained control/estimation
problems, the optimal coding scheme can be described from
IV. D ISTRIBUTED C ONTROL AND S TATE E STIMATION
two aspects: 1) deriving the optimal control/filtering perfor-
P ROBLEMS W ITH L IMITED C OMMUNICATION C APACITY
mance with a fixed bit rate of the network, and 2) determining
the minimum information needed to transmit in order to Compared with the simple NCS which only consists of
achieve a given control/filtering task. the single plant and controller/estimator, the distributed net-
1) Optimal control/estimation performance: In the first worked systems are more widespread in the real-world, such
sense, there has been attracting considerable attention in recent as the chemical process systems, the biological systems, the
years, see e.g. [1], [12], [25], [27], [46], [110], [130], [147]. In unmanned aerial vehicles, etc. For those large-scale distributed
[110], the optimal control problem for discrete-time nonlinear systems, each individual can be viewed as an agent which
uncertain systems via digital communication channels with has the capacity of sensing, communication, computation, and
12 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

implementation. In practical applications, the MASs and the 2) Subject to network-induced complexities: In the above
WSNs are two typical examples of the distributed systems, results, a common assumption is that the communication
and the corresponding distributed control and state estimation channel is noiseless, i.e., no time-delays, dropouts, exter-
problems have attracted increasing research attention in recent nal disturbances, etc. during the data transmission process.
years. Since the frequent information exchange among agents However, this is not always true in the engineering practical
would certainly occupy considerable network resource and owing to the limited network resources and the environmental
power energy, the effective data-compression and energy- complexity. In [73], the authors have discussed the average
saving transmission mechanisms are vitally meaningful in real consensus problem of the MASs subject to bit-rate constraint,
engineering systems. The typical framework of the distributed where the communication delays have been taken into consid-
networked system with encoding-decoding communication is eration. By means of constructing a pair of dynamic encoder
given by Fig. 7. In this section, we aim to provide some and decoder, a consensus control protocol has been designed
recent developments with respect to the encoding-decoding- to reach the average consensus performance with bounded
based distributed control problem for the MASs and the state time-delays. Because of the merits in the reduction of the
estimation problem for the WSNs, respectively. energy and network resource, the event-triggering transmis-
sion mechanism has recently been adopted in the distributed
control problem of the MASs with bandwidth restrictions, see
e.g. [6], [29], [54], [141]. In [54], in order to handle the
consensus control problem for a class of discrete-time MASs
via an imperfection communication network, both the event-
triggering transmission mechanism and the dynamic encoding-
decoding scheme have been employed. The proposed encoding
scheme could adaptively select the number of the channel
quantization level at each time step, and the relationship
among the convergence rate, the number of the agents, the
Fig. 7. The architecture of distributed systems with encoding-decoding com-
quantization interval, the network structure, and control gain
munication mechanism.
has been well illustrated. Based on [54], very recently, the
distributed consensus optimization problem for the MASs has
been put forward in [57]. For each agent, a convex cost func-
A. Multi-Agent Systems With Encoding-Decoding-Based In-
tion has been defined and the objective optimization function
formation Transmission
was the sum of those individual cost functions. It also has been
1) Without network-induced phenomena: Consensus is one proved that as long as the system parameters satisfy certain
of the most important collective behaviors for the MASs, conditions, the optimality of the consensus control scheme
and the related control problems have received remarkable can be guaranteed with only one bit information exchange for
research interests in the past decade, see e.g. [32], [74], [85], each pair of adjacent agents. The similar problem with some
[135], [140], [142]. It is known from [64] that the encoding- extensions has been investigated in [59].
decoding-based consensus control protocol for MASs was first
proposed in [5], in which, the average consensus performance
has been investigated for a class of linear MASs by the B. Wireless Sensor Networks With Encoding-Decoding-Based
bit-rate-constrained control protocol. According to different Communication Scheme
quantization strategies, two encoding-decoding schemes have A wireless sensor network (WSN) is usually composed of
been provided which would guarantee the desired average con- numerous simple and low-cost sensor nodes geographically
sensus if the communication network is strongly connected. deployed in certain areas. With the collaboration such as
Under a connected communication network with limited data the integrated sensing, computation and communication, those
transmission capacity, [56] has explored the average consensus sensor nodes can jointly perform various tasks in wide appli-
control problem for the first-order linear discrete-time MASs. cation fields including environmental monitoring, battlefield
By designing a dynamic encoding-decoding-based distributed surveillance, civil infrastructure monitoring and control, intel-
control law, the exponential average consensus has been en- ligent solar farms, etc., [13], [14], [16], [17], [33], [60], [84],
sured for the requirement that only one bit data exchanged [125], [127], [132]. However, it should be stressed that due
between two adjacent agents. Furthermore, an energy index to the frequent communications among sensor nodes powered
has been defined to characterize the communication energy by irreplaceable batteries, the scarce network bandwidth and
consumption for the consensus task, by optimizing which, energy resources would give rise to a severe limitation on
a trade-off between the consensus rate and the quantization the practical applications of the WSNs. Therefore, developing
levels has been presented. Then, [64] has extended to the certain useful communication techniques, which aims to im-
time-varying topologies case which is of jointly connected and prove the efficiency of the bandwidth utilization and reduce
the quantized observer-based encoding-decoding consensus the energy consumption, appears to be an urgent task.
control problem has been considered in [65] for the second- So far, an alternative way to facilitate the information
order discrete-time MASs. The related results with respect to exchange in a bit-rate-constrained WSN is to design appro-
high-order MASs can be found in [101]. priate data encoding-decoding schemes so as to reduce the
WANG et al.: ENCODING-DECODING-BASED CONTROL AND FILTERING OF NSS: INSIGHTS, DEVELOPMENTS AND OPPORTUNITIES 13

number of transmitted bits at each data transmission pro- 1) Channel encoding. So far, most of the existing results
cess, see [22], [107], [108]. In [108], the distributed filtering focus their attention on the aspect of source encoding and
problem has been considered for a class of continuous-time decoding. In fact, the channel encoding-decoding is also very
stochastic systems with a low communication capacity. For important for the reliability of the data transmission in the
each sensor, the observation value is required to be encoded communication network. Therefore, it is quite meaningful to
into one bit information to be transmitted, based on which, study the co-design issue of the source and channel encoding-
a recursive algorithm of the Kalman filtering type has been decoding in NSs.
designed. Also, it has been revealed that encoding-decoding- 2) Combining with other communication protocols. Re-
based filtering algorithm can achieve a close performance cently, some effective data scheduling strategies such as the
with the standard Kalman filtering algorithm. The distributed Round-Robin protocol, the Try-Once-Discard protocol, the
parameter estimation issue via the bandwidth-restricted WNSs stochastic protocol, etc., have been successfully applied to
has been studied in [20], in which the distributed adaptive the NSs in order to cope with the limited communication
quantization approach and a sequential transmission strategy resources. For the purpose of making full use of the network
have been employed. Based on the adaptive quantization resources, the development of several encoding-decoding-
approach, three quantization schemes have been proposed and based data scheduling protocols would be a trend for future
the corresponding maximum likelihood estimators have been research.
designed. 3) Data encryption issues. Since NSs share the common
By using the observations from multiple sensors, [53] network links for data exchange, they are prone to be attacked
has researched the distributed state estimation problem for by adversaries, and the secure communication in NSs becomes
a discrete-time linear system. Because of the data-rate con- more and more significant. Thereby, how to combine the
straint, the individual sensors’ innovations are required to be encoding-decoding schemes with the data encryption tech-
quantized before sending to the fusion center. The stability niques in a unified framework is another interesting topic.
of the estimation scheme can be achieved if the data-rate for 4) Combining with self-triggering mechanism. Although
each sensor is sufficiently large. In addition, an optimal rate there have been some results with respect to the combination
allocation issue for the individual sensors has been discussed. of the event-triggering mechanism and the decoding-decoding
[89] has studied the distributed parameter estimation problem communication schemes recently, the additional hardware is
in an energy-constrained WSN, where only one bit for each still needed to detect the specific “event”. In this sense, by
sensor has been allowed to be sent to the fusion center. On the replacing the detection task according to certain admissi-
basis of the variance of the measurement noises of individual ble “prediction” [150], the self-triggering mechanism is one
sensors, the mean square estimation error can be minimized by preferable candidate to facilitate the encoding-decoding-based
designing an optimal bit assignment scheme. In [105], based control/filtering problems.
on the Kalman filter structure, a distributed state estimation
problem has been dealt with for the microgrid, and the system-
atic convolution encoding approach has been developed to fa- R EFERENCES
cilitate the smart grid communications. By applying the energy [1] S. Azuma and T. Sugie, “Dynamic quantization of nonlinear control
measurements from the bit-rate-constrained sensor networks, systems,” IEEE Transactions on Automatic Control, vol. 57, no. 4,
the target location estimation problem has been addressed pp. 875-888, Apr. 2012.
in [71]. A position-based adaptive quantization scheme has
[2] H. Bai, M. Zhang, A. Wang, M. Liu, and Y. Zhao, “Multiple de-
been provided to improve the estimation performance, and the scription video coding using inter- and intra-description correlation at
maximum likelihood estimator and the corresponding Cramér– macro block level,” IEICE Transactions on Information and Systems,
Rao lower bound have been obtained. In [102], the distributed vol. E97D, no. 2, pp. 384-387, Feb. 2014.
state estimation issue has been handled via a lossy WSN. By [3] Y. Barshalom, “Redundancy and data compression in recursive esti-
using the encoded measurements, the Kalman filter has been mation,” IEEE Transactions on Automatic Control, vol. AC17, no. 5,
constructed to estimate the system state at the gateway, which pp. 684-689, 1972.
also has controlled the transmission power of the radio power
[4] Y. Berrouche and R. E. Bekka, “Improved multiple description wavelet
amplifier and the bit rate needed for each sensor. based image coding using Hadamard transform,” AEU-International
Journal of Electronics and Communications, vol. 68, no. 10, pp. 976-
982, 2014.
V. C ONCLUSIONS A ND F UTURE W ORK
[5] R. Carli, F. Fagnani, and P. Frasca, “Efficient quantized techniques
Throughout this paper, we have summarized the existing re- for consensus algorithms,” NeCST workshop, pp. 1-8, Nancy, France,
sults on the encoding-decoding-based control and state estima- 2007.
tion/filtering issues for NSs. Several typical encoding-decoding
[6] X. Chen, X. Liao, L. Gao, S. Yang, H. Wang, and H. Li, “Event-
schemes and the corresponding encoder-decoder structures triggered consensus for multi-agent networks with switching topology
have been systematically reviewed. Subsequently, for different under quantized communication,” Neurocomputing, vol. 230, pp. 294-
kinds of NSs, a variety of encoding-decoding-based control 301, Mar. 2017.
and state estimation problems have been thoroughly surveyed.
[7] T. M. Cheng, “Robust output feedback stabilization of nonlinear net-
Finally, based on the literature review, possible relative direc- worked systems via a finite data-rate communication channel,” Journal
tions for the further research work are listed as follows. of Systems Science & Complexity, vol. 24, no. 1, pp. 1-13, Feb. 2011.
14 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

[8] G. Como, F. Fagnani, and S. Zampieri, “Anytime reliable transmission [25] Y. W. Feng and G. Ge, “Optimal coding-decoding for systems con-
of real-valued information through digital noisy channels,” SIAM Jour- trolled via a communication channel,” International Journal of Systems
nal on Control and Optimization, vol. 48, no. 6, pp. 3903-3924, 2010. Science, vol. 44, no. 12, pp. 2190-2198, Dec. 2013.

[9] P. Correia, P. A. Assuncao, and V. Silva, “Multiple description of coded [26] A. L. Fradkov, B. Andrievsky, and M. Ananyevskiy, “State estimation
video for path diversity streaming adaptation,” IEEE Transactions on and synchronization of pendula systems over digital communication
Multimedia, vol. 14, no. 3, pp. 923-935, Jun. 2012. channels,” European Physical Journal-Special Topics, vol. 223, no. 4,
pp. 773-793, Apr. 2014.
[10] C. E. Curry, P. Mirchandani, and C. F. Price, “State estimation with
coarsely quantized, high data-rate measurements,” IEEE Transactions [27] A. L. Fradkov, B. Andrievsky, and R. J. Evans, “Chaotic observer-
on Aerospace and Electronic Systems, vol. 11, no. 4, pp. 613-621, 1975. based synchronization under information constraints, Physical Review
E,” vol. 73, no. 6, Article Number: 066209, Jun. 2006.
[11] D. F. Delchamps, “Stabilizing a linear-system with quantized state
feedback,” IEEE Transactions on Automatic Control, vol. 35, no. 8, [28] A. L. Fradkov, B. Andrievsky, and R. J. Evans, “Synchronization of
pp. 916-924, Aug. 1990. passifiable Lurie systems via limited-capacity communication channel,”
IEEE Transactions on Circuits And Systems I-Regular Papers, vol. 56,
[12] S. Dey, A. Chiuso, and L. Schenato, “Event-triggered stabilization no. 2, pp. 430-439, Feb. 2009.
of linear systems under bounded bit rates,” IEEE Transactions on
Automatic Control, vol. 62, no. 6, pp. 3054-3061, Jun. 2017. [29] L. Gao, X. Liao, H. Li, and G. Chen, “Event-triggered control for
multi-agent network with limited digital communication, Nonlinear
[13] D. Ding, Z. Wang, D. W. C. Ho, and G. Wei, “Distributed recursive Dynamics,” vol. 82, no. 4, pp. 1659-1669, Dec. 2015.
filtering for stochastic systems under uniform quantizations and decep-
tion attacks through sensor networks,” Automatica, vol. 78, pp. 231- [30] V. K. Goyal, “Multiple description coding: Compression meets the
240, Feb. 2017. network, IEEE Transactions on Signal Processing Magazine,” vol. 18,
no. 5, pp. 74-93, Sept. 2001.
[14] D. Ding, Z. Wang, B. Shen, and H. Dong, “Event-triggered distributed
H∞ state estimation with packet dropouts through sensor networks,” [31] S. Hattori, K. Kobayashi, H. Okada, and M. Katayama, “On-off
IET Control Theory & Applications, vol. 9, no. 13, pp. 1948-1955, error control coding scheme for minimizing tracking error in wireless
Aug. 2015. feedback,” IEEE Transactions on Industrial Informatics, vol. 11, no. 6,
pp. 1411-1421, Dec. 2015.
[15] D. Ding, G. Wei, S. Zhang, Y. Liu, and F. E. Alsaadi, “On scheduling
of deception attacks for discrete-time networked systems equipped with [32] F. Han, G. Wei, D. Ding, and Y. Song, “Finite-horizon H∞ -consensus
attack detectors,” Neurocomputing, vol. 219, pp. 99-106, Jan. 2017. control for multi-agent systems with random parameters: The local
condition case,” Journal of The Franklin Institute-Engineering and
[16] H. Dong, X. Bu, N. Hou, Y. Liu, F. E. Alsaadi, and T. Hayat, “Event- Applied Mathematics, vol. 354, no. 14, pp. 6078-6097, Sept. 2017.
triggered distributed state estimation for a class of time-varying systems
over sensor networks with redundant channels,” Information Fusion, [33] F. Han, G. Wei, D. Ding, and Y. Song, “Local condition based
vol. 36, pp. 243-250, Jul. 2017. consensus filtering with stochastic nonlinearities and multiple missing
measurements,” IEEE Transactions on Automatic Control, vol. 62,
[17] H. Dong, Z. Wang, J. Lam, and H. Gao, “Distributed filtering in sensor no. 9, pp. 4784-4790, Sept. 2017.
networks with randomly occurring saturations and successive packet
dropouts,” International Journal of Robust and Nonlinear Control, [34] L. He, D. Han, X. Wang, and L. Shi, “State estimation over a lossy
vol. 24, pp. 1743-1759, Aug. 2014. network using linear temporal coding, in Proceedings of the 51st IEEE
Conference on Decision and Control,” pp. 2002-2007, Maui, Hawaii,
[18] A. A. Elgamal, T. M. Cover, “Achievable rates for multiple descrip- USA, Dec. 2012.
tions,” IEEE Transactions on Information Theory, vol. 28, no. 6,
pp. 851-857, Aug. 1982. [35] J. Hespanha, A. Ortega, and L. Vasudevan, “Towards the control of
linear systems with minimum bit-rate, in Proceedings of the 15th
[19] J. Fang and Q.-Q. Liu, “Quantized feedback control over packet International Symposium on Mathematical Theory of Networks and
dropout communication channels,” Information Technology and Con- Systems,” pp. 1-15, Notre Dame, Indiana, USA, Aug. 2002.
trol, vol. 41, no. 3, pp. 229-238, 2012.
[36] N. Hou, H. Dong, Z. Wang, W. Ren, and F. E. Alsaadi, “Non-fragile
[20] J. Fang and H. Li, “Distributed adaptive quantization for wireless state estimation for discrete Markovian jumping neural networks,”
sensor networks: From delta modulation to maximum likelihood,” IEEE Neurocomputing, vol. 179, pp. 238-245, Feb. 2016.
Transactions on Signal Processing, vol. 56, no. 10, pp. 5246-5257,
Oct. 2008. [37] N. Hou, H. Dong, Z. Wang, W. Ren, and F. E. Alsaadi, “H∞ state
estimation for discrete-time neural networks with distributed delays
[21] A. Farhadi, “Feedback channel in linear noiseless dynamic systems and randomly occurring uncertainties through fading channels,” Neural
controlled over the packet erasure network,” International Journal of Networks, vol. 89, pp. 61-73, May 2017.
Control, vol. 88, no. 8, pp. 1490-1503, Aug. 2015.
[38] J. Hu, Z. Wang, and H. Gao, “Recursive filtering with random param-
[22] A. Farhadi and N. U. Ahmed, “Tracking nonlinear noisy dynamic eter matrices, multiple fading measurements and correlated noises,”
systems over noisy communication channels,” IEEE Transactions on Automatica, vol. 49, no. 11, pp. 3440-3448, Nov. 2013.
Communications, vol. 59, no. 4, pp. 955-961, Apr. 2011.
[39] L. Hu, Z. Wang, Q.-L. Han, and X. Liu, “State estimation under
[23] A. Farhadi and C. D. Charalambous, “Robust coding for a class false data injection attacks: Security analysis and system protection,”
of sources: Applications in control and reliable communication over Automatica, vol. 219, pp. 176-183, 2018.
limited capacity channels, Systems & Control Letters,” vol. 57, no. 12,
pp. 1005-1012, Dec. 2008. [40] L. Hu, Z. Wang, Q.-L. Han, and X. Liu, “Event-based input and
state estimation for linear discrete time-varying systems,” International
[24] A. Farhadi and C. D. Charalambous, “Stability and reliable data recon- Journal of Control, in press. DOI: 10.1080/00207179.2016.1269205.
struction of uncertain dynamic systems over finite capacity channels,
Automatica,” vol. 46, no. 5, pp. 889-896, May 2010. [41] J. Hu, Z. Wang, S. Liu, and H. Gao, “A variance-constrained approach
WANG et al.: ENCODING-DECODING-BASED CONTROL AND FILTERING OF NSS: INSIGHTS, DEVELOPMENTS AND OPPORTUNITIES 15

to recursive state estimation for time-varying complex networks with rected topologies and quantized communication,” IEEE Transactions
missing measurements,” Automatica, vol. 64, pp. 155-162, Feb. 2016. on Cybernetics, vol. 47, no. 8, pp. 2044-2057, Aug. 2017.

[42] H. Ishii, B. A. Francis, “Quadratic stabilization of sampled-data sys- [58] L. Li, M. Lemmon, and X. Wang, “Stabilizing bit-rates in quantized
tems with quantization,” Automatica, vol. 39, no. 10, pp. 1793-1800, event triggered control systems,” in Proceedings of The 15th ACM In-
Oct. 2013. ternational Conference on Hybrid Systems: Computation and Control,
pp. 245-254, Beijing, P. R. China, 2012.
[43] H. Ishii, S. Hara, “A subband coding approach to control under limited
data rates and message losses,” Automatica, vol. 44, no. 4, pp. 1141- [59] H. Li, S. Liu, Y. C. Soh, and L. Xie, “Event-triggered communication
1148, Apr. 2008. and data rate constraint for distributed optimization of multiagent sys-
tems,” IEEE Transactions on Systems, Man, and Cybernetics: Systems,
[44] A. Isidori, L. Marconi, C. de Persis, “Remote tracking via encoded in press. DOI: 10.1109/TSMC.2017.2694323.
information for nonlinear systems,” Systems & Control Letters, vol. 55,
no. 10, pp. 809-818, Oct. 2006. [60] Q. Li, B. Shen, Z. Wang, and F. E. Alsaadi, “A sampled-data approach
to distributed H∞ resilient state estimation for a class of nonlinear
[45] Y. Isidori, K. Takaba, and D. E. Quevedo, “Stability analysis of time-delay systems over sensor networks,” Journal of the Franklin
networked control systems subject to packet-dropouts and finite-level Institute, vol. 354, pp. 7139-7157, Oct. 2017.
quantization,” Systems & Control Letters, vol. 60, no. 5, pp. 325-332,
May 2011. [61] L. Li, X. Wang, M. D. Lemmon, “Efficiently attentive event-triggered
systems with limited bandwidth,” IEEE Transactions on Automatic
[46] X.-W. Jiang, B. Hu, Z.-H. Guan, X.-H. Zhang, and L. Yu, “Best achiev- Control, vol. 62, no. 3, pp. 1491-1497, Mar. 2017.
able tracking performance for networked control systems with encoder-
decoder,” Information Sciences, vol. 305, pp. 184-195, Jun. 2015. [62] W. Li, G. Wei, D. Ding, Y. Liu, and F. E. Alsaadi, “A new look at
boundedness of error covariance of kalman filtering,” IEEE Trans-
[47] Z. Jin, V. Gupta, and R. M. Murray, “State estimation over packet actions on Systems, Man, and Cybernetics: Systems, in press. DOI:
dropping networks using multiple description coding,” Automatica, 10.1109/TSMC.2016.2598845.
vol. 42, no. 9, pp. 1141-1152, Sept. 2006.
[63] W. Li, G. Wei, F. Han, and Y. Liu, “Weighted average consensus-
[48] D. Kapetanovic, S. Chatzinotas, and B. Ottersten, “Index assignment based unscented Kalman filtering,” IEEE Transactions on Cybernetics,
for multiple description repair in distributed storage systems,” in vol. 46, no. 2, pp. 558-567, Feb. 2016.
Proceedings of the 2014 IEEE International Conference on Communi-
cations, pp. 3896-3901, Sydney, Australia, Jun. 10-14, 2014. [64] T. Li and L. Xie, “Distributed consensus over digital networks with
limited bandwidth and time-varying topologies,” Automatica, vol. 47,
[49] M. J. Khojasteh, P. Tallapragada, J. Cortes, and M. Franceschetti, no. 9, pp. 2006-2015, Sept. 2011.
“The value of timing information in event-triggered control: The scalar
case,” in Proceedings of the 54th Annual Allerton Conference on [65] T. Li and L. Xie, “Distributed coordination of multi-agent systems with
Communication, Control, and Computing, pp. 1165-1172, Monticello, quantized-observer based encoding-decoding,” IEEE Transactions on
USA, Sept. 27-30, 2016. Automatic Control, vol. 57, no. 12, pp. 3023-3037, Dec. 2012.

[50] M. J. Khojasteh, P. Tallapragada, J. Cortes, and M. Franceschetti, [66] X. Li, C. Li, and M. Z. Q. Chen, “Stabilisation of non-linear DISS
“Time-triggering versus event-triggering control over communication systems with uncertainty via encoded feedback,” IET Control Theory
channels,” in arXiv:1703.10744 [math.OC]. & Applications, vol. 11, no. 5, pp. 732-739, Mar. 2017.

[51] E. Kofman and J. H. Braslavsky, “Level crossing sampling in feedback [67] D. Liberzon, “On stabilization of linear systems with limited informa-
stabilization under data-rate constraints,” in Proceedings of the 45th tion,” IEEE Transactions on Automatic Control, vol. 48, no. 2, pp. 304-
IEEE Conference on Decision and Control, pp. 4423-4428, San Diego, 307, Feb. 2003.
California, USA, Dec. 2006.
[68] D. Liberzon, “Hybrid feedback stabilization of systems with quantized
[52] D. Lehmann and J. Lunze, “Event-based control using quantized state signals,” Automatica, vol. 39, no. 3, pp. 1543-1554, Sept. 2003.
information,” in Proceedings of the 2nd IFAC Workshop on Distributed
Estimation and Control in Networked Systems, pp. 1989-1994, Annecy, [69] D. Liberzon, J. P. Hespanha, “Stabilization of nonlinear systems
France, 2010. with limited information feedback,” IEEE Transactions on Automatic
Control, vol. 50, no. 6, pp. 910-915, Jun. 2005.
[53] A. S. Leong, D. Subhrakanti, and G. N. Nair, “Quantized filtering
schemes for multi-sensor linear state estimation: stability and perfor- [70] Q. Ling, “Bit rate conditions to stabilize a continuous-time scalar linear
mance under high rate quantization,” IEEE Transactions on Signal system based on event triggering,” IEEE Transactions on Automatic
Processing, vol. 61, no. 15, pp. 3852-3865, Aug. 2013. Control, vol. 62, no. 8, pp. 4093-4100, Aug. 2017.

[54] H. Li, G. Chen, T. Huang, Z. Dong, and L. Gao, “Event-triggered [71] G. Liu, H. Liu, H. Chen, C. Zhou, and L. Shu, “Position-based adaptive
distributed average consensus over directed digital networks with quantization for target location estimation in wireless sensor networks
limited communication bandwidth,” IEEE Transactions on Cybernetics, using one-bit data,” Wireless Communications & Mobile Computing,
vol. 46, no. 12, pp. 3098-3110, Dec. 2016. vol. 16, no. 8, pp. 929-941, Jun. 2016.

[55] J. Li, H. Dong, F. Han, N. Hou, and X. Li, “Filter design, fault [72] Q. Liu and J. Fan, “State estimation for networked control systems
estimation and reliable control for networked time-varying systems: using fixed data rates,” International Journal of Systems Science,
A survey,” Systems Science & Control Engineering: An Open Access vol. 48, no. 9, pp. 1818-1828, 2017.
Journal, vol. 5, pp. 331-341, 2017.
[73] S. Liu, T. Li, and L. Xie, “Distributed consensus for multiagent systems
[56] T. Li, M. Fu, L. Xie, and J. Zhang, “Distributed consensus with limited with communication delays and limited data rate,” SIAM Journal on
communication data rate,” IEEE Transactions on Automatic Control, Control and Optimization, vol. 49, no. 6, pp. 2239-2262, 2011.
vol. 56, no. 2, pp. 279-292, Feb. 2011.
[74] S. Liu, T. Li, L. Xie, M. Fu, and J. Zhang, “Continuous-time and
[57] H. Li, C. Huang, G. Chen, X. Liao, and T. Huang, “Distributed sampled-data-based average consensus with logarithmic quantizers,”
consensus optimization in multiagent networks with time-varying di- Automatica, vol. 49, no. 11, pp. 3329-3336, Nov. 2013.
16 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

[75] S. Liu, Y. Song, D. Ding, and Y. Liu, “Event-triggered dynamic output back: tightest attainable bounds,” Systems & Control Letters, vol. 41,
feedback RMPC for polytopic systems with redundant channels: Input- no. 1, pp. 49-56, Sept. 2000.
to-state stability,” Journal of the Franklin Institute-Engineering and
Applied Mathematics, vol.354, no. 7, pp. 2871-2892, May 2017. [91] G. N. Nair and R. J. Evans, “Exponential stabilisability of finite-
dimensional linear systems with limited data rates,” Automatica,
[76] S. Liu, Y. Song, G. Wei, and X. Huang, “RMPC-based security vol. 39, no. 4, pp. 585-593, Apr. 2003.
problem for polytopic uncertain system subject to deception attacks and
persistent disturbances,” IET Control Theory & Applications, vol. 11, [92] G. N. Nair, “A nonstochastic information theory for communication
no. 10, pp. 1611-1618, Jun. 2017. and state estimation,” IEEE Transactions on Automatic Control, vol. 58,
no. 6, pp. 1479-1510, Jun. 2013.
[77] W. Liu, Z. Wang, and N. Mingkang, “Controlled synchronization for
chaotic systems via limited information with data packet dropout,” [93] D. Nešić and D. S. Laila, “A note on input-to-state stabilization for
Automatica, vol. 49, no. 11, pp. 3329-3336, Nov. 2013. nonlinear sampled-data systems,” IEEE Transactions on Automatic
Control, vol. 47, no. 7, pp. 1153-1158, Jul. 2002.
[78] H. Liu, Z. Wang, B. Shen, and X. Liu, “Event-triggered H∞
state estimation for delayed stochastic memristive neural networks [94] J. Ostergaard, D. E. Quevedo, and A. Ahlen, “Predictive power control
with missing measurements: The discrete time case,” IEEE Trans- and multiple-description coding for wireless sensor networks,” in Pro-
actions on Neural Networks and Learning Systems, in press. DOI: ceedings of the IEEE International Conference on Acoustics, Speech,
10.1109/TNNLS.2017.2728639. and Signal Processing, Taipei, Taiwan, China, Apr. 19-24, 2009.

[79] H. Liu, Z. Wang, B. Shen, and F. E. Alsaadi, “H∞ state estimation [95] J. Ostergaard and D. E. Quevedo, “Multiple descriptions for packetized
for discrete-time memristive recurrent neural networks with stochastic predictive control,” Eurasip Journal on Advances in Signal Processing,
time-delays,” International Journal of General Systems, vol. 45, no. 5, Article Number: 45, Apr. 2016.
pp. 633-647, 2016.
[96] C. de Persis and A. Isidori, “Stabilizability by state feedback implies
[80] S. Liu, G. Wei, Y. Song, and Y. Liu, “Extended Kalman filtering for stabilizability by encoded state feedback,” Systems&Control letters,
stochastic nonlinear systems with randomly occurring cyber attacks,” vol. 53, no. 3-4, pp. 249-258, Nov. 2004.
Neurocomputing, vol. 207, pp. 708-716, Sept. 2016.
[97] C. de Persis, “n-Bit stabilization of n-dimensional nonlinear systems in
[81] M. B. Loiola, R. R. Lopes, and J. M. T. Romano, “Modified Kalman feedforward form,” IEEE Transactions on Automatic Control, vol. 50,
filters for channel estimation in orthogonal space-time coded systems,” no. 3, pp. 299-311, Mar. 2005.
IEEE Transactions on Automatic Control, vol. 60, no. 1, pp. 533-538,
Jan. 2012. [98] C. de Persis, “Nonlinear stabilizability via encoded feedback: The case
of integral ISS systems,” Automatica, vol. 42, no. 10, pp. 1813-1816,
[82] Y. Luo, Z. Wang, G. Wei, and F. E. Alsaadi, “Robust H∞ filtering Oct. 2006.
for a class of two-dimensional uncertain fuzzy systems with randomly
occurring mixed delays,” IEEE Transactions on Fuzzy Systems, vol. 25, [99] C. de Persis and D. Nešić, “Practical encoders for controlling nonlinear
no. 1, pp. 70-83, Feb. 2017. systems under communication constraints,” Systems & Control Letters,
vol. 57, no. 8, pp. 654-662, Aug. 2008.
[83] Y. Luo, Z. Wang, G. Wei, F. E. Alsaadi, and T. Hayat, “State estimation
for a class of artificial neural networks with stochastically corrupted [100] C. de Persis, “Minimal data rate stabilization of nonlinear systems
measurements under Round-Robin protocol,” Neural Networks, vol. 77, over networks with large delays,” International Journal of Robust And
pp. 70-79, May 2016. Nonlinear Control, vol. 20, no. 10, pp. 1097-1111, Jul. 2010.

[84] L. Ma, Z. Wang, Q.-L. Han, and H. K. Lam, “Variance-constrained [101] Z. Qiu, L. Xie, and Y. Hong, “Quantized leaderless and leader-
distributed filtering for time-varying systems with multiplicative noises following consensus of high-order multi-agent systems with limited
and deception attacks over sensor networks,” IEEE Sensors Journal, data rate,” IEEE Transactions on Automatic Control, vol. 61, no. 9,
vol. 17, no. 7, pp. 2279-2288, Apr. 2017. pp. 2432-2447, Sept. 2016.

[85] L. Ma, Z. Wang, and H. K. Lam, “Event-triggered mean-square [102] D. E. Quevedo, A. Ahlen, and J. Ostergaard, “Energy efficient state
consensus control for time-varying stochastic multi-agent system with estimation with wireless sensors through the use of predictive power
sensor saturations,” IEEE Transactions on Automatic Control, vol. 62, control and coding,” IEEE Transactions on Signal Processing, vol. 58,
no. 7, pp. 3524-3531, Jun. 2017. no. 9, pp. 4811-4823, Sept. 2010.

[86] V. Malyavej and A. V. Savkin, “Set-valued robust Kalman state esti- [103] D. E. Quevedo, J. Ostergaard, and A. Ahlen, “Power control and
mation via digital communication channels with bit-rate constraints,” coding formulation for state estimation with wireless sensors,” IEEE
in Proceedings of the American Control Conference Denver, Colorado, Transactions on Control Systems Technology, vol. 22, no. 2, pp. 413-
USA, Jun. 4-6, 2003. 427, Feb. 2014.

[87] A. S. Matveev, “State estimation via limited capacity noisy communi- [104] M. Rabi, C. Ramesh, and K. H. Johansson, “Separated design of
cation channels,” Mathematics of Control Signals and Systems, vol. 20, encoder and controller for networked linear quadratic optimal control,”
no. 1, pp. 1-35, Mar. 2008. SIAM Journal on Control and Optimization, vol. 54, no. 2, pp. 662-689,
2016.
[88] A. S. Matveev and A. V. Savkin, “Shannon zero error capacity in the
problems of state estimation and stabilization via noisy communication [105] M. M. Rana, L. Li, and S. Su, “Distributed state estimation using RSC
channels,” International Journal of Control, vol. 80, no. 2, pp. 241-255, coded smart grid communications,” IEEE Access, vol. 3, pp. 1340-
Feb. 2007. 1349, 2015.

[89] S. Movaghati and M. Ardakani, “Optimum bit-sensor assignment [106] M. A. Rana, F. Ndiaye, S. A. Chowdhury, and N. Mansoor, “Multi-
for distributed estimation in inhomogeneous sensor networks,” IEEE ple description image transmission for diversity systems over unreli-
Communications Letters, vol. 18, no. 4, pp. 668-671, Apr. 2014. able communication networks,” in Proceedings of 10th International
Conference on Computer and Information Technology, pp. 240-244,
[90] G. N. Nair and R. J. Evans, “Stabilization with data-rate-limited feed- Dhanmondi, Bangladesh, Dec. 27-29, 2007.
WANG et al.: ENCODING-DECODING-BASED CONTROL AND FILTERING OF NSS: INSIGHTS, DEVELOPMENTS AND OPPORTUNITIES 17

[107] A. Ribeiro and B. Giannakis, “Bandwidth-constrained distributed es- approach to recursive filtering for nonlinear two-dimensional systems
timation for wireless sensor networks-Part I: Gaussian case,” IEEE with measurement degradations,” IEEE Transactions on Cybernetics,
Transactions on Signal Processing, vol. 54, no. 3, pp. 1131-1143, in press. DOI: 10.1109/TCYB.2017.2716400.
Mar. 2006.
[125] L. Wang, Z. Wang, Q.-L. Han, and G. Wei, “Event-based variance-
[108] A. Ribeiro, B. Giannakis, and S. I. “Roumeliotis, SOI-KF: Distributed constrained H∞ filtering for stochastic parameter systems over sensor
Kalman filtering with low-cost communications using the sign of networks with successive missing measurements,” IEEE Transactions
innovations,” IEEE Transactions on Signal Processing, vol. 54, no. 12, on Cybernetics, in press. DOI: 10.1109/TCYB.2017.2671032.
pp. 4782-4795, Dec. 2006.
[126] L. Wang, Z. Wang, Q.-L. Han, and G. Wei, “Synchronization control
[109] A. J. Rojas, J. S. Freudenberg, J. H. Braslavsky, and R. H. Middleton, for a class of discrete-time dynamical networks with packet dropouts: a
“Optimal signal to noise ratio in feedback over communication chan- coding-decoding-based approach,” IEEE Transactions on Cybernetics,
nels with memory,” in Proceedings of the 45th IEEE Conference on in press. DOI: 10.1109/TCYB.2017.2740309.
Decision & Control, pp. 240-244, San Diego, CA, USA, Dec. 13-15,
2006 [127] L. Wang, Z. Wang, T. Huang, and G. Wei, “An event-triggered approach
to state estimation for a class of complex networks with mixed time
[110] A. V. Savkin, “Analysis and synthesis of networked control systems: delays and nonlinearities,” IEEE Transactions on Cybernetics, vol. 46,
Topological entropy, observability, robustness and optimal control,” no. 11, pp. 2497-2508, Nov. 2016.
Automatica, vol. 42, no. 1, pp. 51-62, Jan. 2006.
[128] L. Wang, Z. Wang, G. Wei, and F. E. Alsaadi, “Finite-time state
[111] A. V. Savkin and T. M. Cheng, “Detectability and output feedback estimation for recurrent delayed neural networks with component-based
stabilizability of nonlinear networked control systems,” IEEE Transac- event-triggering protocol,” IEEE Transactions on Neural Networks and
tions on Automatic Control, vol. 52, no. 4, pp. 730-735, Apr. 2007. Learning Systems, in press. DOI: 10.1109/TNNLS.2016.2635080.

[112] C. E. Shannon, “A mathematical theory of communication,” Bell System [129] J. Wang and Z. Yan, “Coding scheme based on spherical polar coordi-
Technical Journal, vol. 27, no. 3, pp. 379-423, Jul. 1948. nate for control over packet erasure channel, International Journal of
Robust and Nonlinear Control, vol. 51, pp. 1159-1176, May 2014.
[113] C. E. Shannon, “A mathematical theory of communication,” Bell System
Technical Journal, vol. 27, no. 4, pp. 623-666, Oct. 1948. [130] X. Wang, E. E. Yaz, and J. Long, “Robust and resilient state-dependent
control of discrete-time nonlinear systems with general performance
[114] A. Shirazinia, A. A. Zaidi, L. Bao, and M. Skoglund, “Dynamic source- criteria,” Systems Science & Control Engineering: An Open Access
channel coding for estimation and control over binary symmetric Journal, vol. 2, pp. 48-54, 2014.
channels,” IET Control Theory & Applications, vol. 9, no. 9, pp. 1444-
1454, Jun. 2015. [131] W. Wei, S. Liu, Y. Song, and Y. Liu, “Probability-guaranteed set-
membership filtering for systems with incomplete measurements,”
[115] B. Shen, Z. Wang, D. Ding, and H. Shu, “H∞ state estimation Automatica, vol. 60, pp. 12-16, Oct. 2015.
for complex networks with uncertain inner coupling and incomplete
measurements,” IEEE Transactions on Neural Networks and Learning [132] G. Wei, S. Liu, L. Wang, and Y. Wang, “Event-based distributed set-
Systems, vol. 24, no. 12, pp. 2027-2037, Dec. 2013. membership filtering for a class of time-varying non-linear systems
over sensor networks with saturation effects,” International Journal of
[116] B. Shen, Z. Wang, and H. Qiao, “Event-triggered state estimation for General Systems, vol. 45, no. 5, pp. 532-547, 2016.
discrete-time multidelayed neural networks with stochastic parameters
and incomplete measurements,” IEEE Transactions on Neural Networks [133] N. Wiener, “Cybernetics: Or Control and Communication in the Animal
and Learning Systems, vol. 28, no. 5, pp. 1152-1163, May 2017. and the Machine.” Paris, (Hermann & Cie) & Camb. Mass. (MIT
Press), 1948.
[117] T. Sui, K. You, and M. Fu, “Kalman filtering with intermittent
observations using measurements coding,” in 10th IEEE International [134] W. S. Wong and R. W. Brockett, “Systems with finite communica-
Conference on Control and Automation, pp. 1127-1132, 2013. tion bandwidth constraints-II: Stabilization with limited information
feedback,” IEEE Transactions on Automatic Control, vol. 44, no. 5,
[118] T. Sui, K. You, M. Fu, and D. Marelli, “Stability of MMSE state esti- pp. 1049-1053, May 1999.
mators over lossy networks using linear coding,” Automatica, vol. 51,
pp. 167-174, Jan. 2015. [135] J. Wu, H. Li, and X. Chen, “Leader-following consensus of nonlinear
discrete-time multi-agent systems with limited communication channel
[119] P. Tallapragada and J. Cortes, “Event-triggered stabilization of linear capacity,” Journal of The Franklin Institute-Engineering and Applied
systems under bounded bit rates,” IEEE Transactions on Automatic Mathematics, vol. 354, no. 10, pp. 4179-4195, Jul. 2017.
Control, vol. 61, no. 6, pp. 1575-1589, Jun. 2016.
[136] X. Xiao, L. Zhou, and G. Lu, “Detection of singular systems via a
[120] S. Tatikonda and S. Mitter, “Control-under communication constraints,” limited communication channel with missing measurements capacity,”
IEEE Transactions on Automatic Control, vol. 49, no. 7, pp. 1056-1068, Information Sciences, vol. 228, pp. 192-202, Apr. 2013.
Jul. 2004.
[137] Y. Xu, C. Zhu, and L. Yu, “Multipath routing of multiple description
[121] V. A. Vaishampayan, “Design of multiple description scalar quantizers,” coded images in wireless networks,” Journal of Computer Science and
IEEE Transactions on Information Theory, vol. 39, no. 3, pp. 8214-834, Technology, vol. 29, no. 4, pp. 576-588, Jul. 2014.
May 1993.
[138] K. You, “Control over communication networks: A personal perspec-
[122] V. A. Vaishampayan and J. C. Batllo, “Asymptotic analysis of multiple tive,” in Proceedings of the 11th World Congress on Intelligent Control
description quantizers,” IEEE Transactions on Information Theory, and Automation (WCICA’2014), pp. 2002-2007, Shenyang, P. R. China,
vol. 44, no. 1, pp. 278-284, Jan. 1998. Jun.-Jul. 2012.

[123] J. Wang and H. Li, “Stabilization of a continuous linear system over [139] K. You, “Recursive algorithms for parameter estimation with adaptive
channel with network-induced delay and communication constraints,” quantizer,” Automatica, vol. 52, pp. 192-201, Feb. 2015.
European Journal of Control, vol. 31, pp. 72-78, Sept. 2016.
[140] K. You and L. Xie, “Network topology and communication data
[124] F. Wang, J. Liang, Z. Wang, and X. Liu, “A variance-constrained rate for consensusability of discrete-time multi-agent systems,” IEEE
18 IEEE/CAA JOURNAL OF AUTOMATICA SINICA, VOL. 5, NO. 1, JANUARY 2018

Transactions on Automatic Control, vol. 56, no. 10, pp. 2262-2275, Licheng Wang received the B.Sc. degree in au-
Oct. 2011. tomation in 2011 from Weifang University, Weifang,
China, and the M.Sc. degree in control science and
[141] M. Yu, C. Yan, and D. Xie, “Event-triggered control for couple-group engineering in 2014 from University of Shanghai
multi-agent systems with logarithmic quantizers and communication for Science and Technology, Shanghai, China. He
delays,” Asian Journal of Control, vol. 19, no. 2, pp. 681-691, is now currently pursuing the Ph.D. degree in Con-
Mar. 2017. trol Science and Engineering at the University of
Shanghai for Science and Technology, Shanghai,
[142] W. Yu, G. Chen, and M. Cao, “Some necessary and sufficient condi- China. Since Nov. 2016, he has been a visiting
tions for second-order consensus in multi-agent dynamical systems,” Ph.D. student in the Department of Electronic and
Automatica, vol. 46, no. 6, pp. 1089-1095, Jun. 2010. Computer Engineering at Brunel University London
in the UK. His research interests include nonlinear stochastic control and
filtering, as well as complex networks and sensor networks. Mr. Wang is a
[143] Y. Yuan, P. Zhang, L. Guo, and H. Yang, “Towards quantifying the
reviewer for some international journals.
impact of randomly occurred attacks on a class of networked control
systems,” Journal of the Franklin Institute-Engineering and Applied
Mathematics, vol.354, no. 12, pp. 4966-4988, Aug. 2017.

[144] S. Zhang, D. Ding, G. Wei, J. Mao, Y. Liu, and F. E. Alsaadi,

“Design and analysis of H∞ filter for a class of T-S fuzzy system
with redundant channels and multiplicative noises,” Neurocomputing,
vol. 260, pp. 257-264, Oct. 2017.

[145] S. Zhang, D. Ding, G. Wei, Y. Liu, and F. E. Alsaadi, “H∞ state

estimation for artificial neural networks over redundant channels,”
Neurocomputing, vol. 226, pp. 117-125, Feb. 2017.

[146] W. Zhang, S. Vedantam, and U. Mitra, “Joint transmission and state es-
timation: A constrained channel coding approach,” IEEE Transactions
on Information Theory, vol. 57, no. 10, pp. 7084-7095, Oct 2011. Shuai Liu received her B.Sc. degree in School of
Mathematics and Statistics Science in 2011 from
Ludong University, Yantai, China and the M.Sc.
[147] G. Zherlitsyn and A. S. Matveev, “Min-max optimal data encoding and Degree in College of Science in 2014 from Uni-
fusion in sensor networks,” Automatica, vol. 46, no. 9, pp. 1546-1552, versity of Shanghai for Science and Technology,
Sept. 2010. Shanghai, China. She is now currently working
toward the Ph.D. degree in Business School from
[148] L. Zhou and G. Lu, “Detection and stabilization for discrete-time University of Shanghai for Science and Technology,
descriptor systems via a limited capacity communication channel,” Shanghai, China. Since 2017, she has been a Visiting
Automatica, vol. 45, no. 10, pp. 2272-2277, Oct. 2009. Ph.D. Student with the Department of Information
Systems and Computing, Brunel University Lon-
[149] L. Zou, Z. Wang, and H. Gao, “Set-membership filtering for time- don, Uxbridge, U.K. Her research interests include sensor networks, set-
varying systems with mixed time-delays under Round-Robin and membership filtering, model predictive control as well as nonlinear filtering.
Weighted Try-Once-Discard protocols,” Automatica, vol. 74, pp. 341-
348, Dec. 2016.

[150] L. Zou, Z. Wang, and D. H. Zhou, “Event-based control and filtering

of networked systems: A survey,” International Journal of Automation
and Computing, vol. 14, no. 3, pp. 239-253, Jun. 2017.

Zidong Wang (SM’03-F’14) was born in Jiangsu,

China, in 1966. He received the B.Sc. degree
in mathematics from Suzhou University, Suzhou,
China, in 1986, and the M.Sc. degree in applied
mathematics and the Ph.D. degree in electrical en- Guoliang Wei received the B.Sc. degree in math-
gineering from the Nanjing University of Science ematics from Henan Normal University, Xinxiang,
and Technology, Nanjing, China, in 1990 and 1994, China, in 1997 and the M.Sc. degree in applied
respectively. mathematics and the Ph.D. degree in control engi-
He is currently a Professor of dynamical systems neering, both from Donghua University, Shanghai,
and computing with the Department of Information China, in 2005 and 2008, respectively. He is cur-
Systems and Computing, Brunel University London, rently a Professor with the Department of Control
Uxbridge, U.K. From 1990 to 2002, he held teaching and research appoint- Science and Engineering, University of Shanghai for
ments in universities in China, Germany, and the U.K. He has published over Science and Technology, Shanghai, China.
300 papers in refereed international journals. His current research interests From March 2010 to May 2011, he was an
include dynamical systems, signal processing, bioinformatics, control theory, Alexander von Humboldt Research Fellow in the In-
and applications. stitute for Automatic Control and Complex Systems, University of Duisburg-
Prof. Wang was a recipient of the Alexander von Humboldt Research Essen, Germany. From March 2009 to February 2010, he was a post doctoral
Fellowship of Germany, the JSPS Research Fellowship of Japan, and the research fellow in the Department of Information Systems and Computing,
William Mong Visiting Research Fellowship of Hong Kong. He serves (or Brunel University, Uxbridge, UK, sponsored by the Leverhulme Trust of the
has served) as the Editor-in-Chief of Neurocomputing, and an Associate Editor UK From June to August 2007, he was a Research Assistant at the University
for 12 international journals, including the IEEE Transactions on Automatic of HongKong. From March to May 2008, he was a Research Assistant at the
Control, IEEE Transactions on Control Systems Technology, IEEE Transac- City University of Hong Kong.
tions on Neural Networks, IEEE Transactions on Signal Processing, and IEEE His research interests include nonlinear systems, stochastic systems, and
Transactions on Systems, Man, and Cybernetics–Part C: Applications and bioinformatics. He has published more than 20 papers in refereed international
Reviews. He is a Fellow of the Royal Statistical Society and a member of journals. He is a very active reviewer for many international journals.
program committee for several international conferences.