International Journal of Automation and Computing 18(6), December 2021, 857-886

www.ijac.net DOI: 10.1007/s11633-021-1306-z

Dynamic Event-triggered Control and

Estimation: A Survey
Xiaohua Ge Qing-Long Han Xian-Ming Zhang Derui Ding
School of Software and Electrical Engineering, Swinburne University of Technology, Melbourne 3122, Australia

Abstract: The efficient utilization of computation and communication resources became a critical design issue in a wide range of net-
worked systems due to the finite computation and processing capabilities of system components (e.g., sensor, controller) and shared net-
work bandwidth. Event-triggered mechanisms (ETMs) are regarded as a major paradigm shift in resource-constrained applications com-
pared to the classical time-triggered mechanisms, which allows a trade-off to be achieved between desired control/estimation perform-
ance and improved resource efficiency. In recent years, dynamic event-triggered mechanisms (DETMs) are emerging as a promising ena-
bler to fulfill more resource-efficient and flexible design requirements. This paper provides a comprehensive review of the latest develop-
ments in dynamic event-triggered control and estimation for networked systems. Firstly, a unified event-triggered control and estima-
tion framework is established, which empowers several fundamental issues associated with the construction and implementation of the
desired ETM and controller/estimator to be systematically investigated. Secondly, the motivations of DETMs and their main features
and benefits are outlined. Then, two typical classes of DETMs based on auxiliary dynamic variables (ADVs) and dynamic threshold
parameters (DTPs) are elaborated. In addition, the main techniques of constructing ADVs and DTPs are classified, and their corres-
ponding analysis and design methods are discussed. Furthermore, three application examples are provided to evaluate different ETMs
and verify how and under what conditions DETMs are superior to their static and periodic counterparts. Finally, several challenging is-
sues are envisioned to direct the future research.

Keywords: Networked systems, dynamic event-triggered control, dynamic event-triggered estimation, dynamic event-triggered
mechanisms, vehicle active suspension system, water distribution and supply system.

Citation: X. Ge, Q. L. Han, X. M. Zhang, D. Ding. Dynamic event-triggered control and estimation: a survey. International Journal of
Automation and Computing, vol.18, no.6, pp.857–886, 2021. http://doi.org/10.1007/s11633-021-1306-z

1 Introduction lers and state estimators are deployed spatially and re-
motely over some wireless and digital network mediums,
Advances in communication and computer techniques where control and estimation tasks persistently share pre-
are leading to a new paradigm of control and estimation cious resources with other neighboring tasks[7]. Under this
for modern networked systems. In the past, the plant was new paradigm, the constrained resource issue, posed by fi-
controlled and monitored via networks configured with nite bandwidth and scarce computation and processing
sufficient communication resources. Accordingly, desired capabilities of battery-powered system components,
control and state estimation actions are implemented in a should be adequately taken into account during control-
time-triggered fashion (namely, at predetermined and ler and estimator design. It is thus of theoretical and
periodic instants of time) since this allows system per- practical significance to address resource-efficient control
formance analysis and design procedures to be readily and estimation problems for networked systems.
In order to efficiently utilize limited computation and
performed by using the celebrated sampled-data system
communication resources, data sampling and transmis-
theory[1−4]. However, it is well acknowledged that these
sion actions or control and estimate updates should be
time-triggered control and estimation approaches often
kept to the minimum required to preserve desired con-
lead to over-utilization of available computation and com-
trol and estimation performance, which motivates an
munication resources, given that embedded system com-
event-triggered mechanism (ETM)1. It is noteworthy that
ponents are constantly battery-operated[5] and data com-
event-triggered ideas have been prevalently adopted to
munication is often energy-costly[6]. Nowadays, control- account for a wide array of capabilities, such as data
Review sampling, transmission, communication, scheduling, and
Manuscript received April 15, 2021; accepted May 12, 2021;
published online June 11, 2021
controller/estimator update. Understanding that event-
Recommended by Associate Editor Yuan-Qing Xia 1We use the terms “event-triggered mechanism” and “event
Colored figures are available in the online version at https://link.
springer.com/journal/11633 trigger” interchangeably throughout the paper whenever without
© The Author(s) 2021 causing confusion.
 858 International Journal of Automation and Computing 18(6), December 2021

based controller and estimator take the triggered and re- between the last triggering instant tm and the present
leased data packets as their inputs, we employ, hereafter, time t̃ with the error being denoted as e(t̃) =z(tm ) − z(t̃);
the term ETM to refer to either a sampling event or a and a threshold function σf (Z(t̃)) that is used to quantit-
transmission event of the data of interest (e.g., system atively characterize the change/variation of the data
state or measurement output), which depends on the trig- amplitude, where g(·) and f (·) are class K functions, and
gering strategies in terms of event-triggered sampling the other notations are given in Table 1. Then, the tradi-
(ETS) and event-triggered transmission (ETT). Specific- tional static event-triggered mechanism (SETM) decides
ally, an ETM decides when or how often data samplings when to sample and/or transmit the data z(t̃) at every
and/or transmissions should be performed based on some instant of time t̃ according to the following decision rule
well-defined events rather than at fixed points of time.
From this perspective, ETMs can be regarded as an addi- tm+1 = inf{t̃ > tm | g(e(t̃)) > σf (Z(t̃))}. (1)
tion of certain intelligence to conventional sampling
and/or transmission decisions in a time-triggered mechan- By properly designing the triggering condition
ism (TTM). Hence, ETMs are capable of significantly re- g(e(t̃)) > σf (Z(t̃)), it is obvious that the triggering ac-
ducing the number of data samplings and/or transmis- tions (or the instants tm+1) can be only invoked when the
sions compared with TTMs, while retaining satisfactory
data z(t̃) is truly needed to be sampled and/or transmit-
system performance. On the other hand, it is noted that
ted for ensuring stability and performance requirements,
reducing unnecessary data transmissions contributes to a
which thus leads to a noticeable reduction of occupancy
relief of network traffic or congestion, which in turn alle-
of the available computation and communication re-
viates network-induced phenomena (e.g., transmission
delays, packet dropouts) that inevitably affect desired sources. The following two facts are noted from the above
control/estimation performance. In this sense, ETMs may SETM: 1) The threshold function σf (Z(t̃)) directly ac-
also be beneficial for meeting a fundamental quality-of- counts for the triggering frequency of events. Specifically,
service requirement during networked controller/estimat- the larger the threshold, the less the number and thus the
or design. frequency of the events. The existing literature has con-
An ETM is composed of two essential components: an sidered a wide variety of threshold functions to cater to
error function g(e(t̃)) that evaluates the data of interest different system models and problem formulations, e.g.,

Table 1 Mathematical notations used in ETMs

Notation Meaning

h>0 The constant sampling period

{kh| k ∈ N} The monotonically increasing time sequence of sampling instants on sensor/sampler

The monotonically increasing time sequence of event triggering/releasing instants, where tm ∈ R in ETS,
{tm | m ∈ N; t0 = 0} tm ∈ {kh} in ETT, and tm ∈ N in the discrete-time case

{sm | sm = tm + mh, m = 0, The monotonically increasing time sequence of sampling instants between any two consecutive triggering
1, · · ·, tm+1 − tm − 1} instants in ETT

Tm = tm+1 − tm The m -th IET between two consecutive events

z(t) The data of interest, i.e., the system state x(t) or the system measurement output y(t)

z̃(t) ≜ z(tm ) The last triggered/released data, i.e., x̃(t) ≜ x(tm ) or ỹ(t) ≜ y(tm ) at time tm

The present instant of time, i.e., t̃ ≜ t ∈ R in the ETS case or t̃ ≜ kh ∈ R in the ETT case or t̃ ≜ t ∈ N in
t̃
the discrete-time case

The data at the present instant of time t or kh , i.e., z(t̃) ≜ x(t) or z(t̃) ≜ y(t) in ETS or discrete-time case,
z(t̃)
and z(t̃) ≜ x(kh) or z(t̃) ≜ y(kh) in ETT

Z(t̃) The last triggered data z̃(t) or the present data z(t̃)

e(t̃) = z̃(t) − z(t̃) The triggering error between the last triggered data and the present data

Φ>0 The weighting matrix in the relevant triggering condition to be designed

σ≥0 The static threshold parameter in the relevant triggering condition

σ(t̃) ≥ 0 The DTP in the relevant triggering condition

λ(t̃) ≥ 0 The ADV (or internal dynamic variable) in the relevant triggering condition
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 859

Section 2.4 for some typical threshold functions. 2) There lease the data at every sampling instant of time. In con-
is a fundamental trade-off between desired control/estim- trast, an SETM represents a single decision-maker via
ation performance and expected resource efficiency. In merely the predefined threshold function. The promise of
other words, along with decreased resource occupancy, it further decreasing the frequency of samplings on sensor
is not uncommon that the overall control/estimation per- devices and/or data packet transmissions over some
formance is degraded to some extent because less data shared communication medium serves as the primary mo-
from the plant is transmitted for controller/estimator tivation that stimulates recent developments of this class
design and implementation. of ADV-based DETMs in networked systems.
Based on the same parameter selection as in SETM Although the advantages of the ADV-based DETMs
(1), a question naturally arises: How can one further re- are well motivated, these DETMs are implemented in a
duce the number of events without sacrificing too much decisive manner to reduce the frequency of sampling
the desired control/estimation performance? A solution to and/or transmission actions as much as possible.
this question is to enlarge the threshold function σf (Z(t̃)) However, in many practical situations, one needs ways of
by adding a non-negative (or strictly positive) auxiliary implementing dynamic triggering ideas in a more flexible
dynamic variable (ADV) to the right-hand side of the and versatile manner. For example, network conditions
triggering condition such that only the more significantly intrinsically vary over time, and network bandwidth may
changed data can cross the newly defined threshold, be only busy during some specific peak periods but idle
which motivates a dynamic event-triggered mechanism during other periods. In this sense, an intelligent DETM
(DETM)[8−10]. For a simple illustration, a comparative ex- should trigger events more often when actual network
ample of triggering the signal x(t) = sin(πt)e−0.3t , t ∈ R bandwidth is idle but less frequently when bandwidth is
under TTM, SETM and DETM, respectively, is provided busy. On the other hand, from a convergence perspective,
in Fig. 1. It can be seen that introducing a positive ADV more data packets are expected to be released either at
λ(t) results in significantly sporadic events compared with an early stage of system evolution or when a system un-
the SETM. Apparently, such a DETM incorporates addi- dergoes external disturbances to seek fast transient re-
tional dynamics into deciding when to sample and/or re- sponse and quick settling, while fewer data packets can

1.0 1.0
x(t) x(t)
0.5 x(tm) 0.5 x(tm)

0 0
0.2
−0.5 0.1 −0.5
0
6.4 6.5
−1.0 −1.0
0 5 10 0 5 10
Time (s) Time (s)
(a) TTM (b) SETM

1.0
x(t) λ(t)
0.20
0.5 x(tm)

0.15
0

0.10
−0.5

0.05
−1.0
0 5 10 0 5 10
Time (s) Time (s)
(c) DETM (d) ADV
Fig. 1 Illustration of periodic triggering and event triggering of the signal of interest x(t) = sin(πt)e−0.3t, t ∈ R : (a) TTM (or periodic
sampling) with tm = mh, m ∈ N; h = 0.01s; (b) SETM of the form tm+1 = inf{t > tm | |x(tm ) − x(t)| > 0.2|x(tm )|}; (c) ADV-based
DETM of the form tm+1 = inf{t > tm | |x(tm ) − x(t)| > 0.2|x(tm )| + 0.5λ(t)}; (d) ADV λ(t) > 0 evolving along
λ̇(t) = −0.1λ(t) − (|x(tm ) − x(t)| − 0.2|x(tm )|), λ(0) = 0.2
860 International Journal of Automation and Computing 18(6), December 2021

be triggered when a system is approaching its steady- features and benefits, and the relevant analysis and
state, and no disturbance is acting on it to relieve net- design techniques for event-triggered control and estima-
work traffic. This also necessitates a DETM to be dynam- tion of resource-constrained networked systems. A gener-
ic and adaptive to respond to different system stability al event-triggered control/estimation framework is firstly
and performance requirements. Letting the fixed presented, which enables several key issues during con-
threshold parameter σ in SETM (1) be dynamically or struction, design, and implementation of desired event
adaptively adjusted, namely in the form of σ(t̃), consti- trigger and controller/estimator to be comprehensively
tutes one of the possible solutions, which leads to the so- examined. An emphasis is then placed on two typical
called dynamic threshold parameter based (DTP-based) classes of DETMs based on ADVs and DTPs. The main
DETM[10]. In a nutshell, DETMs based on either ADVs techniques of constructing ADVs and DTPs and the cor-
responding analysis and design methods are elaborated.
or DTPs introduce extra dynamics and further design
Furthermore, several application examples are presented
freedom to event-triggered systems, which can thus be re-
to demonstrate how and when DETMs outperform the
garded as a promising alternative to the traditional
classical SETMs and TTMs. It is shown through fair
SETMs.
comparisons that DETMs are more flexible and intelli-
The rationale behind an ETM is to selectively ex-
gent to strike a trade-off between desired control/estima-
ecute sampling and/or transmission actions in order to ef-
tion performance and satisfactory resource efficiency,
ficiently accomplish various control and estimation tasks.
which is useful for identifying scenarios where dynamic
Hence, ETMs intrinsically trade real-time control and es-
triggering offers potential benefits.
timation performance for resource efficiency since the de-
Notice that there are several reviews of the advances
signed event-based controllers and estimators are merely
in ETMs in the published literature on different design
executed intermittently. Furthermore, the majority of ex-
objectives and networked systems, such as [11−14] on
isting DETMs make sampling and/or transmission ac-
static event-triggered control of networked control sys-
tions work in a more sporadic fashion to confront a severe
tems, [15] on static sampled-data-based event-triggered
shortage of computation and communication resources. It
control and filtering of networked systems, [16, 17] on
is therefore imaginable that most existing dynamic event-
static event-triggered control and filtering/estimation of
triggered control and estimation approaches may sacri-
networked systems, [18] on static event-triggered distrib-
fice more real-time control and estimation performance in
uted estimation of wireless sensor network-based monitor-
exchange for significantly decreased resource utilization.
ing systems, [19, 20] that focus on static event-triggered
Still, understanding that no single best DETM can meet
consensus of multi-agent systems, and [10] on dynamic
all design objectives and application requirements, there event-triggered distributed coordination control of multi-
is a clear need to present ways of evaluating different agent systems. Meanwhile, a book is edited in [6] to cov-
DETMs that successfully achieve the same control/estim- er the latest developments of static event-based control
ation task in order to understand which one is advantage- and signal processing for a variety of networked systems.
ous and at what cost. It is also noted that some existing A bibliometric analysis of the published results on event-
DETM strategies in the literature have certain limita- based control in the last twenty years is conducted in
tions that hamper their implementation in practice. An [21], which identifies the most relevant articles, authors,
insightful examination of existing DETMs for various institutions, and journals. This survey, however, is dedic-
event-triggered control and estimation problems is also ated to DETMs and reviewing the related studies that
needed. are not covered in the surveys and book mentioned
Albeit the control and estimation theory of sampled- above, paying special attention to those published in re-
data systems and networked control systems has been cent seven years. Moreover, this paper presents a compre-
well developed, there is a lack of mature theory for event- hensive compilation of state-of-the-art dynamic trigger-
triggered systems, not mentioning dynamic triggering. As ing techniques, which will serve as a direct reference for
dynamic triggering ideas gain increasing popularity and interested readers in the field of event-triggered control
application in the field of event-triggered control and es- and estimation. A snapshot of the structure of this sur-
timation, any progress made in DETMs will benefit this vey is shown in Fig. 2.
emerging field as well as the widespread areas of systems
and control, detection, and optimization. The overall aim 2 A general event-triggered control and
of this survey is to emphasize the motivations of DETMs estimation framework
and the wide technical context of dynamic event-
triggered control and estimation, and further promote the
dynamic triggering ideas to other related tasks that can 2.1 A plant
be implemented in a resource-efficient and intelligent
manner. Specifically, this survey presents a comprehens- For simplicity of exposition, the dynamics of the plant
ive review of dynamic triggering techniques, their main are modeled by the following state-space equations:
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 861
{
1 Introduction x̂∆ (t) = Ax̂(t) + Bu(t) + L(y(t) − C x̂(t))
(4)
2 A General framework p̂(t) = F x̂(t) + Gu(t)
2.1 A plant
2.2 An ETM where x̂(t) ∈ Rnx denotes the state estimate of the system
2.3 A Co-design problem state x(t) at time t ; x̂∆ (t) denotes the differential
2.4 Fundamental issues
˙
operator x̂(t) in the continuous-time case, and the one
2.5 A DETM step ahead operator x̂(t + 1) in the discrete-time case,
3 DETMs based on ADVs respectively; p̂(t) ∈ Rnp denotes the output of the
3.1 ETS case estimator; K and L represent the controller and
3.2 ETT case estimator gain matrices to be designed. Obviously, the
3.3 An application example system state x(t) (or measurement output y(t)) needs to
4 DETMs based on DTPs be continually supplied to the controller (3) (or the
4.1 ETS case estimator (4)) at all times t ∈ R in the continuous-time
4.2 ETT case case or t ∈ N in the discrete-time case. Such a
4.3 An application example requirement is inapplicable or even invalid in a networked
4.4 Another application example communication setting where only digitized and sampled
5 Conclusion and some challenging issues
data packets are permitted to be intermittently
Fig. 2 Skeleton structure of this survey transmitted between networked and distributed system
 components.
 ∆
x (t) = Ax(t) + Bu(t) + Ew(t) A common form of an event-triggered control and
y(t) = Cx(t) + Dv(t) (2) scheduling policy πmx
for system (2) is given by


p(t) = F x(t) + Gu(t)
{
x u(t) = Kx(tm ), t ∈ [tm , tm+1 )
πm : (5)
where t denotes the time variable with t ∈ R = {t|t ≥ 0} tm+1 = inf{t̃ > tm | g(ex (t̃)) > σf (X(t̃))}
in the continuous-time case and t ∈ N = {t|t =
0, 1, 2, · · · } in the discrete-time case; x(t) ∈ Rnx denotes and similarly, an event-triggered estimation and
y
the state vector; x∆ (t) denotes the differential operator scheduling policy πm for system (2) is described as
ẋ(t) in the continuous-time case and the one step ahead  ∆
operator x(t + 1) in the discrete-time case, respectively; 
 x̂ (t) = Ax̂(t)+Bu(t)+L(y(tm )−C x̂(t)),

 t ∈ [t , t
u(t) ∈ Rnu denotes the desired control input; w(t) ∈ Rnw y m )
m+1
πm : (6)
represents the general unknown input (e.g., process noise, 
 p̂(t) = F x̂(t) + Gu(t)


external disturbance); y(t) ∈ Rny stands for the system tm+1 = inf{t̃ > tm | g(ey (t̃)) > σf (Y (t̃))}
measurement output recorded by an on-board sensor
device; v(t) ∈ Rnv denotes the unknown measurement where g(ex (t̃)) (or g(ey (t̃))) denotes a function of the
noise; p(t) ∈ Rnp stands for the controlled system output; triggering error ex (t̃) (or ey (t̃)) and X(t̃) (or Y (t̃)) denotes
x(0) represents an initial state; A, B, C, D, E, F and G are either the last triggered or present state (or output).
matrices of appropriate dimensions. It can be seen from the above control (or estimation)
x y
and scheduling policy πm (or πm ) that the data x(t) (or
2.2 An event-triggered mechanism y(t)) will be sampled and/or transmitted only when the
relevant triggering condition is satisfied but not over the
We first recall the traditional state-feedback control- releasing interval [tm , tm+1 ) in the continuous-time case
ler and state estimator (or observer) that employ continu- or in the set of releasing instants {tm , tm + 1, · · · ,
ous or periodic system information. We then present a tm+1 − 1} in the discrete-time case. In this sense, it is ex-
general form of the event-based state-feedback controller pected that the number of data samplings and/or trans-
and state estimator and discuss several key issues that missions over a communication network can be signific-
need to be carefully addressed. For concision and conveni- antly reduced under (5) and (6), which further leads to
ent development, the mathematical notations used in less resource consumption.
ETMs are clarified in Table 1.
The traditional state-feedback controller for system 2.3 An event-triggered control/estima-
(2) takes the following form: tion and scheduling co-design problem

u(t) = Kx(t) (3) The co-design problem to be pursued for system (2) is
stated as follows: For system (2), the objective is to
and the state estimator (or observer) for system (2) is design a suitable event-triggered control (or estimation)
x y
described as and scheduling policy πm of the form (5) (or πm in the
862 International Journal of Automation and Computing 18(6), December 2021

form of (6)) such that the equilibrium point of the result- transmission. The decision of when to transmit the
ing closed-loop system (or estimation error system) is uni- sampled data packet is made by the ETT-based event
formly ultimately bounded, namely, if there exists a com- trigger at every sampling instant kh . It is thus clear that:
pact set S ⊂ Rnx, then for all x(0) ∈ S , there exists a i) The actions of sampling and transmitting take place at
bound δ ≥ 0 and a time ts (δ, x(0)) such that ∥x(t)∥ ≤ δ discrete but different instants. ii) The event triggering in-
(or ∥ψ(t)∥ = ∥x(t) − x̂(t)∥ ≤ δ ) for all t ≥ ts. stants belong to the set of sampling instants, i.e.,
We note the following two points for the problem tm ∈ {kh}. iii) There is no need to persistently monitor
above. the continuous data x(t) (or y(t)). In the literature, there
1) Stability, performance and optimization: Due to the are some alternative terminologies for the ETT-based
presence of the generally unknown inputs w(t) and v(t) in mechanism, such as sampled-data-based ETM[13, 15, 19, 43, 44]
system (2), the bound δ needs to be suitably regulated to or periodic ETM[45−48].
an acceptable level in a practical scenario, which leads to In the discrete-time system case, although the time
bounded/practical solutions of the resulting closed-loop variable t of system (2) and policies (5) and (6) takes val-
system (or estimation error system). Generally, one may ues in the set of non-negative integers, i.e., t ∈ N, it is
distinguish different concepts of stability for the closed- noted that the discrete-time system is often derived via
loop dynamics (or estimation error dynamics), or conver- approximating the original continuous-time system at dis-
gence of the established control/estimation algorithms, cretized instants of time, e.g., t = kh in a periodic
depending on various noise assumptions and different sampling case. From this perspective, the ETM in the
control (or estimation) objectives, such as asymptotic sta- discrete-time system case is similar to the ETT scenario
bility[22, 23], exponential stability[24−26], mean-square stabil- in the continuous-time system case. Hence, unless other-
ity[27−29], finite-time and fixed-time stability[30−34]. wise clarified, we do not explicitly distinguish the dis-
Apart from stability, performance evaluation and op- crete-time and continuous-time cases in the subsequent
timization are also of great importance. In order to meas- discussions.
ure the response quality of the closed-loop (or estimation
error) dynamics, several performance and optimization in- 2.4 Fundamental issues
dices can be suitably explored to evaluate the closed-loop
(or estimation error) system responses, such as bounded The triggers in (5) and (6) can be unified in a general
error covariance[35, 36], H∞ performance[22, 26, 37], mixed form of the trigger in (1). It is noteworthy that a variety
H2 /H∞ performance[38, 39], and set-valued performan- of triggering conditions have been explored in the exist-
ce[40−42]. Nevertheless, formalizing a suitable performance ing literature for the static event trigger (1). For ex-
index and a notion of optimality generally depends on the ample, let the error function be
type of disturbance and noise as well as the concerned 1
system model. g(e(t̃)) ≜ ∥Φ 2 e(t̃)∥2 . (7)
2) Continuous-time and discrete-time system cases: In
the continuous-time system case, the event triggering in- Some common form of the threshold functions
stants {tm } in the policy πm x
(or πmy
) can be determined σf (Z(t̃)) can be given as follows.
by the following two triggering strategies. Constant threshold[41, 49−51]:
ETS. The continuous-time system state x(t) (or meas-
urement output y(t)) is sampled and transmitted over σf (Z(t̃)) ≜ σ, σ ≥ 0. (8a)
some communication networks at the same time. Under
Continuous-data-dependent threshold[52–55]:
such an ETS: i) The actions of sampling and transmit-
ting occur simultaneously. ii) Each triggering instant sat- 1
σf (Z(t̃)) ≜ σ∥Φ 2 z(t)∥2 , σ ≥ 0. (8b)
isfies tm ∈ R and the set {tm } ⊂ [0, ∞) holds. iii) Some
extra hardware/device may be required to continuously
Sampled-data-dependent threshold[43, 56, 57]:
monitor the system state x(t) or measurement output
y(t) in order to judge the triggering condition 1
σf (Z(t̃)) ≜ σ∥Φ 2 z(kh)∥2 , σ ≥ 0. (8c)
g(ex (t̃)) ≥ σf (X(t̃)) (or g(ey (t̃)) ≥ σf (Y (t̃))) and decide
whether an event should be released at any time t̃ . Un- Triggered-data-dependent threshold[44, 58, 59]:
doubtedly, this may increase system expenditure and dif-
ficulty of practical implementation of the trigger. 1
σf (Z(t̃)) ≜ σ∥Φ 2 z(tm )∥2 , σ ≥ 0. (8d)
ETT. The continuous-time system state x(t) (or out-
put y(t)) is firstly sampled at discretized and equidistant Time-dependent threshold[60–62]:
instants of time {kh|k ∈ N}. The sampled data x(kh) (or
y(kh)) with its time-stamp k are then encapsulated into a σf (Z(t̃)) ≜ σ + σ1 e−σ2 t , σ, σ1 ≥ 0, σ2 > 0. (8e)
single data packet (k, x(kh)) (or (k, y(kh))) for possible
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 863

It can be also a combination of the above thresholds[63−66]. least Tw units of time, the other is to introduce an addi-
To guarantee practical implementation of the event tional positive constant threshold into the right-hand side
trigger (1) and well-posedness of the event-triggered con- of the triggering condition[63, 66] (e.g., in (8a) and (8e)) at
trol/estimation and scheduling co-design problem, the fol- the expense of a sacrifice in accurate stability (asymptot-
lowing three fundamental issues are required to be well ic convergence) in exchange for practical/bounded sys-
addressed. tem stability (convergence to a neighborhood around
1) Minimal inter-event time (IET): Event triggers are zero).
often embedded in advanced sensing or transceiver 3) Triggering condition constraint: Under the event
devices. A critical design issue is thus to determine how trigger (1), it is clear that within each IET, there is no
fast the device should release the so-called events, or occurring event, namely,
equivalently, how often the desired remote controller/es-
timator should be updated with the newly arrived data. g(e(t̃)) ≤ σf (Z(t̃)), ∀ t ∈ [tm , tm+1 ).
Such an issue is commonly interpreted as Zeno-freeness or
exclusion of Zeno behavior of the event-based sensor/con- Therefore, the above inequality constraint is required
troller/estimator, i.e., there must exist a strictly positive to be suitably accommodated in the desired analysis and
minimal IET Tmin > 0 such that Tm ≥ Tmin . Only in this design criteria so as to preserve the existence of an ad-
way the event-based sensor/controller/estimator will not missible event-based controller or estimator.
perform an infinite number of updates in a finite time It is further noted that in the continuous-time case,
period on digital platforms. the system behaviour under event-triggered control (or
In the ETT case, it is clear that the IETs satisfy that estimation) is inherently hybrid, which means that both
Tm ≥ h > 0 for all m ∈ N , and the sampling period h continuous as well as discrete signals are incorporated in
guarantees the strict positiveness of the minimal IET. the resulting closed-loop dynamics (or estimation error
However, it is generally difficult to prove the existence of dynamics). This also poses a challenge to the analysis and
a strictly positive lower bound of the IETs for ETS-based design procedures of the event-triggered controller/estim-
triggers. Therefore, ETTs may offer more simplicity and ator.
convenience than ETSs for real-time implementation of
event-based controllers and estimators. 2.5 A dynamic event-triggered mechanism
2) Continuous monitoring VS. periodic sampling: As
mentioned in Section 2.3, the ETS-based triggers dictate Among several ETMs for networked systems, the sub-
the continuous system state x(t) or output y(t) to be sequent focus is put on DETMs because of the intro-
available at all times t ∈ R , which means that some ded- duced extra dynamics and the potential design freedom
icated hardware is demanded to meet such a continuous that will be unfolded hereinafter. Before elaborating on
monitoring requirement. This requirement may increase dynamic event triggers, our discussion shall begin with
the overheads of system monitoring and operation or may the conventional static event triggers that have been
find its technical impossibility in a cyber-physical applica- widely studied in the literature.
tion scenario. Whereas, the ETT-based triggers operate Static event triggers can be arguably referred to as the
only at sampling instants of time, e.g., {kh, k ∈ N}, triggers whose threshold functions σf (Z(t̃)) are depend-
which naturally excludes continuous monitoring and ent on only the system information Z(t̃) (e.g., state, out-
makes them better suited for practical implementation in put) and/or the time information (e.g., e−σ2 t) during the
standard time-sliced embedded hardware and software ar- entire implementation of the ETM. For example, the
chitectures. However, under such an ETT-based trigger, event triggers equipping the thresholds (8a)−(8e) are typ-
the sensor or trigger is configured to sample the system ical SETMs.
state x(t) or output y(t) after each fixed interval of time, Dynamic event triggers are classified as the triggers
no matter whether it is actually needed for preserving whose threshold functions include not only the system in-
stability and performance. This may shorten sensor life- formation Z(t̃) (e.g., state, output) and/or the time in-
time because one of the main causes for energy consump- formation but also some auxiliary variables or dynamic
tion of a real-world sensor devise arises from its persist- parameters possessing their own dynamics. A versatile
ent message listening and sampling[18]. structure of such a DETM can be given as
In the context of ETSs, there are two common tech-
niques that can be adopted to eliminate the continuous 1
monitoring of the system data at all times t ∈ R : One is tm+1 = inf{t̃ > tm | g(e(t̃)) > σ(t̃)f (Z(t̃))+ λ(t̃)} (9)
ϵ
called time regularization[67–69], where a positive time
threshold Tw (acting as time regularization or waiting with ϵ > 0 being a prescribed constant, σ(t̃) ≥ 0 denoting
time between two contiguous events) is inserted during a DTP and λ(t̃) ≥ 0 denoting an ADV. Compared with
the verification of the triggering condition such that the SETM (1), it is clear that the introduction of DTP σ(t̃)
next triggering instant tm+1 is always produced after at and/or ADV λ(t̃) into (9) brings extra dynamics and
864 International Journal of Automation and Computing 18(6), December 2021

sometimes adaptiveness to event triggering decisions. In a (12) in terms of guaranteed resource efficiency, it is
particular case, by fixing σ(t̃) ≡ σ and letting ϵ → +∞, shown in [8, 68] through simulations that the generated
DETM (9) reduces to SETM (1). The DETM of the form IETs under the relevant DETMs are typically larger than
(9) thus offers a comprehensive trade-off analysis between those of their static counterparts (i.e., SETM (12) for [8]
desired system performance and satisfactory resource and SETM under the triggering condition ϕ(o(t), 0) < 0
efficiency than the traditional SETM (1). for [68]). Furthermore, it is formally proved in [8] that for
Dynamic triggering ideas have long been pursued to a given state x(tm ), the next triggering time under
deal with various event-triggered control and estimation DETM (10) will be larger than or equal to that under
DETM (10) SETM (12)
problems in the literature. For example, an ADV-based SETM (12), namely, tm+1 ≥ tm+1 . In other
DETM of the following form: words, the minimal IET for the DETM (10) cannot be
smaller than that for the SETM (12), which serves as the
1
tm+1 = inf{t > tm |g(∥e(t)∥) > σf (∥x(t)∥)+ λ(t)} (10) essential profit motive of the ADV-based DETMs for net-
ϵ
worked systems. Since the minimum IET can be inter-
is firstly developed in [8] to investigate the stability of the preted as how far or safe the event trigger is away from
resulting continuous-time closed-loop nonlinear and linear the Zeno behavior, it is generally believed that an ADV-
control systems under guaranteed minimal IET, where based DETM produces more sporadic data
the ADV λ(t) is given by samplings/transmissions than the relevant static ana-
logue.
λ̇(t) = −α(λ(t)) + σf (∥x(t)∥) − g(∥e(t)∥) (11)
3 Dynamic event-triggered mechanisms
with α(·) denoting a locally Lipschtiz continuous K∞ based on auxiliary dynamic variables
function. The discrete-time case of DETM (10) under
ADV (11) for stability analysis is studied in [9]. It is In this section, depending on the triggering strategies,
proved in [8] that ADV λ(t) in the form of (11) is non- i.e., ETS or ETT, the existing DETMs based on the ADV
negative for any t ∈ R . This implies that in comparison technique are classified and discussed. Note that an focus
to the following SETM: is placed on only the construction of the triggering mech-
anism, specifically, the ADV, while the Zeno-freeness ana-
tm+1 = inf{t > tm | g(∥e(t)∥) > σf (∥x(t)∥)} (12) lysis is left out as one may either formally prove the ex-
istence of a strictly positive minimal IET or employ the
it will be much more stringent for the triggering error time regularization technique or constant threshold tech-
function g(∥e(t)∥) of DETM (10) to exceed the new niques aforementioned.
1 1
threshold σf (∥x(t)∥) + λ(t) because λ(t) ≥ 0 for any
ϵ ϵ
t ∈ R . In [68], a DETM in the form of 3.1 Event-triggered sampling case

tm+1 = inf{t > tm + Tw | λ(t) < 0} (13) Consider the following ADV- and ETS-based trigger-
ing mechanism:
is presented to reduce communication cost while
guaranteeing desired stability and performance criteria 1
tm+1 = inf{t > tm | q σ (t) > λ(t)} (15)
despite the presence of packet losses, where the ADV ϵ
evolves according to 1 1
where q σ (t) ≜ ∥Φ 2 e(t)∥2 − σ∥Φ 2 Z(t̃)∥2 with e(t) = z(tm )−
 z(t) and σ ≥ 0 , ϵ > 0 being two given constants.
λ̇(t) = ϕ(o(t), λ(t)), λ(t) ≥ 0

1) Continuous-time case: The ADV λ(t) in (15) can be
λ(t+ ) = λ0 (o(t)), λ(t) = 0 & release successful (14) defined as follows:

 +
λ(t ) = λ(t), λ(t) = 0 & release failed
λ̇(t) = −µλ(t) − χq σ (t), t ∈ [tm , tm+1 ) (16)
with o(t) denoting all the information locally available at
the trigger. On account of the time regularization, the with µ ≥ 1/ϵ > 0 and χ > 0 denoting two prescribed
DETM (13) subject to (14) generates the next event constants and λ(0) ≥ 0 being a given initial condition.
always after at least Tw time units even in the presence of Note that a salient feature of the DETM (15) is that the
disturbance, which thus guarantees Zeno-freeness and introduced ADV λ(t) of the form (16) is non-negative,
preserves robustness of the event trigger. Similar DETMs i.e., λ(t) ≥ 0 holds for all t ∈ R . Indeed, it is clear that
of (13) subject to (14) are studied in [67, 69, 70] for no event is triggered for t ∈ [tm , tm+1 ) and thus the
different problem formulations. 1
following condition that q σ (t) ≤ λ(t) holds. Recalling
Although it is generally difficult to theoretically prove ϵ
χ
that the DETM in the form of (10) outperforms SETM (16), it can be derived that λ̇(t) ≥ −µλ(t) − λ(t) =
ϵ
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 865
( χ)
− µ+ λ(t) ≜−µ̃λ(t) for t ∈ [tm , tm+1 ). Then, condition. Similarly, the ADV λ(t) ≥ 0 holds at all time
ϵ
multiplying its both sides by e µ̃t
and noting that steps t ∈ N. Actually, it is readily seen from (15) that
d µ̃t 1
µ̃t µ̃t
(e λ(t)) = µ̃e λ(t) + e λ̇(t), we further have that q σ (t) ≤ λ(t) for any t ∈ {tm , tm + 1, · · · , tm+1 − 1}.
dt ϵ
d µ̃t Recalling (17), it can be inferred that λ(t + 1) ≥
(e λ(t)) ≥ 0 , t ∈ [tm , tm+1 ). Then integrating its both χ
dt µλ(t) − λ(t) ≜ µ̃λ(t) ≥ · · · ≥ µ̃t+1 λ(0) ≥ 0 by recalling
sides from tm to t and noting µ̃ > 0 , one obtains ϵ
that µ̃ ≜ µ − χ/ϵ ≥ 0 and λ(0) ≥ 0.
that λ(t) ≥ λ(tm )e−µ̃(t−tm ) ≥ λ(tm−1 )e−µ̃(t−tm−1 ) ≥ · · · ≥
Analogously, one may adopt the following Lypaunov
λ(0)e−µ̃t ≥ 0 for t ∈ [tm , tm+1 ) and thus all t ∈ R =
∪∞ function term Vλ = λ(t) to cope with the relevant trigger-
m=0 [tm , tm+1 ). ing condition constraint. Then, one has that ∆Vλ =
During system performance analysis and control
λ(t + 1) − λ(t) = −(1 − µ)λ(t) − χq σ (t) ≤ −(ϵ(1 − µ) + χ)·
design, an additional Lyapunov function term Vλ = λ(t) 1 1
q σ (t) ≜ −µ̃q σ (t) = −µ̃∥Φ 2 e(t)∥2 + µ̃σ∥Φ 2 Z(t̃)∥2.
may be introduced to deal with the inequality constraint
For a class of nonlinear discrete-time systems repres-
posed by the triggering condition under the non-negative
ented by polynomial fuzzy models, a decentralized ver-
ADV (16). Then, it is easy to derive that V̇λ ≤
( ) sion of the DETM (15) subject to (17) is adopted in [78]
1 1
V̇λ + λ(t) − q (t) =
σ
− µ λ(t) − (χ + 1)q σ (t) ≤ for solving an H∞ state feedback control problem. In [79],
ϵ ϵ1 1
−(χ + 1)q σ (t) = −(χ + 1)∥Φ 2 e(t)∥2 + (χ + 1)σ∥Φ 2 Z(t̃)∥2 an ADV-based event-triggered H∞ dynamic output feed-
( )
1 back control method is devised for a class of sensor satur-
since − µ λ(t) ≤ 0 .
ϵ ated systems with external disturbances. In [80], a dy-
The continuous ADV λ(t) in the form of (16) has been namic event-triggered estimation approach under ADV
intensively investigated for various dynamic event- 1
(17) and q σ (t) ≜ ∥Φ 2 e(t)∥2 − σ is presented for a class of
triggered multi-agent coordination control problems. For discrete-time singularly perturbed systems with distrib-
example, two dynamic event-triggered control laws are uted time-delays. In a multi-sensor network setting, a
proposed in [71] to cope with the average consensus prob- novel dynamic event-triggered distributed set-member-
lem for a class of first-order continuous-time multi-agent ship estimation approach is developed in [40] for a class
systems under undirected and connected graphs. In [72],
of discrete-time linear time-varying systems subject to
via the time regularization technique, a hybrid dynamic
unknown-but-bounded process and measurement noises.
event trigger is devised to solve the consensus problem for
A cluster of ellipsoidal sets centered at the computed
general linear multi-agent systems with external disturb-
state estimates are derived for the distributed and net-
ances. In [73], both the dynamic event-triggered leader-
worked sensors such that the plant's true state always
less and leader-follower consensus problems of general lin-
resides in each sensor's bounding ellipsoid at each time
ear multi-agent systems are studied. In [74], the forma-
step regardless of the process and measurement noises. In
tion-containment control of general linear multi-agent
[81], the recursive distributed filtering problem under the
systems is addressed, where the leader-to-leader and fol-
DETM (15) of ADV (17) and a constant threshold is
lower-to-follower communications are regulated by a dy-
studied for a class of discrete nonlinear time-varying sys-
namic event trigger. Recent advances in dynamic event-
tems over Gilbert-Elliott channels.
triggered distributed coordination control can be found in
the survey [10]. For a linear time-invariant networked
system, an ETS-based dynamic event-triggered controller 3.2 Event-triggered transmission case
is designed in [75] with an L2-gain performance guaran-
Consider the ADV-based and ETT-based triggering
tee. In the context of event-triggered estimation, the con-
mechanism of the following form:
tinuous ADV λ(t) has also been widely investigated. To
name a few, a DETM of the form (15) with ADV (16)
1
tm+1 = inf{sm > tm | q σ (sm ) > (18)
1
and q σ (t) ≜ ∥Φ 2 e(t)∥2 − σ is studied in [76] for a class of ϵ
λ(sm )}
continuous-time polynomial nonlinear systems with ex-
1 1
ternal disturbance. In [77], the DETM (15) is employed where q σ (sm ) ≜ ∥Φ 2 e(sm )∥2 − σ∥Φ 2 Z(t̃)∥2 with e(sm ) =
to address the fault estimation and accommodation prob- z(tm ) − z(sm ) and the ADV λ(sm ) being defined as
lem for continuous-time linear systems.
2) Discrete-time case: The ADV λ(t) in (15) can be λ̇(t) = −µλ(t) − χq σ (sm ), t ∈ [sm , sm+1 ) (19)
described by
with µ > 0 and χ > 0 denoting two prescribed constants
λ(t + 1) = µλ(t) − χq σ (t), t ∈ {tm , tm + 1, · · · , tm+1 − 1} and satisfying that ϵµ < χ, and λ(0) ≥ 0 being a given
(17) initial condition.
For any t ∈ [sm , sm+1 ), solving the differential equation
with ϵ > 0, χ > 0, µ ∈ (0, 1) and µ ≥ χ/ϵ denoting some 1
prescribed constants and λ(0) ≥ 0 being a given initial (19) yields that λ(t) = e−µ(t−sm ) λ(sm ) + (1 − e−µ(t−sm ) )·
µ
866 International Journal of Automation and Computing 18(6), December 2021
( −µ(t−sm )
)
(ϵµ + χ)e −χ
(−χq σ (sm )) ≥ λ(sm ). Notice listening and sampling. To further promote resource-effi-
ϵµ
that for all t ∈ [s0 , s1 ), one has that λ(t) ≥ cient control and estimation applications, it is desirable
( ) ( ) to develop refined ETT-based dynamic event triggers
(ϵµ+χ)e−µ(t−s0 ) −χ (ϵµ+χ)e−µh −χ
λ(s0 ) ≥ λ(s0 ) . based on aperiodic/nonuniform sampling instants
ϵµ ϵµ
Letting ϵ ≥ χ(e − 1)/µ , it is clear that λ(t) ≥ 0 for all
µh {sm |m ∈ N, sm ∈ R, 0 < s ≤ sm+1 − sm ≤ s} . In this
t ∈ [s0 , s1 ). Since λ(t) is a continuous function with re- case, the triggers employ only sporadically sampled data
spect to the time variable t , one then has λ(s1 ) ≥ 0. Us- z(sm ) to decide whether or not the sampled data packet
ing the mathematical induction, one can conclude that should be released at time sm. For example, for a class of
stochastic linear systems, two ADV-based dynamic event-
λ(t) ≥ 0 holds for all t ∈ R if ϵ ≥ χ(eµh − 1)/µ (and thus
triggered control strategies are proposed in [88] to pre-
µh < ln(2)).
serve the stability of the resulting closed-loop system un-
During analysis and design, one may consider an addi-
der sporadic measurements and communication delays.
tional Lyapunov function term[82, 83] Vλ = λ(t)|t=sm.
The ADV λ(t) therein is regulated according to some im-
Then, it is derived that V̇λ |t=sm = −µλ(sm ) − χq σ (sm ) ≤
1 1 pulsive differential equations.
(µϵ − χ)q σ (sm ) ≜ −µ̃q σ (sm ) = −µ̃∥Φ 2 e(sm )∥2 + µ̃σ∥Φ 2 Z(t̃)∥2
As in [8], it can be formally proved that for a given
given that µ̃ > 0 . For example, to deal with the limited signal z(tm ), the next triggering time under DETM (15)
bandwidth allocation, the above dynamic event trigger is or (18) or (20) will not be smaller than that under the re-
adopted in [82] for primary-redundancy communication lative SETM in either of the following form:
channels, where both event-triggered primary and re-
dundancy state feedback control laws are designed to tm+1 = inf{t > tm | q σ (t) > 0} (21a)
guarantee the exponential stability of the resulting closed-
loop switched delay system. In [83], the observer-based tm+1 = inf{sm > tm | q σ (sm ) > 0}. (21b)
dynamic event-triggered control problem under power-
constrained denial-of-service attacks is considered. A dy- In other words, the minimum IET of DETM (15) or
namic event trigger based on sampled state estimates is (18) or (20) cannot be less than that of the relative
(21a) (15)
employed to economize the limited bandwidth. SETM, namely, min{Tm } ≤ min{Tm } and
(21b) (18) or (20)
Alternatively, DETM (18) can be modified as min{Tm } ≤ min{Tm }.

1 3.3 An application to dynamic event-

tm+1 = inf{sm > tm | q σ (sm ) > λ(t)} (20)
ϵ
triggered control of an in-vehicle net-
1 worked active suspension system
which involves a continuous ADV λ(t) under µ ≥ . A
ϵ
benefit of such a triggering mechanism is that the ADV
In this section, for comparison purposes, we conduct
λ(t) can be readily shown to be non-negative for all some quantitative performance analysis between the
t ∈ R . More specifically, (19) implies that λ̇(t) = −µλ(t)− ADV-based DETM and the relative SETM by investigat-
χq σ (sm ) ≥ −(µ + χ/ϵ)λ(t) ≜ −µ̃λ(t) for all t ∈ [sm , sm+1 ), ing an event-triggered control problem of an in-vehicle
which further means that λ(t) ≥ λ(sm )e−µ̃(t−sm ) ≥ networked active suspension system which has been
λ(sm−1 )e−µ̃(t−sm−1 ) ≥ · · · ≥ λ(0)e−µ̃t ≥ 0. Then, employing widely studied in the literature[84, 89−92].
1 The schematic diagram of the studied event-triggered
Vλ = λ(t), it can be derived that V̇λ ≤ V̇λ − q σ (sm ) + λ(t)
( ) ϵ active suspension control system is demonstrated in
1 1
= − µ λ(t) − (χ + 1)q (sm ) ≤ −(χ + 1)∥Φ 2 e(sm )∥2 +
σ
Fig. 3, where the quarter vehicle suspension system, con-
ϵ ( )
1 1 sisting of an one-quarter vehicle chassis (represented by a
(χ + 1)σ∥Φ 2 z(tm )∥2 given that − µ λ(t) ≤ 0 .
ϵ sprung mass ms ), a vehicle wheel assembly (represented
However, one drawback of the DETM (20) is that it by an unsprung mass mu ), suspension spring (fs =
demands the continuous ADV λ(t), which seems to break ks (xs (t) − xu (t))) and damper (fd = cs (ẋs (t) − ẋu (t))),
a promise of eliminating the continuous monitoring of any and an elastic pneumatic tire component (with the stiff-
signal under ETT in (20). Similar DETMs with hybrid ness kt and the constant damping coefficient ct), is con-
signals can be also found in [84−87], where the triggering trolled over a digital and shared controller area network.
conditions therein are checked based on both periodically The dynamical model of the 2-DOF quarter vehicle sus-
sampled data and continuously evolving threshold pension system can be described by the continuous-time
parameter σ(t). As a result, these DETMs still need to be state-space equation (2) with x(t) = [x1 (t), x2 (t), x3 (t),
implemented and performed continuously. x4 (t)]T = [xs (t) − xu (t), xu (t) − xr (t), ẋs (t), ẋu (t)]T ∈ R4
As discussed in Section 2.4, most ETT-based DETMs denoting the stacked state vector, w(t) = ẋr (t) ∈ R de-
leverage a periodic sampling paradigm, which may cause noting the road disturbance input vector, u(t) ∈ R rep-
excessive energy consumption on persistent message resenting the desired actuator force, and
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 867

Electronic control
unit (ECU)/
controller
CAN bus

Controller-to-
actuator link
Sensor-to-
Actuator & controller
suspension strut link
Wheel/tire
Vehicle body sensor sensor
Wheel & tire

x(tk) x(kh) x(tk) Delay dk

Controller area network (CAN) Data loss θ(t)
θ(t) x(tk−dk)

Controller
trigger

SETM
Event

1 Vehicle body
−
x(kh) 4 xs
sprung mass ms Zero-order
hold
Sensors

x(tk) fs fd u
x(kh)
Actuator
Wheel xu
DETM unsprung mass mu

ADV/ Tire
kt ct
DTP xr
Road surface
Fig. 3 An event-triggered active suspension control configuration over a controller area network[92]
 0 0 1 −1  For the in-vehicle networked uncertain quarter vehicle
 0 0 0 1 
  active suspension system subject to the external road dis-
 ks cs cs 
A =− 0 −  turbance input w(t) ∈ L2 [0, ∞), the objective is to design
 ms m ms 
 s
 an ETT-based and ADV-based event-based control and
ks kt cs cs + c t
− − x
scheduling policy πm of the following form:
m mu mu mu
 u 
0  0  
 0  
   −1  u(t) = K(t)x̄θ (t) = K(t)θ(t)x(tm ),

 1    t ∈ [tm + dm , tm+1 + dm+1 )
B =
 ms 
 , E =
 0 .
 (22)
   c  

 1  t
tm+1 = inf{sm > tm | q σ (sm ) > 1 λ(sm )}
− ϵ
mu mu
where θ(t) ∈ [θ, θ] ∈ [0, 1] characterizes the data loss ratio
Note that the vehicle loads may not be precisely of the sensor transmission links; dm ∈ [d, d] stands for the
known in practice because of payload changes, which res- transmission delay at time step m induced by the
1 1
ults in the uncertain and time-varying sprung and un- network; q σ (sm ) ≜ ∥Φ 2 (x(tm ) − x(sm ))∥2 − σ∥Φ 2 x(tm )∥2 ,
sprung masses ms (t) and mu (t). In this sense, the Takagi- such that 1) (System stability) the resulting closed-loop
Sugeno fuzzy modeling approach can be employed to ap- system with w(t) ≡ 0 is asymptotically stable; 2) (Ride
proximate and represent the uncertain active suspension comfort) under zero initial condition and nonzero
system in terms of a finite number of fuzzy rules. The in- w(t) ∈ L2 [0, ∞), the vehicle body acceleration
[
terested readers are referred to [92] in more detail for the ks
p(t) = ẍs (t) = ẋ3 (t) = F x(t) + Gu(t) with F = − ,
Takagi-Sugeno fuzzy active suspension system. ] ms (t)
cs cs
The main problem to be addressed is now stated as: 0, − , and G = ms1(t) is minimized, namely,
ms (t) ms (t)
868 International Journal of Automation and Computing 18(6), December 2021

∥p(t)∥2 In the subsequent simulation, we examine the result-

sup ≤ γ for a prescribed performance level
w(t)̸=0 ∥w(t)∥ 2 ing active suspension and control performance as well the
γ > 0; 3) (Road holding) the dynamic tire load resource efficiency under three different ETMs: 1) SETM
kt x2 (t) = kt (xu (t) − xr (t)) does not exceed the minimal in the form of (22) without λ(sm ) and under a fixed
static load (ms + mu )g, i.e., r1 (t) = H1 x(t) ≤ 1 with threshold parameter σ = 0.08; 2) DETMm in the form of
H1 = [0, kt /((ms + mu )g), 0, 0]; and 4) (Suspension
(22) under the sampled-data version of λ(sm ); and 3)
stroke) the suspension deflection x1 (t) = xs (t) − xu (t)
DETMt in the form of (22) under the continuous-time
does not surpass the allowable limit xmax, i.e.,
version of λ(t). In the last two cases, we set λ(0) =
|r2 (t)| = |H2 x(t)| ≤ 1 with H2 = [1/xmax , 0, 0, 0].
2, χ = 1.2, µ = 0.6, ϵ = 1.8, which straightforwardly im-
To solve the problem formulated above, Algorithm 1
1
is presented to achieve the co-design of the desired event- plies that χ > µϵ in DETMm and µ > in DETMt. The
ϵ
triggered controller and DETM co-design algorithm. A vehicle dynamics parameters are the same as those in
detailed derivation of the inequalities can be found [92]. The other design parameters are set as h = 1 ms,
in [92]. γ = 40, ϖ = 0.015, η = 0.01, ν = 10 . By applying the co-
Algorithm 1. Dynamic event-triggered control and design algorithm above, it is found that the problem is
scheduling co-design
feasible in all cases. The bump responses of the con-
1) For suitable scalars, h , γ, d, d, θ , θ , ν , ϖ , η , σ , µ ,
trolled vehicle suspension system under the bump road
ϵ, χ, solve the following linear matrix inequalities: ( ( ))
hb 2πv0
disturbance input, xr (t) = 1 − cos t for
  [ ] 2 lb
Ξij ϕ1i ϕT lb
 T
2j
 t ∈ 0, and xr (t) = 0 otherwise, are presented in
 ϕ1i −ν −2 I 0  < 0, 1 ≤ i ≤ j ≤ 4 (23) v0
  Fig. 4, where hb = 0.06 m and lb = 5 m are the height and
ϕ2j 0 −ν I
2
length of the bump, respectively, and v0 = 45 km/h is the
vehicle forward velocity. One can observe that regardless
[ √ ]
−P ϖU T HqT of the simultaneous presence of the road disturbance
√ < 0, q = 1, 2 (24) xr (t), transmission delay (dm = 2 ms) and data loss
ϖHq U −I
(θ(t) = 0.7), 1) the controlled vehicle body vertical accel-
to find the positive matrices P, Q, R1 , R2 , Φ̃ and matrices eration is greatly suppressed compared with the open-
Ũ , K̃j, where Ξij = 2eT T T
1 P e2 + e1 Qe1 − e3 Qe3 + loop case; and 2) the suspension deflection and the dy-
T 2 T namic tyre load are all regulated below the specified lim-
e2 (d R1 + (d − d + h) R2 )e2 − (e1 − e3 ) R1 (e1 − e3 )−
2

π2 ( its. A further comparison between the resulting control

γ 2 eT6 e6 − (e3 − e4 )T R2 (e3 − e4 ) + 2(eT T
1 + ηe2 ) − U e2 +
4o ) performance and the anticipated resource efficiency un-
Ai U e1 + θ Bi K̃j (e4 + e5 ) + Ei e6 + µ̃σ(e4 + e5 )T × der the three ETMs is provided in Fig. 5. It can be seen
T T T T T o T T
Φ̃(e4 + e5 ) − µ̃e5 Φ̃e5 + 2e1 U Fi e7 + 2e4 θ K̃j Gi e7 + that 1) via inserting the ADV λ(sm ) or λ(t), the DET-
o T T
2eT T
5 θ K̃j Gi e7 − e7 e7 , ϕ1i = [BiT , ηBiT , 0, 0, 0, 0, GT T
i ] , Mm and DETMt can significantly reduce the data trans-
o
ϕ2j = [0, 0, 0, θo K̃j , θo K̃j , 0, 0]; µ̃ = χ − µϵ; θ = (θ + θ)/2, missions than the SETM case due to the prolonged aver-
θo = (θ − θ)/2; and ei, ∀ i = 1, 2, · · · , 7, denoting some age IETs, and 2) although the greatly improved resource
block entry matrices, e.g., e6 = [0, 0, 0, 0, 0, I, 0]. efficiency under DETMm and DETMt, the control per-
if (23) and (24) are infeasible, repeat Step 1) other- formance is compromised to a certain degree compared
wise, determine a feasible solution (Kj , Φ) = (K̃j U −1, with the SETM case, which can be clearly observed from
U −T Φ̃U −1 ) and continue the trajectory of ∥x(t)∥.
2) Set the simulation time Tsim, m = 0, tm = 0 and It is noted from the above simulation results that:
x̃ = x(0) 1) In the context of event-triggered control or estimation,
3) for k = 0 : Tsim /h − 1 do the existing literature testifies vastly via numerical veri-
4) Compute u(t) in (22) under the newly received data fication that the DETM in the form of (9) is capable of
x̃ releasing much fewer data packets over networks than the
5) Derive the system dynamics of x(t) under u(t) relative SETM of the form (1), thereby offering great po-
6) Obtain the present sampled data packet (k, x(kh)) tential for saving more resources. However, the price to
from the sensor/sampler at time step k be paid is that the resulting control or estimation per-
7) Determine the ADV λ(t) in (19) formance of the concerned system is often compromised
8) if q σ (k) > 1ϵ λ(k) at time step k then to some extent. This is reasonable as the central motiva-
9) Release the sampled data packet (k, x(kh)) tion of an ETM is to trade a certain level of system per-
10) Update m = m + 1, tm+1 ← kh and x̃ ← x(tm+1 ) formance for improved resource efficiency, which is espe-
11) else Keep x̃ unchanged cially profitable when communication networks exhibit
12) end if insufficient bandwidth or wireless sensor devices possess
13) end for some finite battery for continual data sampling and
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 869

0.08
2
xmax
0.06

x·3(t) (m/s2)
|x1(t)| (m) 0.04
0
Open-loop

0.02 −2 Closed-loop (SETM)

Closed-loop (DETMm)
Closed-loop (DETMt)
0 −4
0 1 2 3 4 0 1 2 3 4
Time (s) Time (s)

3 000 500

2 000 (ms + mu)g

kt x2(t) (N)

u(t) (N)
0
1 000
Open-loop
Closed-loop (SETM)
0
−500 Closed-loop (DETMm)
Closed-loop (DETMt)
−1 000
0 1
2 3 4 0 1 2 3 4
Time (s) Time (s)
Fig. 4 Bump responses of the open-loop and the resulting closed-loop active suspension system in terms of the suspension stroke
|x1 (t)|, vehicle body vertical acceleration ẋ3 (t), dynamic tire load kt x2 (t), and actuator power u(t) under SETM, DETMm, and DETMt

0.10
SETM
500 DETMm
Tm (s)
|u(t)|

DETMt 0.05

0 0
0 1 2 3 4 0 1 2 3 4
Time (s) SETM tm (s)

1.5
Open-loop
0.5 Closed-loop (SETM) 1.0
Tm (s)

Closed-loop (DETMm)
Closed-loop (DETMt) 0.5
0.4
×10−3
20 0
0.3 0 1 2 3 4
|x(t)|

DETMm tm (s)
10
3
0.2
0
0 2 4 2
Tm (s)

0.1
1

0 0
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
0 1 2 3 4 DETMt tm (s)
Time (s)
4
DETMm 2.65
ADV λ(t)

JTm, h (%)

DETMt
2
2
0.70
0.35
0 0
2 0 1 3 4 SETM DETMm DETMt
Time (s)
Fig. 5 Comparison of the control performance (in terms of control power |u(t)| and trajectory of ∥x(t)∥) and the resource efficiency (in
terms of IET Tm and data packet transmission rate JTm ,h in percentage) of the vehicle active suspension system between SETM,
DETMm and DETMt
870 International Journal of Automation and Computing 18(6), December 2021

broadcasting. 2) There, however, exist certain applica- 4 Dynamic event-triggered mechanisms

tion scenarios that desired control or estimation perform-
based on dynamic threshold parame-
ance should not be sacrificed too much. For example, for
a mobile target tracking application, the latest target ters
state information (e.g., position and velocity) is the key
The threshold parameter σ plays a vital role in
to accomplish accurate and successful target state estima-
scheduling data samplings and/or transmissions under the
tion. In this sense, when a dynamic event-triggered estim-
SETM of the form (1). It is generally true that a larger σ
ation objective is pursued, it is expected that the DETM leads to a lower frequency of data packet transmi-
should release sensor data packets as often as possible ssions[10, 44, 45]. Similarly, different values of σ(t̃) in DETM
such that the desired tracking performance can be main-
(9) affect the ADV λ(t̃) and thus result in varying
tained at a satisfactory level or not disrupted. 1
A maneuvering target tracking example under differ- threshold functions σ(t̃)f (Z(t̃)) + λ(t̃). Therefore, the
ϵ
ent measurement transmission rates: Consider a maneuv- threshold parameter could be deemed as a scheduling
ering target equipped with a global positioning system for parameter that corresponds to data sampling/transmis-
measuring its position. The objective is to design a state sion rate over networks.
estimator based on only the noisy and intermittent posi- It is undesirable to permanently fix the threshold
tion measurements such that the real-time target motion parameter σ during the implementation of the existing
SETMs mainly due to the following two reasons.
can be estimated as accurately as possible.
1) The resilience requirement of an event trigger:
The target motion dynamics under a prescribed com-
Event triggers are often embedded in smart devices such
mand acceleration[93, 94] can be described as the discrete-
as sensors as they are required to monitor, either continu-
time state-space equations (2), where x(t) = [pn (t),
ously or periodically, the data of interest. These devices,
pe (t), vn (t), ve (t)]T with (pn (t), pe (t)) being the northerly
in most cases, possess restricted energy resources, which
and easterly positions and (vn (t), ve (t)) being the relative means that event triggers may not be able to perform ac-
velocity components, A = [1, 0, h, 0; 0, 1, 0, h; 0, 0, 1, 0; 0, curately at a fixed level of scheduling performance due to
0, 0, 1], B = [0, 0, h sin(φ), h cos(φ)]T , E = [2, 1, 1, 1]T , varying power allocations, finite chipset's processing capa-
T
C = [1, 0, 0, 0; 0, 1, 0, 0], D = [1, 2] , F = [0.25, 0.25, 0.25, city, inherent parameter shifts, and changes or runtime
0.25], G = 0, the sampling period h = 1.5 s, the road ori- errors. This thus requests desired event triggers to pos-
entation angle φ = 120 deg (in the counterclockwise direc- sess a certain level of resilience to tolerate these uncer-
tion from due east), the commanded acceleration tainties. On the other hand, the resilience of an event
u(t) = 0.15 m/s2 for t ≤ 60 and u(t) = −0.15 m/s2 for trigger may emanate from inaccurate event detection
t > 60, the disturbance w(t) = 0.3 sin(0.8t)e−0.05t and the caused by either exogenous disturbance, measurement
measurement noise v(t) = 0.2 cos(t)e−0.05t . The target outlier, or even malicious attack/injection. For example,
starts with the initial positions of (10 m, 10 m) during normal operation of an ETS-based static event
trigger with a fixed threshold of 0.1, whenever the trig-
and velocities of (5 m/s, −3 m/s). Choosing 1
gering function that ∥Φ 2 (z(t) − z(tm ))∥2 > 0.1 holds, the
[ ]T continuous data z(t) will be sampled and released.
1.709 8 0.499 7 0.511 9 0.293 9
L= However, when there is a persistent exogenous disturb-
−0.117 6 0.963 7 −0.050 2 0.197 1
ance or malicious data injection, causing the weighting er-
for the estimator (4), it is easy to verify that the matrix ror function always to be greater than 0.1, the notorious
Zeno phenomenon inevitably occurs. In this sense, event
A − LC is Schur stable. To demonstrate the significance
triggers should be resilient enough to inaccurate event de-
of the real-time target motion information for preserving
tection and remain functional. It becomes apparent that a
accurate state estimation performance, we examine three
time-varying or dynamically adjusted threshold paramet-
scheduling scenarios of sensor measurements y(t): 1) 100%
er represents a possible way to circumvent this inaccur-
transmission rate; 2) 80.83% transmission rate; and 3)
ate event detection, and thus may contribute to such a
27.50% transmission rate, where the last two cases are resilience requirement.
randomly determined. The estimation (tracking) 2) The time-varying network traffic and bandwidth
performance of the estimator is provided in Fig. 6. It can status: The majority of existing ETMs are designed to al-
be inferred that the desired estimation performance leviate the utilization of limited computation and/or com-
becomes increasingly deteriorated when the measurement munication resources during the entire system operation
transmission rate decreases. Apparently, the ADV-based and runtime. In other words, they are decisively engaged
DETMs should be carefully exploited in the in preventing data samplings and/or transmissions over
communication scheduling scenarios that the accurate networks as long as the event trigger and the
control/estimation performance is a major concern or the controller/estimator are implemented. It is true, in most
concerned dynamical system exhibits fast evolutions. network communication scenarios, that the computation
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 871

400
Target moving trajectory
90 Estimator (100.00%)
350 Estimator (80.83%)
Estimator (27.50%)
80
Start position
300 End position
70

60 250
||x(t) − x(t)||

pn(t) (m)
50
^

200

40
150
350
30

100 300
20
250
10 50
200
0 550 600 650
0
0 20 40 60 80 100 120 0 200 400 600
Time step t pe(t) (m)
Fig. 6 Comparison of the estimation (tracking) performance (in terms of ∥x(t) − x̂(t)∥) of the moving target tracking system under
different data transmission rates

and/or communication resources are often shared in a iments of the established event-triggered control/estima-
multipurpose way or accessed by other neighboring pro- tion algorithm. In contrast, the DTP-based DETMs,
cesses and tasks, which thus requires the precious re- thanks to the dynamic nature of the threshold paramet-
sources to be occupied as little as possible. Meanwhile, it ers, alleviate these requirements. Still, more importantly,
is also stressed that network conditions essentially vary the DTPs add a certain level of flexibility and intelli-
from time to time. Consequently, the bandwidth may be gence to the desired event triggers via directed schedul-
only busy during some specific peak periods but idle dur- ing in such a way as to permit data transmissions under
ing others. If the traditional SETM and ADV-based real-time network conditions, which will be elaborated in
DETM are employed in this case, some useful data pack- Sections 4.1 and 4.2.
ets (that are important for accomplishing real-time and In what follows, the existing DTP-based DETMs are
accurate control/estimation tasks) might still be preven- classified and reviewed in detail.
ted from being transmitted even during off-peak periods.
An intelligent DETM is thus expected to schedule data 4.1 Event-triggered sampling case
samplings and/or transmissions more frequently when ac-
tual network traffic is low, and real-time bandwidth Consider the following DTP- and ETS-based trigger-
status is idle to gain better control and estimation per- ing mechanism:
formance, while releasing data packets less often when
1 1
network traffic is high, and the bandwidth is busy to tm+1 = inf{t > tm |∥Φ 2 e(t)∥2 −σ(t)∥Φ 2 Z(t̃)∥2 > 0} (25)
yield better resource efficiency. This also motivates a
DTP-based DETM. where e(t) = z(tm ) − z(t) and σ(t) ≥ 0 is a DTP to be
It should be further noted that the existing SETMs suitably designed. Specifically, some common strategies to
and ADV-based DETMs are still possible to sustain a dynamically adjust the DTP are summarized as follows.
trade-off between desired system performance and satis- 1) A monotonically nonincreasing continuous function
factory resource consumption via carefully prescribing the σ(t) is in the following form:
threshold and ADV parameters. Nevertheless, the ascer-
1
tainment of a suitable threshold or ADV parameter for σ̇(t) = −ϵσ 2 (t)∥Φ 2 e(t)∥2 (26)
the desired SETM or DETM often relies on either cer-
tain design experience or extensive simulations and exper- with ϵ > 0 and 0 < σ(0) ≤ σ < 1 being a given initial
872 International Journal of Automation and Computing 18(6), December 2021

condition. Clearly, (26) implies that σ̇(t) ≤ 0, which distributed event-based communication mechanism under
straightforwardly shows the monotonically nonincreasing a time-varying threshold parameter is proposed to cope
property of the DTP and further σ(t) ≤ σ(0) ≤ σ < 1 for with the leader-following consensus problem for multi-
all t ∈ R . Next, one only needs to ensure that σ(t) > 0, agent systems with unknown but-bounded process and
namely, lower-bounded. From (26), one has that measurement noises.
σ̇(t) 1 The DTP in (27b) has yet been widely employed to
− 2 = ϵ∥Φ 2 e(t)∥2, which means that
σ (t) address various event-triggered control and estimation
d(σ −1 (t)) 1 problems under an ETT strategy; see, e.g., [86] on event-
= ϵ∥Φ 2 e(t)∥2. Integrating both sides of the
dt triggered stabilization of a class of networked Takagi-Sug-
equation from 0 to t, we obtain that
∫ t −1 ∫t 1 2
eno fuzzy systems, [87] on decentralized event-triggered
dσ (s) = ϵ∥Φ 2 e(s)∥ ds , which further yields that
0
( 0
)−1 H∞ filtering of a class of interconnected Takagi-Sugeno
∫t 1
σ(t) = σ −1 (0) + 0 ϵ∥Φ 2 e(s)∥2 ds > 0 by noting that fuzzy systems, [84, 92] on event-triggered H∞ control of
σ(0) > 0, ϵ > 0 and Φ > 0 . Therefore, the continuous vehicle active suspension control systems, and [85] on
function σ(t) is monotonically nonincreasing for all t ∈ R automatic steering control of autonomous ground
and satisfies 0 < σ(t) ≤ σ(0) ≤ σ < 1. During the analysis vehicles. Nevertheless, as noted in Section 3.2, when the
and design procedures, one may consider the worst-case continuous DTP σ(t) is embedded in an ETT-based dy-
triggering scenario by including the upper bound σ of namic event trigger, the triggering law therein may still
σ(t) in the derived criteria or employ an additional need to perform continuously rather than periodically,
Lyapunov function term Vσ = (σ −1 (t) − κ)2 /2ϵ to deal which represents a clear limitation of such a dynamic
with the triggering condition constraint under the DTP event trigger.
(26), where κ is a prescribed positive scalar. Then, a 3) An adaptive continuous function σ(t) is one of the
simple calculation yields that V̇σ = (σ −1 (t) − κ) following forms:
1 1 1
∥Φ 2 e(t)∥ ≤ −κ∥Φ 2 e(t)∥ + ∥Φ 2 z(tm )∥ . For example,
2 2 2
1
z(t)∥2
σ(t) = σ + σ1 e−σ2 ∥Φ , t∈R
2
the above DETM is adopted to solve a distributed leader- (28a)
follower consensus problem for a nonlinear multi-agent
system subject to input saturation[95]. 1
z(tm )∥2
σ(t) = σ + σ1 e−σ2 ∥Φ , t ∈ [tm , tm+1 )
2

2) A non-monotonic continuous function σ(t) is one of

(28b)
the following forms:
1
σ ≤ σ(t) ≤ σ, 0 < σ ≤ σ < 1 e(t)∥2
(27a) σ(t) = σ + σ1 e−σ2 ∥Φ , t ∈ [tm , tm+1 )
2
(28c)

1 ( 1 ) 1 where σ2 ≥ 0 and σ1 = σ − σ . It is clear that σ(t) ∈ [σ, σ]

σ̇(t) = − ϵ ∥Φ 2 e(t)∥2 (27b)
σ(t) σ(t) for all the DTPs in (28). Furthermore, a key feature of
the above DTPs is that their values can be adaptively
where 0 < σ ≤ σ(0) and ϵ > 0 in (27b). It is noteworthy adjusted on the interval [σ, σ] based on the weighting
that the DTP σ(t) in (27b) does not exhibit a strictly data or error term. More specifically, when the system
monotonic property. In order to derive the desired data z(t) (or z(tm ) or the triggering error e(t)) suffers
analysis and design criteria under (27b), an additional large fluctuation, a smaller value of σ(t) will be selected
1
Lyapunov function term Vσ = σ 2 (t) can be employed to to verify the event trigger. Generally, a smaller threshold
2
deal with the triggering condition constraint. parameter leads to more data packets to be sampled
Furthermore, it should be noted that σ(t) in (27b) will and/or transmitted over networks with an aim to achieve
eventually converge to some finite steady value if the faster convergence of the resulting closed-loop system or
system approaches its equilibrium point in the presence of estimation error system. On the other hand, when the
vanishing disturbance/noise. On the other hand, the DTP system data z(t) (or z(tm ) or the triggering error e(t))
σ(t) in (27a) allows to be generally unknown and time- experiences little fluctuation, it may mean that the
varying with attainable upper and lower bounds. These system is now approaching its equilibrium point without
bounds can be further exploited in the analysis and external disturbance/noise, and thus a larger threshold
design criteria to guarantee robust control/estimation. parameter σ(t) should be prescribed to reduce
For example, such a DTP (27a) is adopted to deal with a unnecessary data samplings and/or transmissions. From
distributed event-triggered H∞ consensus filtering this perspective, the DTPs in (28) offer certain
problem for a class of discrete-time linear systems over adaptiveness between maintaining desired system
sensor networks[96]. Under a polytope-like transformation performance and resource efficiency.
regarding the DTP, a threshold-parameter-dependent It is shown in [48] that dynamic triggering with an ad-
approach is developed to determine both the desired aptive threshold can be employed to better shape the res-
distributed consensus filters and event triggers. In [97], a ulting network traffic over some shared medium when
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 873

successive packet dropouts occur. Specifically, the pro- where e(sm ) = z(tm ) − z(sm ) and σ(sm ) ≥ 0 is a DTP to
posed adaption technique for adjusting the threshold be designed. Some common strategies for constructing the
parameter depends on the prediction horizon and net- DTP are elaborated below.
work congestion status. For example, when packet losses 1) A monotonically nonincreasing discrete function
take place due to network congestion, the prediction hori- σ(sm ) is of the following forms:
zon becomes shortened, and a larger threshold is thus se-
lected to reduce the transmissions. Obviously, such a dy- σ(kh)
σ((k + 1)h) = 1 (30)
namic triggering mechanism requires an acknowledge- 1 + ϵ1 σ(kh)∥Φ 2 e(kh)∥2
ment scheme that determines the last successful transmis-
sion and the number of consecutive dropouts. A similar with ϵ ≥ 0 and 0 ≤ σ(0) ≤ σ ∈ [0, 1) being a given initial
acknowledgement-based event-triggered protocol is con- condition. Using the mathematical induction, it can be
sidered in [68] for dynamic event-triggered control sys- easily shown that σ((k + 1)h) ≤ σ(kh) ≤ σ(0) ≤ σ < 1
1
tems subject to packet losses, however, under a different because 1 + ϵσ(kh)∥Φ 2 e(kh)∥2 ≥ 1. Therefore, the
dynamic triggering mechanism and problem setup. sequence {σ1 (kh)} is monotonically nonincreasing and
Adaptive techniques have been intensively studied in lower-bounded for all k ∈ N and satisfies that
conventional control literature. It also seems natural to 0 ≤ σ1 (kh) ≤ σ1 (0) ≤ σ < 1.
make the DTP and thus the event trigger dependent on Due to the monotonic nonincreasing property of the
some adaptive gain/parameter in such a way as to gain DTP σ(t) in (27b) or σ(sm ) in (30), more and more data
more adaptiveness during the scheduling and control co- packets may be transmitted and released over networks
design. For example, an adaptive event-triggered control before the system reaches its steady-state or when there
method is presented in [98] for a class of single-input and remains external disturbance acting on the system. As
single-output uncertain nonlinear systems, where the such, the DETM equipped DTP σ(t) (27b) or DTP σ(sm )
DTP and the controller gain are both adaptively adjus- (30) may be advantageous when the network bandwidth
ted via some adaptive weights. However, such an adapt- is identified as idle, or the system seeks fast convergence
ive event-triggered control method may exhibit a poten- and high-performance requirement during its operation.
tial limitation of practical implementation since it re- The DTP of the form (30) has been well explored in
quires both the controller, normally remotely located, and several different control problem formulations; see, e.g.,
the event-trigger, often locally embedded in an intelli- [100] on distributed formation control of linear multi-
gent sensor device, to be synchronously orchestrated at agent systems, [95] on leader-follower consensus control of
all times. In [99], an adaptive event-triggered output-feed- a class of nonlinear and input-saturated multi-agent sys-
back control scheme is developed for a class of upper-tri- tems, [89, 92] on vehicle active suspension control, and
angular uncertain nonlinear systems. Whether or not the [101] on vehicle platooning control of a group of wire-
lessly connected automated vehicles.
observer state should be transmitted is decided by an ad-
2) A monotonically nondecreasing discrete function
aptive event trigger whose threshold parameter is adapt-
σ(sm ) is one of the following forms:
ively adjusted via a dynamic observer gain.
A closer look at the DTPs (26) and (28) reveals that 1

0 < σ(t)|(26) ≤ σ ≤ σ(t)|(28) ≤ σ . This further indicates σ(kh)∥Φ 2 e(kh)∥2 + ϵσ

σ((k + 1)h) = 1 (31a)
that for a given signal z(tm ), the next triggering times for ϵ + ∥Φ 2 e(kh)∥2
the event triggers (25) under DTP (26) and DTP (28)
1
and the static threshold parameter σ(t) ≡ σ satisfy σ(kh) + ϵσ∥Φ 2 e(kh)∥2
DTP(26) σ(t)≡σ DTP(28) σ((k + 1)h) = (31b)
≤ tm+1 ≤ tm+1 . Therefore, the theoretical
1
tm+1 1 + ϵ∥Φ e(kh)∥2
2

relationship between the minimum IETs can be expressed

as min{Tm
DTP(26)
} ≤ min{Tm
σ(t)≡σ
} ≤ min{Tm
DTP(28)
}. with ϵ ≥ 0 and 0 < σ ≤ σ(0) ≤ σ being a given initial
condition. Notice that the upper bound of σ(sm ) in (31)
4.2 Event-triggered transmission case can be easily proved by the mathematical induction. As a
matter of fact, it follows from (31a) and (31b) that
1
In this section, we review some existing DTP-based (σ(kh) − σ)∥Φ 2 e(kh)∥2
σ((k + 1)h) − σ = 1 ≤ 0 and σ((k+
and ETT-based triggering mechanisms. It is noted that ϵ + ∥Φ 2 e(kh)∥2
the exclusion of Zeno behavior follows naturally under σ(kh) − σ
1)h) − σ = ≤ 0 by recalling that σ(kh)−
ETT owe to the positive sampling period h . 1
1 + ϵ∥Φ 2 e(kh)∥2
Consider the following ETT-based triggering mechan-
σ ≤ 0 , ϵ ≥ 0 and Φ > 0 . On the other hand, the
ism in the form of
monotonic and nondecreasing property of {σ(kh)} in (31)
1 1 can be straightforwardly verified by noting that σ((k+
tm+1 = inf{sm > tm | ∥Φ 2 e(sm )∥2 −σ(sm )∥Φ 2 Z(t̃)∥2 > 0}
ϵ(σ − σ2 (kh))
(29) 1)h) − σ(kh) = 1 ≥ 0 and σ((k + 1)h)−
ϵ2 + ∥Φ 2 e(kh)∥2
874 International Journal of Automation and Computing 18(6), December 2021

ϵ(σ − σ(kh))
σ(kh) = 1 ≥ 0 . Therefore, the sequence Notice that a defining feature of the invert tangent
1 + ϵ∥Φ 2 e(kh)∥2 π
function arctan(·) is that it is naturally bounded by ,
{σ(kh)} in (31) is monotonically nondecreasing and ( π π) 2
i.e., arctan(·) ∈ − , . This feature is fully explored
upper-bounded for all k ∈ N and satisfies that 2 2
0 ≤ σ ≤ σ(0) ≤ σ(kh) ≤ σ . As a result, the data in [108, 109] to construct an adaptive DTP-based trigger-
samplings and/or transmissions are to be scheduled in a ing mechanism for addressing event-triggered control
directed manner, i.e., less and less data packets are problems. Specifically, the DTP σ(sm ) in (29) is replaced
expected to be sampled and/or transmitted over networks by σ(tm ) which takes the following form:
to alleviate resource shortage. Clearly, such a DETM may
be beneficial to account for heavy traffic load and busy σ(tm ) = max{σ, κσ(tm−1 )} (32c)
bandwidth scenarios during system operation[102]. ( )
2ϵ ∥z(tm )∥−∥z(tm−1 )∥
An event-triggered filtering scheme based on a similar where σ(0) = σ and κ = 1− arctan
π ∥z(tm )∥
form of DTP (31a) is proposed in [103] to deal with a with ϵ > 0 being a prescribed constant. The following
probabilistic-constrained filter design problem for a class facts are noted: 1) σ(tm ) ≥ σ holds for all tm; 2) if
of time-varying systems with stochastic nonlinearities and ∥z(tm )∥ = ∥z(tm−1 )∥, σ(tm ) = σ ; if ∥z(tm )∥ > ∥z(tm−1 )∥,
state constraints. In [104], an event-triggered predictive
it is clear that κ ∈ (1 − ϵ, 1) and σ(tm ) < σ(tm−1 ); and if
control scheme under a similar DTP in (31b) is studied to
∥z(tm )∥ < ∥z(tm−1 )∥, one has that κ ∈ (1, 1 + ϵ) and thus
tackle the leader-follower consensus problem of discrete-
σ(tm ) > σ(tm−1 ); and 3) the constant ϵ affects the change
time multi-agent systems with communication delays. In
rate of the DTP. The larger ϵ, the larger the change rate
[102], a unified DETM incorporating both the DTP (30)
of σ(tm ). It is clear that the DTP σ(tm ) can adaptively
and DTP (31a) is exploited to deal with the co-design of
adjust its value at each triggering instant based on the
resource-efficient scheduling and platooning control for a
convoy of automated vehicles. Based on a bandwidth ac- latest and past transmitted data to regulate the data
knowledgement parameter, it is shown that the vehicle- transmission rate. For example, if ∥z(tm )∥ < ∥z(tm−1 )∥, a
to-vehicle data transmissions can be dynamically regu- larger DTP σ(tm ) is chosen, which further means that a
lated in accord with the bandwidth status. lower transmission frequency of the sampled data packets
3) An adaptive discrete function σ(sm ) is one of the is configured to gain improved resource efficiency.
following forms: It is also noted that the DTP σ(tm ) in (32c) is rarely
lower-bounded as σ . In many situations, it seems natural
1
to also pose an upper bound σ on the DTP. As a result,
σ(kh) = σ + σ1 e−σ2 ∥Φ z(kh)∥
, k∈N
2
(32a)
the adaptive DTP (32c) can be refined as
1
σ(kh) = σ + σ1 e−σ2 ∥Φ e(kh)∥
, k∈N σ(tm ) = min{max{σ, κσ(tm−1 )}, σ}. (32d)
2
(32b)

where σ2 ≥ 0 and σ1 = σ − σ . Similar to (28), one has Recalling the boundedness of the DTPs (30)−(32), it is
that σ(kh) ∈ [σ, σ]. easy to infer that for a given signal z(tm ), the next trig-
An adaptive DTP in the form of (32a) (with gering times for the event triggers (29) under DTPs
z(kh) = x(sm )) is employed in [105] to initiate a memory- (30)−(32), and the static threshold parameter σ(sm ) ≡ σ
DTP(30) σ(sm )≡σ DTP(31)
based dynamic event-triggered H∞ control approach for a satisfy that tm+1 ≤ tm+1 ≤tm+1 and
DTP(30) σ(s )≡σ DTP(32)
class of continuous-time linear systems with network-in- tm+1 ≤ tm+1m
≤ tm+1 . Analogously, one can es-
duced delays. It is shown that by incorporating a series of tablish the theoretical relationship between the minimum
DTP(30) σ(s )≡σ
previously released data packets, the control performance IETs as 0 < h ≤ min{Tm } ≤ min{Tm m }≤
DTP(31) DTP(30)
can be greatly improved since the latest released dynam- min{Tm } and 0 < h ≤ min{Tm }≤
ic information is well used. Such a memory-based event σ(s )≡σ DTP(32)
min{Tm m } ≤ min{Tm }.
trigger is further explored in [106] to deal with H∞ load
frequency control for power systems over a bandwidth- 4.3 An application to dynamic event-
constrained network subject to deception attacks. Re- triggered control of a mass-spring-
cently, based on the adaptive DTP (32b), a decentral- damper mechanical system
ized co-design approach for dynamic event-triggered com-
munication and active suspension control is developed in In this section, we employ a mass-spring-damper
[107] for a class of in-vehicle networked in-wheel motor- mechanical system as an example to evaluate the effect-
driven electric vehicles subject to dynamic damping. It is iveness and performance of different DTP-based and
demonstrated that the co-design approach is promising ETT-based dynamic event triggers presented in Section
for guaranteeing both prescribed suspension performance 4.2. Specifically, consider the following nonlinear mass-
and satisfactory resource efficiency. spring-damper mechanical system, as also shown in Fig. 7,
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 875

x(tk) x(tk−dk) For comparison purposes, we next examine the follow-

Controller
θ(t)

Network
ing five different ETMs. 1) SETM1: Under a constant
small threshold parameter σ(sm ) = σ ≡ 0.05 ; 2) SETM2:
Zero-order
hold Under a constant large threshold parameter σ(sm ) =
σ ≡ 0.75; 3) DETM1: Under a monotonically nonincreas-
trigger
Event
DTP

Actuator
ing DTP (30) with σ = 0.05 and ϵ1 = 1.5; 4) DETM2:
Under a monotonically nondecreasing DTP (31b) with
x(kh) W u
σ = 0.05, σ = 0.75, and ϵ2 = 0.2; and 5) DETM3: Under
Sensor Mass an adaptive DTP (32b) with σ = 0.05, σ1 = 0.7 and
x
σ2 = 50.
The co-design Algorithm 1 (without (24) and with

Damper
Spring
µ̃ = 1, dm = 20 ms, h = 10 ms, θ(t) = 0.8, γ = 0.7 ,
η = 0.008, ν = 1) is adopted here to obtain the comparat-
ive simulation results. Fig. 8 shows a comparison between
the resulting control performance and the preserved re-
Fig. 7 A dynamic event-triggered control configuration for a source efficiency under the five different ETMs. It can be
mass-spring-damper mechanical system
observed that 1) a larger threshold parameter leads to a
{
much lower number of data packet transmissions over the
ẋ1 (t) = x2 (t)
network. Not surprisingly, SETM2 contributes to the low-
ẋ2 (t) = −0.01x1 (t) − 0.67x31 (t) + w(t) + u(t)
est data packet transmission rate among the other ETMs
due to the largest value of σ(sm ) = σ ≡ 0.75 . In contrast,
where x1 (t) ∈ [−1, 1] denotes the mass displacement; x2 (t)
during its implementation, DETM1 employs the smallest
represents the mass velocity; w(t) = 2.5 sin(πt)e−0.2t
threshold parameter owe to its monotonic nonincreasing
stands for the external disturbance; and u(t) is the
property and upper bound of 0.05. It is thus reasonable
desired control force. Detailed parameter selections of the that DETM1 results in the highest data transmission
above system can be found in [110]. Choosing the fuzzy rate; and 2) although the significantly reduced data
membership functions as α1 (x1 (t)) = 1 − x21 (t) and transmissions and thus improved resource efficiency un-
α2 (x1 (t)) = x21 (t), the inferred Takagi-Sugeno fuzzy der SETM2, the control performance is compromised
system can be given as [56] greatly compared with the SETM1 case, which can be
seen from the trajectory of ∥x(t)∥. Similar conclusions can

 ∑2
be drawn between DETM2 and DETM3 and the other
ẋ(t) = αi (x1 (t)){Ai x(t) + Bi u(t) + Ei w(t)}
ETMs.


i=1
p(t) = F x(t) = [1 0]x(t)
4.4 An application to dynamic event-
[ ] [ ]
0 1 0 1 triggered estimation of a water distri-
with A1 = , A2 = , B1 = B2 = bution and supply system
−0.01 0 −0.68 0

[0, 1]T , E1 = E2 = [0, 1]T, and the initial condition Supervisory control and data acquisition (SCADA)
x(0) = [1, 0]T. represents a system of software and hardware elements. It
The objective is to design an ETT-based and DTP- empowers real-time monitoring and control of geographic-
based event-based control and scheduling policy πmx
of ally dispersed assets, such as electrical power grids, water,
the following form: oil and gas pipelines, and sewage treatment plants, to be
conducted reliably, timely, and remotely. SCADA is re-
 garded as the backbone of modern critical infrastructure,
 ∑2
u(t) = αi Kj θ(t)x(tm ), t ∈ [tm +dm , tm+1 +dm+1 ) including water distribution and supply systems. Among


j=1 the many functions of SCADA systems, state estimation
tm+1 = inf{sm > tm | q σ (sm ) > 0} plays an essential role in achieving an effective monitor-
ing task as it allows the unavailable full system state to
where θ(t) ∈ [θ, θ] ∈ [0, 1] characterizes the data loss ratio; be estimated/observed on the basis of partial and noisy
dm ∈ [d, d] corresponds to the transmission delay; sensor measurements.
1 1
q σ (sm ) ≜ ∥Φ 2 (x(tm ) − x(sm ))∥2 − σ(sm )∥Φ 2 x(tm )∥2 such In what follows, we outline a dynamic event-triggered
that the resulting closed-loop system with w(t) ≡ 0 is estimation framework for a remote SCADA water distri-
asymptotically stable; and under zero initial condition bution and supply system, as shown in Fig. 9. Some ma-
and nonzero w(t) ∈ L2 [0, ∞), the following performance jor components of the concerned system include: 1) two
∥p(t)∥2 waste water treatment plants; two water storage reser-
index sup ≤ γ holds for prescribed γ > 0. voirs (R1 and R2); one water tank (T1); numerous water
w(t)̸=0 ∥w(t)∥2
876 International Journal of Automation and Computing 18(6), December 2021

6 0.3
Closed-loop (SETM1)
Closed-loop (DETM1) 0.2

Tm (s)
5 Closed-loop (DETM2)
Closed-loop (DETM3) 0.1
Closed-loop (SETM2)
4
0
1.5 0 5 10 15 20
|u(t)|

SETM1 tm (s)
3
1.0
0.2
2 0.5

Tm (s)
0.1
1 0
4 5 6
0
0 0 5 10 15 20
0 5 10 15 20 SETM1 tm (s)
Time (s)
1.4 1.0

0.35
1.2
Tm (s)

0.5
0.30
1.0
0.25
0
0.8 0 5 10 15 20
0.20
||x(t)||

DETM2 tm (s)

0.6 0.15

0.10 0.6
0.4
Tm (s)

0.05 0.4
4 5 6
0.2 0.2

0
0 0 5 10 15 20
0 5 10 15 20 DETM3 tm (s)
Time (s)
1.0
0.8
Tm (s)

0.5

0.6
0
0 5 10 15 20
σ (kh)

SETM2 tm (s)
0.4
30
SETM1 24.60
SETM2
20 17.30
JTm, h (%)

DETM1
0.2 DETM2
DETM3
10 6.80 7.45 6.65

0 0
0 5 10 15 20
1

2
M

M
T

ET

ET

ET

Time (s)
SE

SE
D

Fig. 8 Comparison of the control performance (in terms of control power |u(t)| and trajectory of ∥x(t)∥) and the resource efficiency (in
terms of IET Tm and data packet transmission rate JTm ,h in percentage) of the mass-spring-damper system between SETM1, SETM2,
DETM1, DETM2, and DETM3
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 877

State estimator
Anomaly monitor
Supervisory controller
Other functions

y (tm)
u(t)
Wireless communication network

trigger
Event

DTP
y (t) Sink/base station

User 2 Sensor S2
Sensor S1 w(2)(t) y(2)(t)
Reservoir R1
(1) User 1
y (t) Reservoir R2
w(1)(t)
Flow
x(1)(t) x(2)(t) input
Flow Sensor S4
input y(4)(t) u(2)(t)
(1)
u (t)
Sensor S3

Water supply pipeline

y(3)(t)

Water
(3)
tank T1
x (t)

Treatment Treatment
plant 1 plant 2
Pump P1 Pump P2

Fig. 9 A dynamic event-triggered estimation configuration for remote monitoring of a water distribution and supply system

supply pipelines and junctions; two pumps (P1 and P2) w(t) ∈ E(0, Q, 1) and v(t) ∈ E(0, R, 1), where E(c(t),
{ }
for regulating the flow rates of R1 and R2; four wireless Θ, ∆) ≜ z(t) : (z(t) − c(t))T Θ(z(t) − c(t)) ≤ ∆ repres-
sensors for measuring the water pressure heads of R1, R2, ents an ellipsoid enclosing a real vector z(t) ∈ R with a
nz

T1, and the main junction, respectively; two end-users; real vector c(t) ∈ Rnz being the center, a real-valued
and some other hydraulic devices; 2) a remote SCADA matrix Θ = ΘT > 0 being the shape matrix, and a posit-
control center for real-time monitoring and supervising ive scalar ∆ > 0 being the radius of the ellipsoid, respect-
control; and 3) a wireless communication network for en-
ively.
abling data transmissions from wireless sensors to remote
The dynamic event-triggered estimation and schedul-
control center. As in [10, 111−113], the discrete-time ver-
ing co-design problem for the concerned SCADA water
sion of the state-space model (2) is employed to describe
distribution and supply system is formulated as follows:
the SCADA system, where x(t) = [x1 (t),x2 (t), x3 (t)]T de-
For any unknown but bounded disturbance and noise in-
notes the system state that incorporates the pressure
puts w(t) ∈ E(0, Q, 1) and v(t) ∈ E(0, R, 1), the objective
heads of R1, R2 and T1, respectively; u(t) =[u1 (t), u2 (t)]T
is to design a DTP-based event-triggered estimation and
represents the control signals sent to local pumps for reg- y
scheduling policy πm in the form of (6) with the trigger-
ulating the flow rates of R1 and R2; w(t) =[w1 (t), w2 (t)]T
ing law being specified by
stands for the water consumption by two end-users and
represents the external disturbance input to the system; 
tm+1 = inf{t > tm | ∥y(tm )−y(sm )∥ > σ(sm )}
2
y(t) = [y1 (t), y2 (t), y3 (t), y4 (t)]T denotes the sensor meas- 


urements; v(t) is the measurement noise affecting all σ(sm ) = min{max{σ, κσ(sm − 1)}, σ}
( )

 2ϵ ∥y(sm )∥ − ∥y(tm )∥
sensor readings. It is assumed that the disturbance and 
κ = 1 − arctan
noise are unknown but bounded and satisfy π ∥y(sm )∥
878 International Journal of Automation and Computing 18(6), December 2021

such that the SCADA system′s true state x(t + 1) ∈ to perform a trade-off analysis between the desired estim-
E(x̂(t + 1), P, ∆) always holds at every time step t ation performance and the expected event-triggered
regardless of the simultaneous disturbance w(t) and scheduling performance in a unified framework.
measurement noise v(t), namely, there exists a bounding In the following simulation, the system and measure-
ellipsoidal set E(x̂(t + 1), P, ∆) for any t ∈ Z to guarantee ment matrices are given as
always the enclosing of the true state x(t + 1), where the
[0.995 1 0.000 9 0.004 0]
ellipsoid center, represented by the desired state estimate
x̂(t + 1), the shape matrix P = P T > 0, and the radius A = 0.001 2 0.992 2 0.006 6
∆ > 0 are to be determined. 0.016 2 0.019 8 0.996 4
To solve the above problem, Algorithm 2, which out- [0.625 0 0 ]
lines the main steps for the co-design of the DTP-based B= 0 0.833 0
y
event-triggered estimation and scheduling policy πm is 0 0
provided.  
−0.229 3 −0.054 0
Algorithm 2. Dynamic event-triggered estimation
E = −0.095 9 −0.365 7
and scheduling co-design
−1.295 0 −1.187 1
1) For positive scalars σ and η , solve the following
1 0 0 
linear matrix inequality:
 0 1 0 
[ ] C= 
−P Φ 0 0 1
≤0 (33) 0.366 9 0.115 1 0.518 0
ΦT −Λ
1
to find the positive matrix P , matrix L̃, and positive 1
D =  .
scalars ρ1 , ρ2 , ρ3 and ∆ , where Φ = [0, P A − L̃C, 1
−L̃, P E, −L̃D]; Λ = diag{(1 − η)∆ − ρ1 − ρ2 − ρ3 σ , ηP , 1
ρ3 Iny, ρ1 Q, ρ2 R}.
During system operation and runtime, the reservoirs
if (33) is infeasible, repeat Step 1) otherwise, determ- R1 and R2 are replenished with constant flow rates
ine a feasible solution L = P −1 L̃ and continue
u(t) = [0.453 3, 0.553 9]T m3/s; the first customer′s water
2) Set the simulation time Tsim, m = 0, tm = 0 and
ỹ = y(0) usage demand w(1) (t) varies randomly between
0.8~1.0 m3/s for t < 100 and fluctuates randomly between
3) for t = 0 : Tsim do
1.3~1.5 m3/s for t ≥ 100 ; the second customer′s demand
4) Derive x̂(t + 1) in (6) under L and ỹ = y(tm )
w(2) (t) varies randomly between 0.9~1.1 m3/s for t < 120
5) Compute the ellipsoidal state estimate set
1
{x(t + 1) : x(t + 1) = x̂(t + 1) + ∆ 2 Eα, α ∈ Rnx , ∥α∥ ≤ 1} and fluctuates randomly between 1.0~1.2 m3/s for
t ≥ 120 ; and the four sensors experience some persistent
based on x̂(t + 1) and P −1 = EE T.
random measurement noise v(t) that causes the changes
6) Determine the DTP σ(t) at time step t
of −0.3~0.3 m on the their readings. Furthermore, set
7) if ∥y(tm ) − y(t)∥2 > σ(t) at time step t then 1 1
8) Release the sampled data packet (t, y(t)) Q = Inw, R = Inv, x(0) = [100, 80, 60]T m, x̂(0) = [101, 81,
3 2
9) Update m = m + 1, tm+1 ← t and ỹ ← y(tm+1 ) 59]T m, σ(0) = σ = 0.1σ , ϵ = 200, η =0.95. We next exam-
10) else Keep ỹ unchanged ine the resulting estimation performance and event-
11) end if triggered scheduling performance of the SCADA monitor-
12) end for ing system under four different DTPs σ(sm ), i.e., σ = 1,
It is noted from the above co-design algorithm that 1) σ = 10, σ = 20, σ = 50, respectively.
1
the calculated state estimates x̂(t + 1) + ∆ 2 Eα at any Implementing the co-design Algorithm 2 straightfor-
t + 1 in Step 5) form an ellipsoidal set in state-space wardly implies the feasibility of the formulated co-design
rather than a single vector generated by some traditional problem in different cases of DTP σ(sm ). Fig. 10 presents
estimation methods such as Kalman filtering and H∞ es- the comparison results of the resulting estimation and
timation. In this sense, the resulting state estimate ellips- scheduling performance under the four different DTPs. It
oid E(x̂(t + 1), P, ∆) guarantees to always contain all pos- is noted that in the context of ellipsoidal estimation, the
sible values of the true SCADA system state x(t + 1) for conservatism of the resultant bounding ellipsoid lies in its
any t ∈ {0, 1, · · · , Tsim } regardless of the unknown but tightness, i.e., the width between the upper and lower
bounded disturbance w(t) and measurement noise v(t); bounds centered at the state estimate x̂(t). Generally, the
and 2) the derived criterion in terms of inequality (33) tighter the ellipsoid, the less conservative the ellipsoidal
enables the upper bound σ of the DTP and the shape estimation method. It can be observed from Fig. 10 that
matrix P and radius ∆ of the ellipsoid to be jointly de- 1) a smaller DTP generally results in more sensor meas-
signed. Therefore, the co-design algorithm empowers us urement transmissions over the network (and thus sacri-
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 879

110 100

5
105 95

Tm
x1(t) (−), x^1 (-●), bounds (- -)

x2(t) (−), x^2 (-●), bounds (- -)

100 90
0
0 100 200
95 85 tm (Case 1)

20
90 80

Tm
10
85 75

0
80 70 0 100 200
0 100 200 0 100 200 tm (Case 2)
Time step t Time step t
140 60
20
50
DTP σ (sm)

40

Tm
120 30 10
20
x3(t) (−), x^3 (-●), bounds (- -)

10
0 0
100 0 100 200 0 100 200
Time step t tm (Case 3)

80 40
67.50
JTm, h (%)

50
Tm

20
60 25.50
19.00
13.00
0 0
40 0 100 200
1

4
se

se

se

se

0 100 200 tm (Case 4)

Ca

Ca

Ca

Ca

Time step t
Fig. 10 Comparison of the estimation performance (in terms of trajectories of x(t), x̂(t) and their bounds as well as the adaptive DTPs
σ(sm )) and the resource efficiency (in terms of IET Tm and data packet transmission rate JTm ,h in percentage) of the SCADA water
distribution and supply system under different adaptive DTPs: Case 1 (red): σ(sm ) ∈ [σ, σ] = [0.1, 1]; Case 2 (blue):
σ(sm ) ∈ [σ, σ] = [1, 10]; Case 3 (magenta): σ(sm ) ∈ [σ, σ] = [2, 20]; Case 4 (green): σ(sm ) ∈ [σ, σ] = [5, 50].

ficed resource efficiency) and leads to tighter ellipsoids, tions, and the relevant analysis and design techniques.
and 2) a larger DTP contributes to fewer data transmis- Two representative classes of DETMs based on ADVs
sions but gives rise to a larger bound width of the ellips- and DTPs have been discussed in detail, followed by a re-
oidal estimate set. This further means that the designed view of the existing results on event-triggered control and
state estimator sacrifices its confidence to provide an ac- estimation that use these DETMs. Furthermore, several
curate estimate ellipsoid in exchange for a resource ex- practically motivated examples have been provided to
penditure reduction. evaluate the performance of different DETMs.
Research on dynamic event-triggered control and es-
5 Conclusions and some challenging timation has attracted intensive attention in the past sev-
issues eral years. This paper covers a small proportion of the
vast literature and is by no means complete. For ex-
The recent advances in dynamic event-triggered con- ample, we have not discussed some closely relevant top-
trol and estimation of networked systems have been re- ics in the field, such as dynamic event-triggered optimiza-
viewed. In order to cater to various control and estima- tion[114, 115], self-triggered control and estimation[12, 116−118],
tion objectives, a general event-triggered control and es- and stochastic event-triggered control and estimation
timation framework has been presented. Then, a focus based on random thresholds[119−121]. In what follows, we
has been placed on the introduction and motivation of outline some challenging issues worthy of further study
DETMs, and their main features, benefits and limita- for dynamic event-triggered control and estimation.
880 International Journal of Automation and Computing 18(6), December 2021

1) Novel DETMs with less or easily-tunable trigger be efficiently co-designed.

parameters: To evaluate a traditional SETM of the form 4) Significance-based DETMs: Under an ETM of the
(1), the determination of the static threshold parameter form (1), the data of interest z(t̃) is sampled and/or
σ for preserving both desired control/estimation perform- transmitted only if it is deemed as significant for achiev-
ance and satisfactory resource efficiency often requires ing the desired system performance instead of based on
either certain design experience or extensive simulations the progression of time. Specifically, the significance of
and experiments. This is particularly the case in a data is characterized by its amplitude variation g(e(t̃))
DETM, where several trigger parameters need to be suit- exceeding some well-defined threshold σf (Z(t̃)). The ex-
ably chosen and tuned to achieve desired control/estima- isting DETMs mostly focus on designing appropriate and
tion performance. Developing a suitable DETM, which in- flexible thresholds such that the triggers are more sensit-
volves less or easily tunable trigger parameters, for gener- ive to the data amplitude variations. An interesting yet
al event-triggered control/estimation applications, de- open question is how to develop an effective DETM for
serves further investigation. better system performance but with fewer released events.
2) Asynchronous DETMs over feedback (sensor-to- The question, however, seems paradoxical because fewer
controller) and forward (controller-to-actuator) channels: triggering events indicate fewer data packets for control-
In networked control systems, control loops are closed via ler/estimator implementation and design, and thus gener-
communication networks, which makes clock synchroniza- ally degraded system performance. A possible way of ad-
tion between local sensors and remote controllers essen- dressing this question is to look closely at what data is
tially challenging and expensive. On the other hand, it is virtually significant for better performance guarantees.
not uncommon that transmission delays, data packet dro- For example, some attempts have been made via employ-
pouts, and packet disorder may occur when the triggered ing a range of previously triggered data packets at the
sensor measurements and control commands are propag- trigger's side to characterize data amplitude variations,
ated over communication networks. This may result in leading to the so-called memory-based ETMs[105, 106]. In
asynchronous time series of triggered data packets. The this case, the significance can be enhanced by the suit-
existing literature on dynamic event-triggered control of- ably weighted historical data packets. However, it is
ten assumes that the timing regime is the same for all shown that these memory-based ETMs may result in
system components. This may lead to inapplicability in more triggering events near troughs/crests of the system
cyber-physical scenarios, where sensors and actuators are trajectories or when the system exhibits drastic evolu-
collocated with the plant while controllers are spatially tions. How they can be further exploited for both better
distributed and remotely configured. Hence, how to tackle system performance and resource efficiency remains open.
asynchronous dynamic event-triggered control requires On the other hand, in some control systems, frequent
further exploration. data feedings are not preferable for desired control per-
3) Bandwidth-aware DETMs: Some DTP-based formance improvement. For example, it is demonstrated
DETMs presented in Section 4 have the potential to par- in [122] that intentionally discarding some control input
tially address the bandwidth-aware scheduling issue. For packets can reduce heading deviation and rudder oscilla-
example, the DETMs under DTP (27b) and DTP (30) tion of an unmanned surface vehicle in network environ-
generally lead to more often data samplings and/or trans- ments. When a DETM is adopted, this means that those
missions than their static counterparts to seek better con- data packets due to fast/drastic amplitude changes may
trol/estimation performance. As a result, they may be not need to be released in order to mitigate unnecessary
employed when the bandwidth resource is sufficient, or deviation and oscillation. In this case, the significance
the system demands fast convergence during its opera- evaluated by the DETM may not necessarily be related
tion. In contrast, the DETMs under DTP (31) and the to significant data amplitude variations. Hence, how to
ADV-based DETMs in Section 3 may find their applicab- develop novel DETMs that emphasize the significance of
ility when the bandwidth appears constantly busy. Never- the data of interest for better system performance and re-
theless, some novel DETMs have not been adequately ex- source efficiency requires deep investigation.
plored in the event-triggered control/estimation literat- 5) Resilient dynamic event-triggered control/estima-
ure. These DETMs are aware of the real-time bandwidth tion approaches against malicious attacks: Compared
status, where events are generated less often if the real- with the traditional SETMs, most DETMs are capable of
time network channels are overloaded, and vice-versa. It transmitting data packets much less frequently. This im-
should be mentioned that the research of event-triggered plies that only those data packets that are deemed as sig-
control/estimation necessitates an integrated view of both nificant by the DETM are released over the network to
control/estimation theory and communication theory. preserve desired control/estimation performance.
There is a clear need to develop new dynamic event- However, a sophisticated attacker may leverage this fact
triggered control/estimation techniques that incorporate to maliciously manipulate these significant data packets
real-time network dynamics and bandwidth status such during network transmission to disrupt the system per-
that bandwidth allocation and controller/estimator can formance. Typical attacks on the data transmission chan-
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 881

nels include false data injection attacks, which tamper [6] M. Miskowicz. Event-Based Control and Signal Pro-
data integrity, and denial-of-service (DoS) attacks, which cessing, Boca Raton, USA: CRC Press, 2016.
cause data interruption/unavailability. This thus makes [7] L. Zou, Z. D. Wang, Q. L. Han, D. H. Zhou. Moving hori-
zon estimation for networked time-delay systems under
the resilient dynamic event-triggered control/estimation
Round-Robin protocol. IEEE Transactions on Automat-
issue in the presence of attacks particularly important. ic Control, vol. 64, no. 12, pp. 5191–5198, 2019. DOI:
However, to the best of the authors' knowledge, develop- 10.1109/TAC.2019.2910167.
ing resilient dynamic event-triggered control/estimation [8] A. Girard. Dynamic triggering mechanisms for event-
approaches that consider realistic attack models and en- triggered control. IEEE Transactions on Automatic Con-
sure the survivability of the event-triggered system des- trol, vol. 60, no. 7, pp. 1992–1997, 2015. DOI: 10.1109/
TAC.2014.2366855.
pite malicious attacks has not been fully addressed, which
[9] S. L. Hu, D. Yue, X. X. Yin, X. P. Xie, Y. Ma. Adaptive
calls for additional research effort.
event-triggered control for nonlinear discrete-time sys-
tems. International Journal of Robust and Nonlinear
Open Access Control, vol. 26, no. 18, pp. 4104–4125, 2016. DOI:
10.1002/rnc.3550.
This article is licensed under a Creative Commons At- [10] X. H. Ge, Q. L. Han, L. Ding, Y. L. Wang, X. M. Zhang.
tribution 4.0 International License, which permits use, Dynamic event-triggered distributed coordination con-
sharing, adaptation, distribution and reproduction in any trol and its applications: A survey of trends and tech-
niques. IEEE Transactions on Systems, Man, and Cyber-
medium or format, as long as you give appropriate credit netics: Systems, vol. 50, no. 9, pp. 3112–3125, 2020. DOI:
to the original author(s) and the source, provide a link to 10.1109/TSMC.2020.3010825.
the Creative Commons licence, and indicate if changes [11] M. Lemmon. Event-triggered feedback in control, estima-
were made. tion, and optimization. Lecture Notes in Control and In-
The images or other third party material in this art- formation Science, vol. 406, pp. 293−358, 2010. DOI:
10.1007/978-0-85729-033-5_9.
icle are included in the article′s Creative Commons li-
cence, unless indicated otherwise in a credit line to the [12] W. P. M. H. Heemels, K. H. Johansson, P. Tabuada. An
introduction to event-triggered and self-triggered control.
material. If material is not included in the article′s Creat- In Proceedings of the 51st IEEE Conference on Decision
ive Commons licence and your intended use is not per- and Control, IEEE, Maui, USA, pp. 3270-3285, 2012.
mitted by statutory regulation or exceeds the permitted DOI: 10.1109/CDC.2012.6425820.
use, you will need to obtain permission directly from the [13] C. Peng, F. Q. Li. A survey on recent advances in event-
copyright holder. triggered communication and control. Information Sci-
To view a copy of this licence, visit http://creative- ences, vol. 457-458, pp. 113–125, 2018. DOI: 10.1016/
j.ins.2018.04.055.
commons.org/licenses/by/4.0/.
[14] X. M. Zhang, Q. L. Han, X. H. Ge, D. R. Ding, L. Ding,
D. Yue, C. Peng. Networked control systems: A survey of
Acknowledgements trends and techniques. IEEE/CAA Journal of Automat-
ica Sinica, vol. 7, no. 1, pp. 1–17, 2020. DOI: 10.1109/
This work was supported by the Australian Research JAS.2019.1911651.
Council Discovery Early Career Researcher Award (No. [15] X. M. Zhang, Q. L. Han, B. L. Zhang. An overview and
DE200101128). deep investigation on sampled-data-based event-
triggered control and filtering for networked systems.
IEEE Transactions on Industrial Informatics, vol. 13,
References no. 1, pp. 4–16, 2017. DOI: 10.1109/TII.2016.2607150.
[1] J. Ackermann. Sampled-Data Control Systems: Analysis [16] L. Zou, Z. D. Wang, D. H. Zhou. Event-based control and
and Synthesis, Robust System Design, Berlin, Germany: filtering of networked systems: A survey. International
Springer-Verlag, 1985. Journal of Automation and Computing, vol. 14, no. 3,
pp. 239–253, 2017. DOI: 10.1007/s11633-017-1077-8.
[2] T. W. Chen, B. Francis. Optimal Sampled-Data Control
Systems, London, UK: Springer-Verlag, 1995. [17] Z. Y. Chen, Q. L. Han, Y. M. Yan, Z. G. Wu. How often
should one update control and estimation: Review of net-
[3] L. Hetel, C. Fiter, H. Omran, A. Seuret, E. Fridman, J. P.
worked triggering techniques. Science China Information
Richard, S. L. Niculescu. Recent developments on the
Sciences, vol. 63, no. 5, Article number 150201, 2020. DOI:
stability of systems with aperiodic sampling: An over-
10.1007/s11432-019-2637-9.
view. Automatica, vol. 76, pp. 309–335, 2017. DOI:
10.1016/j.automatica.2016.10.023. [18] X. H. Ge, Q. L. Han, X. M. Zhang, L. Ding, F. W. Yang.
Distributed event-triggered estimation over sensor net-
[4] L. F. Ma, Z. D. Wang, Q. L. Han, H. K. Lam. Variance-
works: A survey. IEEE Transactions on Cybernetics,
constrained distributed filtering for time-varying systems
vol. 50, no. 3, pp. 1306–1320, 2020. DOI: 10.1109/TCYB.
with multiplicative noises and deception attacks over
2019.2917179.
sensor networks. IEEE Sensors Journal, vol. 17, no. 7,
pp. 2279–2288, 2017. DOI: 10.1109/JSEN.2017.2654325. [19] L. Ding, Q. L. Han, X. H. Ge, X. M. Zhang. An overview
of recent advances in event-triggered consensus of multia-
[5] H. El Ghor, E. H. M. Aggoune. Energy efficient scheduler
gent systems. IEEE Transactions on Cybernetics, vol. 48,
of aperiodic jobs for real-time embedded systems. Inter-
no. 4, pp. 1110–1123, 2018. DOI: 10.1109/TCYB.2017.
national Journal of Automation and Computing, vol. 17,
2771560.
no. 5, pp. 733–743, 2020. DOI: 10.1007/s11633-016-0993-
3. [20] C. Nowzari, E. Garcia, J. Cortés. Event-triggered com-
882 International Journal of Automation and Computing 18(6), December 2021

munication and control of networked systems for multi- DOI: 10.1109/TII.2018.2817248.

agent consensus. Automatica, vol. 105, pp. 1–27, 2019.
[35] D. Han, Y. L. Mo, J. F. Wu, S. Weerakkody, B. Sinopoli,
DOI: 10.1016/j.automatica.2019.03.009.
L. Shi. Stochastic event-triggered sensor schedule for re-
[21] E. Aranda-Escolástico, M. Guinaldo, R. Heradio, J. mote state estimation. IEEE Transactions on Automatic
Chacon, H. Vargas, J. Sánchez, S. Dormido. Event-based Control, vol. 60, no. 10, pp. 2661–2675, 2015. DOI: 10.11
control: A bibliometric analysis of twenty years of re- 09/TAC.2015.2406975.
search. IEEE Access, vol. 8, pp. 47188–47208, 2020. DOI:
[36] L. Wang, Z. Wang, Q. L. Han, G. Wei. Event-based vari-
10.1109/ACCESS.2020.2978174.
ance-constrained H∞ filtering for stochastic parameter
[22] X. M. Zhang, Q. L. Han. Event-based H∞ filtering for systems over sensor networks with successive missing
sampled-data systems. Automatica, vol. 51, pp. 55–69, measurements. IEEE Transactions on Cybernetics,
2015. DOI: 10.1016/j.automatica.2014.10.092. vol. 48, no. 3, pp. 1007–1017, 2018. DOI: 10.1109/TCYB.
2017.2671032.
[23] X. M. Zhang, Q. L. Han, X. H. Ge, L. Ding. Resilient con-
trol design based on a sampled-data model for a class of [37] X. H. Ge, Q. L. Han, M. Y. Zhong, X. M. Zhang. Distrib-
networked control systems under denial-of-service at- uted Krein space-based attack detection over sensor net-
tacks. IEEE Transactions on Cybernetics, vol. 50, no. 8, works under deception attacks. Automatica, vol. 109,
pp. 3616–3626, 2020. DOI: 10.1109/TCYB.2019.2956137. Article number 108557, 2019. DOI: 10.1016/j.automatica.
2019.108557.
[24] X. F. Wang, M. Lemmon. On event design in event-
triggered feedback systems. Automatica, vol. 47, no. 10, [38] D. J. Du, B. Qi, M. R. Fei, C. Peng. Multiple event-
pp. 2319–2322, 2011. DOI: 10.1016/j.automatica.2011. triggered H2/H∞ filtering for hybrid wired-wireless net-
05.027. worked systems with random network-induced delays. In-
formation Sciences, vol. 325, pp. 393–408, 2015. DOI:
[25] P. Tallapragada, J. Cortés. Event-triggered stabilization
10.1016/j.ins.2015.07.026.
of linear systems under bounded bit rates. IEEE Transac-
tions on Automatic Control, vol. 61, no. 6, pp. 1575–1589, [39] M. Davoodi, N. Meskin, K. Khorasani. Event-triggered
2016. DOI: 10.1109/TAC.2015.2480215. multiobjective control and fault diagnosis: A unified
framework. IEEE Transactions on Industrial Informatics,
[26] A. Selivanov, E. Fridman. Event-Triggered H∞ control: vol. 13, no. 1, pp. 298–311, 2017. DOI: 10.1109/TII.
A switching approach. IEEE Transactions on Automatic 2016.2541669.
Control, vol. 61, no. 10, pp. 3221–3226, 2016. DOI:
10.1109/TAC.2015.2508286. [40] X. H. Ge, Q. L. Han, Z. D. Wang. A dynamic event-
triggered transmission scheme for distributed set-mem-
[27] G. Battistelli, L. Chisci, D. Selvi. A distributed Kalman
bership estimation over wireless sensor networks. IEEE
filter with event-triggered communication and guaran-
Transactions on Cybernetics, vol. 49, no. 1, pp. 171–183,
teed stability. Automatica, vol. 93, pp. 75–82, 2018. DOI:
2019. DOI: 10.1109/TCYB.2017.2769722.
10.1016/j.automatica.2018.03.005.
[41] D. R. Ding, Z. D. Wang, Q. L. Han. A set-membership
[28] Á. Cuenca, D. J. Antunes, A. Castillo, P. García, B. A.
Khashooei, W. P. M. H. Heemels. Periodic event- approach to event-triggered filtering for general nonlin-
triggered sampling and dual-rate control for a wireless ear systems over sensor networks. IEEE Transactions on
networked control system with applications to UAVs. Automatic Control, vol. 65, no. 4, pp. 1792–1799, 2020.
IEEE Transactions on Industrial Electronics, vol. 66, DOI: 10.1109/TAC.2019.2934389.
no. 4, pp. 3157–3166, 2019. DOI: 10.1109/TIE.2018. [42] D. R. Ding, Q. L. Han, X. H. Ge, J. Wang. Secure state
2850018. estimation and control of cyber-physical systems: A sur-
[29] S. X. Luo, F. Q. Deng. On event-triggered control of non- vey. IEEE Transactions on Systems, Man, and Cybernet-
linear stochastic systems. IEEE Transactions on Auto- ics: Systems, vol. 51, no. 1, pp. 176–190, 2021. DOI:
matic Control, vol. 65, no. 1, pp. 369–375, 2020. DOI: 10.1109/TSMC.2020.3041121.
10.1109/TAC.2019.2916285. [43] D. Yue, E. G. Tian, Q. L. Han. A delay system method
[30] B. Hu, Z. H. Guan, M. Y. Fu. Distributed event-driven for designing event-triggered controllers of networked
control for finite-time consensus. Automatica, vol. 103, control systems. IEEE Transactions on Automatic Con-
pp. 88–95, 2019. DOI: 10.1016/j.automatica.2019.01.026. trol, vol. 58, no. 2, pp. 475–481, 2013. DOI: 10.1109/
TAC.2012.2206694.
[31] B. D. Ning, Q. L. Han, Z. Y. Zuo. Practical fixed-time
consensus for integrator-type multi-agent systems: A [44] C. Peng, T. C. Yang. Event-triggered communication
time base generator approach. Automatica, vol. 105, and H∞ control co-design for networked control systems.
pp. 406–414, 2019. DOI: 10.1016/j.automatica.2019. Automatica, vol. 49, no. 5, pp. 1326–1332, 2013. DOI:
04.013. 10.1016/j.automatica.2013.01.038.

[32] B. D. Ning, Q. L. Han. Prescribed finite-time consensus [45] W. P. M. H. Heemels, M. C. F. Donkers, A. R. Teel. Peri-
tracking for multiagent systems with nonholonomic odic event-triggered control for linear systems. IEEE
chained-form dynamics. IEEE Transactions on Automat- Transactions on Automatic Control, vol. 58, no. 4,
ic Control, vol. 64, no. 14, pp. 1686–1693, 2019. DOI: pp. 847–861, 2013. DOI: 10.1109/TAC.2012.2220443.
10.1109/TAC.2018.2852605. [46] E. Garcia, Y. C. Cao, D. W. Casbeer. Periodic event-
[33] H. Yu, T. W. Chen. On Zeno behavior in event-triggered triggered synchronization of linear multi-agent systems
finite-time consensus of multi-agent systems. IEEE with communication delays. IEEE Transactions on Auto-
Transactions on Automatic Control, 2020. DOI: 10.11 matic Control, vol. 62, no. 1, pp. 366–371, 2017. DOI:
09/TAC.2020.3030758. 10.1109/TAC.2016.2555484.
[34] Z. Y. Zuo, Q. L. Han, B. D. Ning, X. H. Ge, X. M. Zhang. [47] X. Y. Meng, L. H. Xie, Y. C. Soh. Asynchronous periodic
An overview of recent advances in fixed-time cooperative event-triggered consensus for multi-agent systems. Auto-
control of multiagent systems. IEEE Transactions on In- matica, vol. 84, pp. 214–220, 2017. DOI: 10.1016/j.auto-
dustrial Informatics, vol. 14, no. 6, pp. 2322–2334, 2018. matica.2017.07.008.
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 883

[48] S. Linsenmayer, D. V. Dimarogonas, F. Allgöwer. Period- [62] D. P. Yang, W. Ren, X. D. Liu, W. S. Chen. Decentral-
ic event-triggered control for networked control systems ized event-triggered consensus for linear multi-agent sys-
based on non-monotonic Lyapunov functions. Automat- tems under general directed graphs. Automatica, vol. 69,
ica, vol. 106, pp. 35–46, 2019. DOI: 10.1016/j.automatica. pp. 242–249, 2016. DOI: 10.1016/j.automatica.2016.
2019.04.039. 03.003.
[49] J. F. Wu, Q. S. Jia, K. H. Johansson, L. Shi. Event-based [63] M. C. F. Donkers, W. P. M. H. Heemels. Output-based
sensor data scheduling: Trade-off between communica- event-triggered control with guaranteed L∞-gain and im-
tion rate and estimation quality. IEEE Transactions on proved and decentralized event-triggering. IEEE Trans-
Automatic Control, vol. 58, no. 4, pp. 1041–1046, 2013. actions on Automatic Control, vol. 57, no. 6,
DOI: 10.1109/TAC.2012.2215253. pp. 1362–1376, 2012. DOI: 10.1109/TAC.2011.2174696.
[50] D. R. Ding, Z. D. Wang, D. W. C. Ho, G. L. Wei. Observ- [64] W. Zhu, Z. P. Jiang. Event-based leader-following con-
er-based event-triggering consensus control for multia- sensus of multi-agent systems with input time delay.
gent systems with lossy sensors and cyber-attacks. IEEE IEEE Transactions on Automatic Control, vol. 60, no. 5,
Transactions on Cybernetics, vol. 47, no. 8, pp. 1936–1947, pp. 1362–1367, 2015. DOI: 10.1109/TAC.2014.2357131.
2017. DOI: 10.1109/TCYB.2016.2582802.
[65] Y. Cheng, V. Ugrinovskii. Event-triggered leader-follow-
[51] M. Kooshkbaghi, H. J. Marquez. Event-triggered dis- ing tracking control for multivariable multi-agent sys-
crete-time cubature Kalman filter for nonlinear dynamic- tems. Automatica, vol. 70, pp. 204–210, 2016. DOI:
al systems with packet dropout. IEEE Transactions on 10.1016/j.automatica.2016.04.003.
Automatic Control, vol. 65, no. 5, pp. 2278–2285, 2020.
DOI: 10.1109/TAC.2019.2945286. [66] C. Peng, D. Yue, M. R. Fei. A higher energy-efficient
sampling scheme for networked control systems over
[52] P. Tabuada. Event-triggered real-time scheduling of sta- IEEE 802.15.4 wireless networks. IEEE Transactions on
bilizing control tasks. IEEE Transactions on Automatic Industrial Informatics, vol. 12, no. 5, pp. 1766–1774, 2016.
Control, vol. 52, no. 9, pp. 1680–1685, 2007. DOI: DOI: 10.1109/TII.2015.2481821.
10.1109/TAC.2007.904277.
[67] V. S. Dolk, D. P. Borgers, W. P. M. H. Heemels. Output-
[53] X. F. Wang, M. D. Lemmon. Event-triggering in distrib- based and decentralized dynamic event-triggered control
uted networked control systems. IEEE Transactions on
with guaranteed Lp-gain performance and Zeno-freeness.
Automatic Control, vol. 56, no. 3, pp. 586–601, 2011. DOI:
IEEE Transactions on Automatic Control, vol. 62, no. 1,
10.1109/TAC.2010.2057951.
pp. 34–49, 2017. DOI: 10.1109/TAC.2016.2536707.
[54] D. V. Dimarogonas, E. Frazzoli, K. H. Johansson. Dis-
[68] V. Dolk, W. Heemels. Event-triggered control systems
tributed event-triggered control for multi-agent systems.
under packet losses. Automatica, vol. 80, pp. 143–155,
IEEE Transactions on Automatic Control, vol. 57, no. 5,
2017. DOI: 10.1016/j.automatica.2017.02.029.
pp. 1291–1297, 2012. DOI: 10.1109/TAC.2011.2174666.
[69] D. P. Borgers, V. S. Dolk, W. P. M. H. Heemels. Riccati-
[55] E. Garcia, P. J. Antsaklis. Model-based event-triggered
based design of event-triggered controllers for linear sys-
control for systems with quantization and time-varying
network delays. IEEE Transactions on Automatic Con- tems with delays. IEEE Transactions on Automatic Con-
trol, vol. 58, no. 2, pp. 422–434, 2013. DOI: 10.1109/ trol, vol. 63, no. 1, pp. 174–188, 2018. DOI: 10.1109/
TAC.2012.2211411. TAC.2017.2713047.

[56] C. Peng, Q. L. Han, D. Yue. To transmit or not to trans- [70] A. Q. Fu, J. A. McCann. Dynamic decentralized periodic
mit: A discrete event-triggered communication scheme event-triggered control for wireless cyber-physical sys-
for networked Takagi-Sugeno fuzzy systems. IEEE tems. IEEE Transactions on Control Systems Techno-
Transactions on Fuzzy Systems, vol. 21, no. 1, logy, 2020. DOI: 10.1109/TCST.2020.3016131.
pp. 164–170, 2013. DOI: 10.1109/TFUZZ.2012.2199994. [71] X. L. Yi, K. Liu, D. V. Dimarogonas, K. H. Johansson.
[57] W. Chen, D. R. Ding, X. H. Ge, Q. L. Han, G. L. Wei. Dynamic event-triggered and self-triggered control for
H∞ containment control of multiagent systems under multi-agent systems. IEEE Transactions on Automatic
event-triggered communication scheduling: The finite-ho- Control, vol. 64, no. 8, pp. 3300–3307, 2019. DOI:
rizon case. IEEE Transactions on Cybernetics, vol. 50, 10.1109/TAC.2018.2874703.
no. 4, pp. 1372–1382, 2020. DOI: 10.1109/TCYB.2018. [72] G. L. Zhao, C. C. Hua. A hybrid dynamic event-triggered
2885567. approach to consensus of multi-agent systems with ex-
[58] X. M. Zhang, Q. L. Han. A decentralized event-triggered ternal disturbances. IEEE Transactions on Automatic
dissipative control scheme for systems with multiple Control, 2020. DOI: 10.1109/TAC.2020.3018437.
sensors to sample the system outputs. IEEE Transac- [73] W. L. He, B. Xu, Q. L. Han, F. Qian. Adaptive consensus
tions on Cybernetics, vol. 46, no. 12, pp. 2745–2757, 2016. control of linear multiagent systems with dynamic event-
DOI: 10.1109/TCYB.2015.2487420. triggered strategies. IEEE Transactions on Cybernetics,
[59] X. H. Ge, Q. L. Han. Distributed event-triggered H∞ fil- vol. 50, no. 7, pp. 2996–3008, 2020. DOI: 10.1109/TCYB.
tering over sensor networks with communication delays. 2019.2920093.
Information Sciences, vol. 291, pp. 128–142, 2015. DOI: [74] A. Amini, A. Asif, A. Mohammadi. Formation-contain-
10.1016/j.ins.2014.08.047. ment control using dynamic event-triggering mechanism
[60] G. S. Seyboth, D. V. Dimarogonas, K. H. Johansson. for multi-agent systems. IEEE/CAA Journal of Auto-
Event-based broadcasting for multi-agent average con- matica Sinica, vol. 7, no. 5, pp. 1235–1248, 2020. DOI:
sensus. Automatica, vol. 49, no. 1, pp. 245–252, 2013. 10.1109/JAS.2020.1003288.
DOI: 10.1016/j.automatica.2012.08.042. [75] D. Liu, G. H. Yang. Dynamic event-triggered control for
[61] J. H. Zhang, G. Feng. Event-driven observer-based out- linear time-invariant systems with H∞-gain perform-
put feedback control for linear systems. Automatica, ance. International Journal of Robust and Nonlinear
vol. 50, no. 7, pp. 1852–1859, 2014. DOI: 10.1016/j.auto- Control, vol. 29, no. 2, pp. 507–518, 2019. DOI: 10.1002/
matica.2014.04.026. rnc.4403.
884 International Journal of Automation and Computing 18(6), December 2021

[76] Y. C. Niu, L. Sheng, M. Gao, D. H. Zhou. Dynamic event- j.automatica.2019.05.028.

triggered state estimation for continuous-time polynomi-
[89] Y. P. Guan, Q. L. Han, H. J. Yao, X. H. Ge. Robust
al nonlinear systems with external disturbances. IEEE
event-triggered H∞ controller design for vehicle active
Transactions on Industrial Informatics, vol. 17, no. 6,
pp. 3962–3970, 2021. DOI: 10.1109/TII.2020.3015004. suspension systems. Nonlinear Dynamics, vol. 94,
pp. 627–638, 2018. DOI: 10.1007/s11071-018-4381-0.
[77] X. D. Wang, Z. Y. Fei, T. Wang, L. Yang. Dynamic
event-triggered actuator fault estimation and accommod- [90] Z. Y. Fei, X. D. Wang, M. Liu, J. Y. Yu. Reliable control
ation for dynamical systems. Information Sciences, for vehicle active suspension systems under event-
vol. 525, pp. 119–133, 2020. DOI: 10.1016/j.ins.2020. triggered scheme with frequency range limitation. IEEE
03.016. Transactions on Systems, Man, and Cybernetics: Sys-
tems, vol. 51, no. 3, pp. 1630–1641, 2021. DOI: 10.1109/
[78] G. L. Zhao, C. C. Hua, X. P. Guan. Decentralized dynam- TSMC.2019.2899942.
ic event-triggered H∞ control for nonlinear systems with
unreliable communication channel and limited band- [91] L. Liu, X. S. Li. Event-triggered tracking control for act-
width. IEEE Transactions on Fuzzy Systems, vol. 29, ive seat suspension systems with time-varying full-state
no. 4, pp. 757–771, 2021. DOI: 10.1109/TFUZZ.2020. constraints. IEEE Transactions on Systems, Man, and
2965877. Cybernetics: Systems published online, 2020. DOI:
10.1109/TSMC.2020.3003368.
[79] Z. Q. Zuo, P. F. Xie, Y. J. Wang. Output-based dynamic
event-triggering control for sensor saturated systems with [92] X. H. Ge, I. Ahmad, Q. L. Han, J. Wang, X. M. Zhang.
external disturbance. Applied Mathematics and Compu- Dynamic event-triggered scheduling and control for
tation, vol. 374, Article number 125043, 2020. DOI: vehicle active suspension over controller area network.
10.1016/j.amc.2020.125043. Mechanical Systems and Signal Processing, vol. 152, Art-
icle number 107481, 2021. DOI: 10.1016/j.ymssp.
[80] L. Ma, Z. D. Wang, C. X. Cai, F. E. Alsaadi. Dynamic 2020.107481.
event-triggered state estimation for discrete-time singu-
larly perturbed systems with distributed time-delays. [93] D. Simon, T. L. Chia. Kalman filtering with state equal-
IEEE Transactions on Systems, Man, and Cybernetics: ity constraints. IEEE Transactions on Aerospace and
Systems, vol. 50, no. 9, pp. 3258–3268, 2020. DOI: Electronic Systems, vol. 38, no. 1, pp. 128–136, 2002. DOI:
10.1109/TSMC.2018.2876203. 10.1109/7.993234.

[81] Q. Li, B. Shen, Z. D. Wang, W. G. Sheng. Recursive dis- [94] E. Mousavinejad, F. W. Yang, Q. L. Han, L. Vlacic. A
tributed filtering over sensor networks on Gilbert-Elliott novel cyber attack detection method in networked con-
channels: A dynamic event-triggered approach. Automat- trol systems. IEEE Transactions on Cybernetics, vol. 48,
ica, vol. 113, Article number 108681, 2020. DOI: no. 11, pp. 3254–3264, 2018. DOI: 10.1109/TCYB.2018.
10.1016/j.automatica.2019.108681. 2843358.

[82] S. L. Hu, Z. H. Cheng, D. Yue, C. X. Dou, Y. S. Xue. [95] X. X. Yin, D. Yue, S. L. Hu. Adaptive periodic event-
Bandwidth allocation-based switched dynamic triggering triggered consensus for multi-agent systems subject to in-
control against DoS attacks. IEEE Transactions on Sys- put saturation. International Journal of Control, vol. 89,
tems, Man, and Cybernetics: Systems, published online, no. 4, pp. 653–667, 2016. DOI: 10.1080/00207179.2015.
2019. DOI: 10.1109/TSMC.2019.2956945. 1088967.
[83] S. L. Hu, D. Yue, Z. H. Cheng, X. P. Xie, X. L. Chen. Co- [96] X. H. Ge, Q. L. Han, Z. D. Wang. A threshold-parameter-
design of dynamic event-triggered communication dependent approach to designing distributed event-
scheme and resilient observer-based control under aperi- triggered H∞ consensus filters over sensor networks.
odic DoS attacks. IEEE Transactions on Cybernetics, IEEE Transactions on Cybernetics, vol. 49, no. 4,
2020. DOI: 10.1109/TCYB.2020.3001187. pp. 1148–1159, 2019. DOI: 10.1109/TCYB.2017.2789296.
[84] H. Y. Li, Z. X. Zhang, H. C. Yan, X. P. Xie. Adaptive [97] X. H. Ge, Q. L. Han, and F. W. Yang. Event-based set-
event-triggered fuzzy control for uncertain active suspen- membership leader-following consensus of networked
sion systems. IEEE Transactions on Cybernetics, vol. 49, multi-agent systems subject to limited communication re-
no. 12, pp. 4388–4397, 2019. DOI: 10.1109/TCYB.2018. sources and unknown-but-bounded noise. IEEE Transac-
2864776. tions on Industrial Electronics, vol. 64, no. 6,
pp. 5045–5054, 2017. DOI: 10.1109/TIE.2016.2613929.
[85] W. F. Li, Z. C. Xie, J. Zhao, P. K. Wong. Velocity-based
robust fault tolerant automatic steering control of [98] Y. X. Li, G. H. Yang. Observer-based fuzzy adaptive
autonomous ground vehicles via adaptive event triggered event-triggered control codesign for a class of uncertain
network communication. Mechanical Systems and Signal nonlinear systems. IEEE Transactions on Fuzzy Systems,
Processing, vol. 143, Article number 106798, 2020. DOI: vol. 26, no. 3, pp. 1589–1599, 2018. DOI: 10.1109/TFUZZ.
10.1016/j.ymssp.2020.106798. 2017.2735944.
[86] Z. Gu, E. G. Tian, J. L. Liu. Adaptive event-triggered [99] F. Z. Li, Y. G. Liu. Adaptive event-triggered output-feed-
control of a class of nonlinear networked systems. Journ- back controller for uncertain nonlinear systems. Auto-
al of The Franklin Institute, vol. 354, no. 9, pp. 3854–3871, matica, vol. 117, Article number 109006, 2020. DOI:
2017. DOI: 10.1016/j.jfranklin.2017.02.026. 10.1016/j.automatica.2020.109006.
[87] Z. Gu, P. Shi, D. Yue, Z. T. Ding. Decentralized adaptive [100] X. H. Ge, Q. L. Han. Distributed formation control of
event-triggered H∞ filtering for a class of networked non- networked multi-agent systems using a dynamic event-
linear interconnected systems. IEEE Transactions on Cy- triggered communication mechanism. IEEE Transac-
bernetics, vol. 49, no. 5, pp. 1570–1579, 2019. DOI: tions on Industrial Electronics, vol. 64, no. 10,
10.1109/TCYB.2018.2802044. pp. 8118–8127, 2017. DOI: 10.1109/TIE.2017.2701778.
[88] S. X. Luo, F. Q. Deng, W. H. Chen. Dynamic event- [101] S. X. Wen, G. Guo, B. Chen, X. E. Gao. Cooperative ad-
triggered control for linear stochastic systems with aptive cruise control of vehicles using a resource-efficient
sporadic measurements and communication delays. communication mechanism. IEEE Transactions on Intel-
Automatica, vol. 107, pp. 86–94, 2019. DOI: 10.1016/ ligent Vehicles, vol. 4, no. 1, pp. 127–140, 2019. DOI:
X. Ge et al. / Dynamic Event-triggered Control and Estimation: A Survey 885

10.1109/TIV.2018.2886676. pp. 969-974, 2018. DOI: 10.1109/CDC.2018.8619311.

[102] X. Ge, S. Xiao, Q. L. Han, X. M. Zhang, D. Ding. Dynam- [115] Y. S. Li, D. W. Gao, W. Gao, H. G. Zhang, J. G. Zhou.
ic event-triggered scheduling and platooning control co- Double-mode energy management for multi-energy sys-
design for automated vehicles over vehicular ad-hoc net- tem via distributed dynamic event-triggered Newton-
works. IEEE/CAA Journal of Automatica Sinica, 2021. Raphson algorithm. IEEE Transactions on Smart Grid,
DOI: 10.1109/JAS.2021.1004060. vol. 11, no. 6, pp. 5339–5356, 2020. DOI: 10.1109/TSG.
2020.3005179.
[103] E. G. Tian, Z. D. Wang, L. Zou, D. Yue. Probabilistic-
constrained filtering for a class of nonlinear systems with [116] X. Wang, M. Lemmon. Self-triggered feedback control
improved static event-triggered communication. Interna- systems with finite-gain L∞ stability. IEEE Transac-
tional Journal of Robust and Nonlinear Control, vol. 29, tions on Automatic Control, vol. 54, no. 3, pp. 452–467,
no. 5, pp. 1484–1498, 2019. DOI: 10.1002/rnc.4447. 2009. DOI: 10.1109/TAC.2009.2012973.
[104] X. X. Yin, D. Yue, S. L. Hu, H. P. Zhang. Distributed ad- [117] U. Tiberi, C. Fischione, K. H. Johansson, M. D. Di Bene-
aptive model-based event-triggered predictive control for detto. Energy-efficient sampling of networked control
consensus of multiagent systems. International Journal of systems over IEEE 802.15.4 wireless networks. Automat-
Robust and Nonlinear Control, vol. 28, no. 18, ica, vol. 49, no. 3, pp. 712–724, 2013. DOI: 10.1016/j.auto-
pp. 6180–6201, 2018. DOI: 10.1002/rnc.4370. matica.2012.11.046.
[105] E. G. Tian, K. Y. Wang, X. Zhao, S. B. Shen, J. L. Liu. [118] Y. Tang, H. J. Gao, J. Kurths. Robust H∞ self-triggered
An improved memory-event-triggered control for net- control of networked systems under packet dropouts.
worked control systems. Journal of The Franklin Insti- IEEE Transactions on Cybernetics, vol. 46, no. 12,
tute, vol. 356, no. 13, pp. 7210–7223, 2019. DOI: pp. 3294–3305, 2016. DOI: 10.1109/TCYB.2015.2502619.
10.1016/j.jfranklin.2019.06.041.
[119] F. D. Brunner, D. Antunes, F. Allgöwer. Stochastic
[106] E. G. Tian, C. Peng. Memory-based event-triggering H∞ thresholds in event-triggered control: A consistent policy
load frequency control for power systems under decep- for quadratic control. Automatica, vol. 89, pp. 376–381,
tion attacks. IEEE Transactions on Cybernetics, vol. 50, 2018. DOI: 10.1016/j.automatica.2017.12.043.
no. 11, pp. 4610–4618, 2020. DOI: 10.1109/TCYB.2020.
2972384. [120] L. Xu, Y. L. Mo, L. H. Xie. Remote state estimation with
stochastic event-triggered sensor schedule and packet
[107] I. Ahmad, X. H. Ge, Q. L. Han. Decentralized dynamic drops. IEEE Transactions on Automatic Control, vol. 65,
event-triggered communication and active suspension no. 11, pp. 4981–4988, 2020. DOI: 10.1109/TAC.2020.300
control of in-wheel motor driven electric vehicles with dy- 4328.
namic damping. IEEE/CAA Journal of Automatica Sin-
ica, vol. 8, no. 5, pp. 971–986, 2021. DOI: 10.1109/ [121] H. Yu, J. Shang, T. W. Chen. On stochastic and determ-
JAS.2021.1003967. inistic event-based state estimation. Automatica,
vol. 123, Article number 109314, 2021. DOI: 10.1016/
[108] C. Peng, M. J. Yang, J. Zhang, M. R. Fei, S. L. Hu. Net- j.automatica.2020.109314.
work-based H∞ control for T-S fuzzy systems with an ad-
aptive event-triggered communication scheme. Fuzzy [122] Y. L. Wang, Q. L. Han. Network-based heading control
Sets and Systems, vol. 329, pp. 61–76, 2017. DOI: and rudder oscillation reduction for unmanned surface
10.1016/j.fss.2016.12.011. vehicles. IEEE Transactions on Control Systems Techno-
logy, vol. 25, no. 5, pp. 1609–1620, 2017. DOI: 10.1109/
[109] C. Peng, J. Zhang, H. C. Yan. Adaptive event-triggering TCST.2016.2617321.
H∞ load frequency control for network-based power sys-
tems. IEEE Transactions on Industrial Electronics,
vol. 65, no. 2, pp. 1685–1694, 2018. DOI: 10.1109/TIE. Xiaohua Ge received the B. Eng. degree
2017.2726965. in electronics and information engineering
[110] K. Tanaka, T. Ikeda, H. O. Wang. Robust stabilization of from Nanchang Hangkong University,
a class of uncertain nonlinear systems via fuzzy control: China in 2008, the M. Eng. degree in con-
Quadratic stabilizability, H∞ control theory, and linear trol theory and control engineering from
matrix inequalities. IEEE Transactions on Fuzzy Sys- Hangzhou Dianzi University, China in
tems, vol. 4, no. 1, pp. 1–13, 1996. DOI: 10.1109/91. 2011, and the Ph. D. degree in computer
481840. engineering from Central Queensland Uni-
versity, Australia in 2014. From 2011 to
[111] Y. L. Mo, R. Chabukswar, B. Sinopoli. Detecting integ-
rity attacks on SCADA systems. IEEE Transactions on 2013, he was a research assistant with Centre for Intelligent and
Control Systems Technology, vol. 22, no. 4, Networked Systems, Central Queensland University, where he
pp. 1396–1407, 2014. DOI: 10.1109/TCST.2013.2280899. was a research fellow in 2014. From 2015 to 2017, he was a re-
search fellow with Griffith School of Engineering, Griffith Uni-
[112] V. L. Do, L. Fillatre, I. Nikiforov, P. Willett. Security of versity, Australia. He is currently a senior lecturer with School of
SCADA systems against cyber-physical attacks. IEEE Software and Electrical Engineering, Swinburne University of
Aerospace and Electronic Systems Magazine, vol. 32, Technology, Australia. He is a Highly Cited Researcher accord-
no. 5, pp. 28–45, 2017. DOI: 10.1109/MAES.2017.160047.
ing to Clarivate Analytics. He is a Senior Member of IEEE. He
[113] X. H. Ge, Q. L. Han, X. M. Zhang, D. R. Ding, F. W. received the 2019 IEEE Systems, Man, and Cybernetics Society
Yang. Resilient and secure remote monitoring for a class Andrew P. Sage Best Transactions Paper Award, the 2020
of cyber-physical systems against attacks. Information IEEE/CAA Journal of Automatica Sinica Outstanding Review-
Sciences, vol. 512, pp. 1592–1605, 2020. DOI: 10.1016/ er Award, the 2017 IEEE Transactions on Cybernetics Outstand-
j.ins.2019.10.057. ing Reviewer Award, and the 2017 International Journal of
[114] W. Du, X. L. Yi, J. George, K. H. Johansson, T. Yang. Automation and Computing Outstanding Reviewer Award. He is
Distributed optimization with dynamic event-triggered an Associate Editor of the IEEE Transaction on Systems, Man,
mechanisms. In Proceedings of the 57th IEEE Confer- and Cybernetics: Systems.
ence on Decision and Control, IEEE, Miami, USA, His research interests include networked, event-triggered, se-
886 International Journal of Automation and Computing 18(6), December 2021

cure, and intelligent control and estimation, and their applica- research fellow and a lecturer with School of Engineering and
tions in autonomous vehicles and connected vehicles. Technology, Central Queensland University, Australia. From
E-mail: [email protected] 2014 to 2016, he was a lecturer with the Griffith School of Engin-
ORCID iD: 0000-0003-0180-0897 eering, Griffith University, Australia. In 2016, he joined the
Swinburne University of Technology, Australia, where he is cur-
rently an associate professor with School of Software and Elec-
Qing-Long Han received the B. Sc. de-
trical Engineering. He is a Highly Cited Researcher according to
gree in mathematics from Shandong Nor-
Clarivate Analytics. He is a Senior Member of IEEE. He was a
mal University, China in 1983, and the
M.Sc. and Ph. D. degrees in control engin- recipient of second National Natural Science Award in China in
eering from East China University of Sci- 2013, and first Hunan Provincial Natural Science Award in Hun-
ence and Technology, China in 1992 and an Province in China in 2011, both jointly with Prof. M. Wu and
1997, respectively. He is Pro Vice-Chancel- Prof. Y. He. He was also a recipient of the 2020 IEEE Transac-
lor (Research Quality) and a Distin- tions on Industrial Informatics Outstanding Paper Award, the
guished Professor at Swinburne Uni- 2019 IEEE Systems, Man, and Cybernetics Society Andrew P.
Sage Best Transactions Paper Award, and the 2016 IET Control
versity of Technology, Australia. He held various academic and
Theory and Applications Premium Award. He is an Associate
management positions at Griffith University and Central
Queensland University, Australia. He was a Highly Cited Re- Editor of the IEEE Transactions on Cybernetics, Journal of the
searcher in both Engineering and Computer Science (Clarivate Franklin Institute, International Journal of Control, Automation,
Analytics, 2019−2020). He was one of Australia′s Top 5 Lifetime and Systems, Neurocomputing, and Neural Processing Letters.
Achievers (Research Superstars) in Engineering and Computer His research interests include H∞ filtering, event-triggered
Science (The Australian′s 2020 Research Magazine). He was the control systems, networked control systems, neural networks,
recipient of the 2021 M. A. Sargent Medal (the Highest Award of distributed systems, and time-delay systems.
the Electrical College Board of Engineers Australia), the 2020 E-mail: [email protected]
IEEE Systems, Man, and Cybernetics Society Andrew P. Sage ORCID iD: 0000-0003-0691-5386
Best Transactions Paper Award, the 2020 IEEE Transactions on
Industrial Informatics Outstanding Paper Award, and the 2019 Derui Ding received the B. Sc. degree in
IEEE SMC Society Andrew P. Sage Best Transactions Paper industry engineering and the M. Sc. de-
Award. He is a Fellow of the Institute of Electrical and Electron- gree in detection technology and automa-
ic Engineers and the Institution of Engineers Australia. He has tion equipment from Anhui Polytechnic
served as an AdCom Member of IEEE Industrial Electronics So- University, China in 2004 and 2007, and
ciety (IES), a Member of IEEE IES Fellow Committee, and the Ph. D. degree in control theory and
Chair of IEEE IES Technical Committee on Networked Control control engineering from Donghua Uni-
Systems. He is Co-Editor of Australian Journal of Electrical and versity, China in 2014. From July 2007 to
Electronic Engineering, an Associate Editor for 12 international December 2014, he was a teaching assist-
journals, including IEEE Transactions on Cybernetics, IEEE ant and then a lecturer in Department of Mathematics, Anhui
Transactions on Industrial Informatics, IEEE Industrial Elec- Polytechnic University, China. From June 2012 to September
tronics Magazine, IEEE/CAA Journal of Automatica Sinica, 2012, he was a research assistant in Department of Mechanical
Control Engineering Practice, Information Sciences, and Interna- Engineering, the University of Hong Kong, China. From March
tional Journal of Automation and Computing, and a Guest Edit- 2013 to March 2014, he was a visiting scholar in Department of
or for 13 special issues. Information Systems and Computing, Brunel University, UK.
His research interests include networked control systems,
He is currently a senior research fellow with School of Software
multi-agent systems, time-delay systems, smart grids, un-
and Electrical Engineering, Swinburne University of Techno-
manned surface vehicles, and neural networks.
logy, Australia. He is a Highly Cited Researcher according to
E-mail: [email protected] (Corresponding author)
Clarivate Analytics. He is a Senior Member of IEEE. He re-
ORCID iD: 0000-0002-7207-0716
ceived the 2020 IEEE Systems, Man, and Cybernetics Society
Andrew P. Sage Best Transactions Paper Award, and the 2018
Xian-Ming Zhang received the M. Sc. IET Control Theory and Applications Premium Award. He has
degree in applied mathematics and the Ph. published more than 80 papers in refereed international journals.
D. degree in control theory and control en- He is an Associate Editor for Neurocomputing and IET Control
gineering from Central South University, Theory & Applications. He is also a very active reviewer for
China in 1992 and 2006, respectively. In many international journals.
1992, he joined Central South University, His research interests include nonlinear stochastic control and
where he was an associate professor with filtering, as well as multi-agent systems and sensor networks.
School of Mathematics and Statistics. E-mail: [email protected]
From 2007 to 2014, he was a post-doctoral ORCID iD: 0000-0001-7402-6682