612 IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, VOL. 10, NO.

2, JUNE 2023

Network-Based Adaptive Multievent-Triggered

Fuzzy Dynamic Positioning Controller Design
for Unmanned Surface Vehicles Against
Denial-of-Service Attacks
Peng Cheng , Chengming Zhang, Wei Xie , Member, IEEE, Weidong Zhang , Senior Member, IEEE,
and Shuping He , Senior Member, IEEE

Abstract—This article focuses on the issue of adaptive low-cost, and highly maneuverable, promoting their applica-
multievent-triggered fuzzy dynamic positioning control of tions in many diverse fields, e.g., cargo transportation, energy
unmanned surface vehicles against denial-of-service at- exploration, military monitoring, and hydrological draw [1].
tacks within a finite-time interval. To improve the data
transmission efficiency, an innovative adaptive multievent- To meet the demands of reliability and high precision in ac-
triggered mechanism is designed to dynamically adjust the tual engineering, plentiful issues regarding the motion control
event-triggered thresholds. Especially, the asynchrony be- of USVs have been discussed, covering coordinated tracking,
tween the membership functions of the controller and that point stabilization, and path following [2], whereas dynamic
of the system is embedded into the adaptive multievent- positioning, focusing on managing the horizontal position and
triggered mechanism, which can help to reduce the con-
servatism of controller design attributed to mismatched heading of maritime vehicles, has received a lot of attention.
membership functions. By employing the linear matrix By introducing a quantized feedback sliding mode control
inequality technology, sufficient conditions are derived scheme, a new fault-tolerant controller for dynamic position-
to ensure that the switched closed-loop fuzzy dynamic ing of USVs was developed in [3]. In [4], a resource-aware
positioning system is finite-time bounded with an H∞ state and uncertainty estimation approach was presented for the
noise attenuation performance. To clarify the validity of
the presented positioning controller under the adaptive dynamic positioning system (DPS) suffering from external en-
multievent-triggered mechanism and denial-of-service at- vironmental disturbances, including currents, waves, and wind.
tacks, we provide and analyze the simulation results. Ma et al. [5] utilized a switched DPS to describe a mass-switched
Index Terms—Adaptive multievent-triggered mechanism, USV and developed a mode-dependent dynamic positioning
denial-of-service (DoS) attacks, dynamic positioning con- control strategy with an event-triggered (ET) communication
troller, finite-time bounded (FTB), unmanned surface vehi- scheme.
cles (USVs). In the actual process, with the increasing demand for USVs,
the dynamic characteristics of USVs has become increasingly
complicated in the presence of nonlinearity, which results in a
I. INTRODUCTION
challenge for system performance analysis. As a popular non-
S CLASSICAL unmanned autonomous robot dynamics,
A unmanned surface vehicles (USVs) have attracted a wide
range of interest. The distinguishing features of USVs are small,
linear processing method, the discussion of the Tagaki–Sugeno
(T-S) fuzzy model never ceases. The employment of linear
systems to represent the local behavior under each fuzzy rule
is a key feature of T-S fuzzy control [6]. After that, a variety
Manuscript received 11 April 2022; revised 6 July 2022; accepted 14 of linear control results is suitable for each linear model. A
August 2022. Date of publication 2 September 2022; date of current T-S fuzzy Markovian switching system was constructed in [7]
version 5 June 2023. This work was supported in part by the Key R&D to investigate the nonstationary control issue by using variable
Program of Guangdong under Grant 2020B1111010002, in part by the
Hainan Province Science and Technology Special Fund under Grant quantization density. With the T-S fuzzy technology, three dif-
ZDYF2021GXJS041, in part by the National Natural Science Foundation ferent performance analyses were fulfilled in [8] for incommen-
of China under Grant U2141234, and in part by the Shanghai Sailing surate fractional-order dynamics subject to nonlinearities and
Program under Grant 22YF1420400. Recommended by Associate Edi-
tor Ming Cao. (Corresponding author: Weidong Zhang.) time-varying delays. In [9], a robust sliding mode controller for
Peng Cheng, Chengming Zhang, Wei Xie, and Weidong Zhang T-S fuzzy systems was figured out by utilizing the quantized
are with the Department of Automation, Shanghai Jiao Tong Uni- state feedback. However, if network channels are considered
versity, Shanghai 200240, China (e-mail: [email protected];
[email protected]; [email protected]; [email protected]). between the concerned system and the controller, the member-
Shuping He is with the School of Electrical Engineering and ship functions of the concerned system and the controller are
Automation, Anhui University, Hefei 230601, China (e-mail: shup- asynchronous. Such a factor is not taken into account in [7],
[email protected]).
Digital Object Identifier 10.1109/TCNS.2022.3203892 [8], [9]. Thus, how to address the asynchrony between the

2325-5870 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
CHENG et al.: NETWORK-BASED ADAPTIVE MULTIEVENT-TRIGGERED FUZZY DYNAMIC POSITIONING CONTROLLER DESIGN 613

normalized membership function of the USV and that of the

controller deserves our further discussion.
It is challenging for USVs to formulate a better scheme
autonomously via only applying their sensors and actuators
when suffering from unfavorable sailing situations, such as
strong winds and large waves. To cater to the aforementioned
severe external environment, the control power can be diverted
from USVs to the land-based control station [10]. The network
Fig. 1. Body-fixed and Earth-fixed reference frames.
system, which serves as a communication channel between the
ship and the control station, allows for quick and easy data
exchange. Thus, a conventional networked control system is
output-based decentralized resilient ET mechanism was formu-
formed by the USV, the land-based control station, and the
lated in [21] to investigate the dissipativity issue for networked
communication network. Based on the aforementioned archi-
control systems under aperiodic DoS attacks. Note that the
tecture, many productive results have been emerged in [11],
ET mechanism under both periodic and aperiodic DoS attacks
[12], and [13]. If the plant state approaches equilibrium, a large
mentioned before are for continuous-time systems, whereas the
number of superfluous data packets will be transmitted over
ET mechanism under DoS attacks for discrete-time systems has
networks, leading to booming communication stress and net-
produced few results so far, which is still a key issue to be further
work resource waste. Thus, the time-triggered (TT) mechanism
studied.
in [11], [12], and [13], can evolve into a more popular method,
Motivated by the aforementioned discussion, the purpose of
called the ET mechanism [14], [15], [16]. The fundamental
this article is to investigate the issue of AMET fuzzy dynamic
idea of the ET mechanism is to judge whether the signal can
positioning control of USVs against DoS attacks within a finite-
be sent to the controller/filter under the given ET condition,
time interval. This seems to be a challenging assignment and two
to minimize communication resource waste while maintaining
essential difficulties are specified as follows: 1) how to design
system performance. A particle filter for the smart grid was
an ET mechanism so that state errors and mismatched mem-
devised by the ET mechanism in [14], which can reduce the
bership functions are taken into consideration, simultaneously
communication stress while maintaining a satisfactory estima-
and 2) how to formulate an AMET mechanism with adjustable
tion precision. An ET mechanism was developed in [15] to
thresholds to seek a tradeoff between the cost of communication
analyze the consistent characteristics of the multiagent systems,
resources and the desired system performance. In what follows,
whereas a locally passive nonlinear was provided to describe
we will seek out satisfactory solutions to the aforementioned
the coupling between adjacent agents. By designing an ET
difficulties, respectively. The main contributions of this article
mechanism-based asynchronous filter, the fault detection issue
are summarized as follows.
for Markov jump systems was worked out in [16]. Note that
1) The aperiodic DoS attacks are considered into the
the thresholds of the ET mechanisms in [14], [15], and [16] are
network-based dynamic positioning control for USVs,
set as a fixed value. When the concerned system dynamically
where the switched closed-loop fuzzy DPS is established
changes, the prescribed threshold cannot adjust itself to make a
to reject the undesired cyberattacks.
balance between the cost of communication resources and the
2) An innovative AMET mechanism by the considering
desired system performance, i.e., the loss of system performance
state behavior and asynchronous characteristics between
in exchange for a reduction of triggering instants. Thus, how
membership functions is formulated to make a balance
to formulate an adaptive multi-ET (AMET) mechanism with
between the cost of communication resources and the
changeable thresholds, especially embedding system states and
desired system performance.
membership functions simultaneously, calls for our further in-
3) In line with the multi-Lyapunov function approach and
vestigation.
iterative technology, sufficient conditions are acquired,
With the growing application of network technology, network
which can guarantee that the switched closed-loop fuzzy
security has sparked a lot of interest. So far, deception attacks and
DPS is finite-time bounded (FTB) with an H∞ noise
denial-of-service (DoS) attacks are the most familiar varieties
attenuation performance.
of cyberattacks. Deception attacks, such as fake data-injection
Notations: In this article, the shorthand “diag{·}” represents
attacks [17] and replay attacks [18], try to tamper with the
a block diagonal matrix. Symbol ∗ means the symmetric term
sent data, resulting in misleading data being received by the
in symmetric block matrices.  ·  refers to the Euclidean norm
controller. On the other hand, DoS attacks aim to disrupt data
of a vector or its induced matrix norm. N + represents a positive
communication between system devices. Then, the precise infor-
integer set. λmax {M } and λmax {M } denote the maximal and
mation will be unable to transmit its intended destination. There
minimal eigenvalue matrix M , respectively.
is no denying that cyberattacks can impact the ET information
over networks and many productive ET mechanism-based re-
sults under cyberattacks have been achieved. Under DoS attacks, II. NETWORK-BASED T-S FUZZY MODELING
a secure cooperative ET controller was devised in [19] for linear Fig. 1 presents the body-fixed reference frame {Ob } and the
multiagent systems. The upper bound of the prediction error Earth-fixed {O} reference frame. xb and yb signify the longitu-
associated with periodic DoS attacks was provided in [20], dinal axis and transverse axis, respectively. x and y signify the
by developing a predictor-based ET control architecture. An Earth-fixed reference diagrams. Note that the USV considered
Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
 614 IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, VOL. 10, NO. 2, JUNE 2023

in this article is not furnished with side thrusters. We can provide ω(t) = [ω1 ω2 ω3 ]T = [τω1 τω2 τω3 ]T
the motion equations of the USV as ⎡ ⎤
−m d11 m22
m11 x3 0 0 0 0
η̇ = R(ψ)v (1) ⎢
11
⎥
⎢ −m 11
m22 x3 −m d22
0 0 0 0⎥
M v̇ + C(v)v + Dv = τ + τ ω ⎢ 22 ⎥
(2) ⎢ m11 −m22 ⎥
⎢ 0 0⎥
A = ⎢ m33 x2 0 − m33 0
d 33

where M = diag{m11 , m22 , m33 } denotes the inertia matrix; ⎥

⎢ ⎥
D = diag{d11 , d22 , d33 } denotes the hydrodynamic damping ⎢ 1 0 0 0 x3 0 ⎥
⎢ ⎥
matrix; η = [x y ψ]T signifies the position vector of the USV ⎣ 0 1 0 −x3 0 0⎦
in Earth-fixed diagram; v = [u v r]T signifies the velocity vector 0 0 1 0 0 0
of the USV in body-fixed diagram, composed by the velocities ⎡ ⎤ ⎡ ⎤
1 1
u in surge and v in sway; and yaw rate r, τ = [τ1 0 τ3 ]T is the m11 0 m11 0 0
control vector, where τ1 and τ3 signify the surge force and the ⎢ 0 ⎥ ⎢ 1 ⎥
⎢ 0 ⎥ ⎢ 0 0 ⎥
yaw torque input, respectively; and τ ω = [τω1 τω2 τω3 ]T is the ⎢ ⎥ ⎢ m22 ⎥
⎢ 0 1 ⎥ ⎢ 1 ⎥
disturbance vector, where τω1 , τω2 , and τω3 are the disturbances B= ⎢ m33 ⎥ , D = ⎢ 0 0 m33 ⎥
⎢ ⎥ ⎢ 0 0 ⎥
of surge, sway, and yaw motions, respectively. The Coriolis ⎢ 0 0 ⎥ ⎢ 0 ⎥
⎢ ⎥ ⎢ ⎥
matrix is given as ⎣ 0 0 ⎦ ⎣ 0 0 0 ⎦
⎡ ⎤ 0 0 0 0 0
0 0 −m22 v
C(v) = ⎣ 0 0 m11 u ⎦ . .
m22 v −m11 u 0 Note that there exist x2 and x3 in A and B, which cause the
obtained USV system to not be linear. Here, we utilize T-S fuzzy
Besides, the rotation matrix is presented as technology to linearize the nonlinear system (6).
⎡ ⎤ Plant Rule p : IF σ1 (k) is ςp1 , σ2 (k) is ςp2 , . . ., and σp (k) is
cos ψ − sin ψ 0
R(ψ) = ⎣ sin ψ cos ψ 0 ⎦ . ςpm , THEN
0 0 1
x(k + 1) = Ap x(k) + Bp u(k) + Dp ω(k)
(7)
Combined with the motion equations (1) and (2), the dynamics z(k) = Cp x(k)
of the USV is formulated as
⎧ where p ∈ S1 = {1, 2, . . . , s}, s denote the number of IF–
⎪
⎪ m11 u̇ − m22 vr + d11 u = τ1 + τω1 THEN rules, ςpq , q ∈ S2 = {1, 2, . . . , m} signifies the fuzzy
⎪
⎪ m22 v̇ + m11 ur + d22 v = τω2
⎪
⎪ sets, and σ(k) = [σ1(k) σ2(k) · · · σp(k) ] represents the premise
⎨ m ṙ + (m − m ) uv + d r = τ + τ
33 22 11 33 3 ω3 variable.
(3)
⎪
⎪ ẋ = u cos ψ − v sin ψ Thus, we can acquire the T-S fuzzy DPS as
⎪
⎪ ẏ = u sin ψ + v cos ψ
⎪
⎪
⎩
ψ̇ = r. x(k + 1) = Ah x(k) + Bh u(k) + Dh ω(k)
(8)
z(k) = Ch x(k)
Inspired by [22], we adopt the following coordinate transfor-
mation as: where
⎧ s s
⎨ g1 = x cos ψ + y sin ψ Ah = p=1 hp (σ(k))Ap , Bh = p=1 hp (σ(k))Bp
s s
g2 = −x sin ψ + y cos ψ (4) Ch = p=1 hp (σ(k))C p , Dh = p=1 hp (σ(k))Dp
⎩ p
ςpq (σq (k))
g3 = ψ and hp (σ(k)) = s q=1p is the normalized
p=1 q=1 ςpq (σq (k))
where g1 , g2 , and g3 represent the mapping variables in the membership function subject to rule p, where ςpq (σq (k)) is
body-fixed coordinate. Taking (4) into (3), it has the grade of membership of σq (k). Besides, for ∀p ∈ S1 , it has
⎧ hp (σ(k)) ≥ 0 with sp=1 hp (σq (k)) = 1.
⎪
⎪ u̇ = m111 (m22 vr − d11 u + τ1 + τω1 )
⎪
⎪
⎪
⎪ 1
⎪ v̇ = m22 (−m11 ur − d22 v + τω2 )
⎪ III. CO-DESIGN OF AMET MECHANISM AND FUZZY
⎨
ṙ = m133 ((m11 − m22 ) uv − d33 r + τ3 + τω3 ) (5) CONTROL LAW UNDER DOS ATTACKS
⎪
⎪
⎪
⎪ ġ 1 = u + g 2 r A. AMET Mechanism Design
⎪
⎪ ġ2 = v − g1 r
⎪
⎪
⎩ ġ = r. The networked diagram for the T-S fuzzy DPS of a USV is
3
exhibited in Fig. 2. The communication between the USV and
From system (5), we obtain the remote-control station is realized by networks. Due to the
ẋ(t) = Ax(t) + Bu(t) + Dω(t) (6) limitation of network bandwidth, data transmission capability is
emphasized. In this work, an innovative AMET mechanism is
where established to reduce the waste of communication resources.
x(t) = [x1 x2 x3 x4 x5 x6 ]T = [u v r g1 g2 g3 ]T u(t) = Only at the released moment kt , t ∈ N + , the present data
[u1 u2 ]T = [τ1 τ3 ]T x(kt ) can be transmitted to the following controller. The AMET

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
CHENG et al.: NETWORK-BASED ADAPTIVE MULTIEVENT-TRIGGERED FUZZY DYNAMIC POSITIONING CONTROLLER DESIGN 615

Fig. 3. Date transmission scheme subject to AMET mechanism and

DoS attacks.

Controller Rule q: IF σ1 (kt ) is ςq1 , σ2 (kt ) is ςq2 , . . ., and

σp (kt ) is ςqp , THEN

u(k) = Kq x (kt ) , k ∈ [kt , kt+1 ) (12)

where Kq is the controller gain to be devised.

Then, the overall T-S fuzzy controller is given as
Fig. 2. Networked diagram for the T-S fuzzy DPS of a USV.

s
u(k) = hq (σ (kt )) Kq x (kt ) , k ∈ [kt , kt+1 ) . (13)
q=1
mechanism can be specifically described as
Remark 3: The AMET mechanism is carried out in the
kt+1 controller design, which suggests that only the data at triggering
   instant kt can be activated. For this reason, there may exist an
= min inf k > kt | eT T
x (k)Ψex (k) ≥ βx (k)x (k)Ψx(k)
asynchrony between the membership functions of the controller
inf {k > kt | |eh (k)| ≥ βh (k)hq (σ(k))}} (9) and that of the system, that is, hq (σ(kt )) = hq (σ(k)). Thus, in
the overall T-S fuzzy controller (13), hq (σ(kt )) is used to replace
where ex (k) = x(kt ) − x(k), eh (k) = hq (σ(kt )) − hq (σ(k)),
hq (σ(k)) to figure out the conservative problem of controller
q ∈ S2 , βx (k) > 0, βh (k) > 0, and 0 < Ψ < 1 means a weight
design resulting from mismatched membership functions.
matrix.
The adaptive laws are shown as
   C. DoS Over Networks
1 1 1
Δβx (k)  −βx0 eT x (k)Ψe x (k)− (Δβ x (k)) 2
Due to the openness characteristic of networks, different
βx (k) βx (k) 2
(10) malicious cyberattacks may appear during the information trans-
   mission, which renders the sent data unsound. In this work, the
1 1 1 concerned DoS attacks is aperiodic, and the active time of that
Δβh (k)  −βh0 |eh (k)|− (Δβh (k))2
βh (k) βh (k) 2 is represented by [Dnon , Dnoff ), n ∈ N + . If a DoS attacks, the
(11) communication channel will be broken off.
In addition, an illustration of the data transmission scheme
where Δβx (k) = βx (k + 1) − βx (k), Δβh (k) = βh (k + 1) − subject to AMET mechanism and DoS attacks is exhibited
βh (k), and βx0 and βh0 are predefined values. in Fig. 3. Obviously, events can be performed as instants
Remark 1: In the AMET mechanism (9), the triggering k0 , k1 , k2 , and k3 . However, due to the intervention of DoS
condition relating to membership functions may increase a attacks, only two instants can be activated. Accordingly, just
few triggering times relative to the ET mechanism, because x(k1 ) and x(k3 ) can be utilized to the controller design.
some instants may meet the aforementioned triggering condition For that matter, we introduce variable α(k) to portray the role
without consideration of the triggering condition based on state of DoS attacks. Then, the overall T-S fuzzy controller (13) can
errors. However, communication resource waste can still be be rewritten as
reduced when the two triggering conditions are not satisfied
simultaneously. 
s

Remark 2: For adaptive laws (10) and (11), the thresholds u(k) = α(k) hq (σ (kt )) Kq x (kt ) (14)
q=1
βx (k) and βh (k) can regulate themselves. If ex (k) approaches
0, Δβx (k) = 0. Therefore, βx (k) tends to be a fixed scalar.
0, k ∈ [kt , kt+1 ) ∩ [Dnon , Dnoff );
Besides, the same law also applies to βh (k). where α(k) =
1, others.
Let (a, k) be the times of DoS OFF/ON transitions and
B. Fuzzy Control Law Design
(a, k) be the total duration in the interval [a, k). To de-
Based on the AMET mechanism, we devise the following scribe the DoS attack, referring to [23], we introduce the
controller. following.

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
 616 IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, VOL. 10, NO. 2, JUNE 2023

Assumption 1 (DoS frequency): There exist scalars ρ ≥ 0 to such peculiarity, the switched closed-loop fuzzy DPS (17) is
and φ > 0 satisfying formulated, which is managed by the switching signal α(k).
k−a Remark 6: Note that the stability analyses in [5], [12], and
(a, k) ≤ ρ + . (15) [25] focused on the asymptotical stability over an infinite-time
φ
interval. However, for many actual engineering, a system with
Assumption 2 (DoS duration): There exist scalars λ ≥ 0 and asymptotic stability over an infinite-time interval cannot imply
Q > 1 satisfying that as it has superior transient peculiarities, such as an informa-
k−a tion transmission system, a flight system, and chemical reaction
(a, k) ≤ λ + . (16) system [26], [27], [28], [29]. For the issue of the network-based
Q
positioning control of a USV, the transient behavior of the USV
Taking the fuzzy controller (14) into the T-S fuzzy DPS (8), is more worthy of attention. When the position changes by the
we can derive the switched closed-loop fuzzy DPS as external disturbances, the USV needs to return to its original
x(k + 1) = Ah,h̄ x(k) + Bh,h̄ e(k) + Dh,h̄ ω(k) position quickly. Thus, it is urgent to judge whether the dynamics
(17) can maintain within the given boundedness over a finite-time
z(k) = Ch,h̄ x(k)
interval or not.
where

s 
s
Ah,h̄ = hq (σ(kt )) hp (σ(k))[Ap + α(k)Bp Kq ] IV. MAIN RESULTS
q=1 p=1
In this section, we are in the position to establish some suffi-

s 
s cient conditions to guarantee that the switched closed-loop fuzzy
Bh,h̄ = hq (σ(kt )) hp (σ(k))α(k)Bp Kq Ch,h̄ DPS (17) is stochastically stable with an H∞ noise attenuation
q=1 p=1 performance.

s Theorem 1: For given scalars ρ ≥ 0, λ ≥ 0, φ > 0, Q > 1,
= hq (σ(kt )) c1 > 0, ε1 > 0, ε2 > 0, N > 0, κ > 0, d > 0, and 2 > 0, the
q=1 switched closed-loop fuzzy DPS (17) is FTB with an H∞ noise

s 
s 
s attenuation level γ̃, if there exist ϑ > 0, γ > 0,  ≥ 1, 0 < π1 <
hp (σ(k))Cp Dh,h̄ = hq (σ(kt )) hp (σ(k))Dp . 1, π2 ≥ 0, π̄3 > 0, βx0 > 0, βh0 > 0, and R > 0 and symmetric
p=1 q=1 p=1 matrices P1 > 0 and P2 > 0 for ∀p, q ∈ S1 , such that
Let hp and h̄q be the simplified expression of hp (σ(k)) and
Γpp < 0 (20)
hq (σ(kt )), respectively. Before proceeding further, we provide
some definitions for the following proof. Γpq + Γqp < 0, p < q (21)
Definition 1 ([23]): Given a time interval [0, N ], scalars c1 >
Ωp < 0 (22)
0 and c2 > 0, with c2 > c1 , and a weight matrix R > 0, the
switched closed-loop fuzzy DPS (17) with ω(k) < 2 is said to P1 ≤ P2 (23)
be FTB with (c1 , c2 , N, R, 2 ), if
P2 ≤ P1 (24)
xT (0)Rx(0) ≤ c1 ⇒ xT (k)Rx(k) < c2 . (18)
εs R < Ps < R, s = 1, 2 (25)
Definition 2 ([23]): The switched closed-loop fuzzy system
ln 
(17) is said to be FTB with an H∞ noise attenuation level γ if 1) ln π̄3 + <0 (26)
the switched closed-loop fuzzy system (17) is FTB with ω(k) < φ
2
, and 2) under the zero-initial condition, the controlled output ln π̄2 − ln π̄3
0< ≤Q−1 (27)
z(k) holds − ln π̄1 + ln π̄3
∞ ∞
⎡ √  ⎤
  −εs c2 π 2 π
3 
c1 π 2 (1 − π3 )N −1 2
T 2 T
z (k)z(k) ≤ γ ω (k)ω(k). (19) ⎣ ∗ −π̄1λ
0 ⎦<0 (28)
k=0 k=0
∗ ∗ −π̄1λ (1 − π3 )
Remark 4: For portraying DoS attacks, Assumptions 1 and
2 are proposed to provide energy constraints related to the fre- where
quency and duration of DoS attacks. Limited energy generated
by DoS attacks not only makes controller design and synthesis ⎡ ⎤
meaningful, but also has a practical motivation. On the other Γ1 ϑ̄ΓT T
2 ϑ̄Γ3
hand, due to the mitigation ability to spread techniques and Γpq = ⎣ ∗ −P1−1 0 ⎦
high-pass filtering, DoS attacks usually occur intermittently ∗ ∗ −I
or randomly [24]. In this article, DoS attacks are performed 1 − κβh0
intermittently by the adversary. Γ1 = diag{−ϑ̄π̄1 P1 + Ψ, −βxo Ψ, −ϑ̄γ 2 I + R}
d
Remark 5: In the cases of α(k) = 1 and α(k) = 0, two
subsystems are implemented from the T-S fuzzy DPS (8). Due Γ2 = [Ap + Bp Kq Bp Kq Dp ], Γ3 = [Cp 0 0]

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
CHENG et al.: NETWORK-BASED ADAPTIVE MULTIEVENT-TRIGGERED FUZZY DYNAMIC POSITIONING CONTROLLER DESIGN 617

⎡ ⎤  T  
Ω1 ΩT2 ΩT3 
s 
s 
s 
s
Ωp = ⎣ ∗ −P2−1 0 ⎦ + ϑ̄ hp hq Γ3p ϑ̄ hp hq Γ3p ξ(k)
∗ ∗ −I p=1 q=1 p=1 q=1

Ω1 = diag{−π̄2 P2 + Ψ, −βxo Ψ, −γ 2 I} 
s 
s
+ hp hq xT (k)[ϑ̄(π1 − 1)P1 + Ψ]x(k)
1
2 = π̄2λ ρ , π
Ω2 = [Ap 0 Dp ], Ω3 = Γ3 , π 3 = π̄3  φ p=1 q=1
  
c1 = c1 + 1, π̄1 = 1 − π1 , π̄2 = 1 + π2
ϑ̄ = 1 + ϑ,  1−κβh0
− βx0 eT (k)Ψe(k)− ϑ̄γ 2 I− R ω T (k)ω(k) .
d
Proof: In the case that no DoS attacks, we give the Lyapunov (33)
function as
With the help of conditions (20) and (21), it renders
Vα1 (k) = VP1 (k) + Vβx (k) + Vβh (k) (29)

s 
s
T 1 2 hp hq Γpq
where VP1 (k) = x (k)P1 x(k), Vβx (k) = 2 βx (k),
Vβh (k) =
1 s 2 p=1 q=1
2 β
q=1 h (k).
Following the trajectory of the switched closed-loop fuzzy 
s 
s−1 
s
DPS (17), one has = h2p Γpp + hp hq (Γpq + Γqp ) < 0. (34)
p=1 p=1 q>p
ΔVP1 (k) = VP1 (k + 1) − VP1 (k)
⎡ T  s s ⎤ In line with inequalities (33) and (34) and Schur complement

s 
s  lemma, it yields
= ξ T (k) ⎣ hp h̄q Γ2 P1 hp h̄q Γ2 ⎦ ξ(k)
p=1 q=1 p=1 q=1 ΔVα1 (k) ≤ −π1 Vα1 (k) − z T (k)z(k) + γ 2 ω T (k)ω(k). (35)

− xT (k)P1 x(k) (30) Note that inequality (35) is achieved based on the assumption
that no DoS attack appears. The following case will discuss the
where ξ(k) = [xT (k) eT (k) ω T (k)]T . switched closed-loop fuzzy DPS (17) subject to DoS attacks.
Moreover, combining the AMET mechanism (9) and the Similarly, we choose
adaptive law (10) renders
Vα0 (k)  VP2 (k) + Vβx (k) (36)
ΔVβx (k) = Vβx (k + 1) − Vβx (k)
where VP2 (k) = xT (k)P2 x(k).
1 2 1 Following the switched closed-loop fuzzy DPS (17), one has
= β (k + 1) − βx2 (k)
2 x 2
ΔVP2 (k)
1 ⎡
= βx (k)Δβx (k) + (Δβx (k))2 T  s s ⎤
2 
s 
s 
  = ξ T (k) ⎣ hp h̄q Ω2 P2 hp h̄q Ω2 ⎦ ξ(k)
1
= − β x 0 eT x (k)Ψex (k) p=1 q=1 p=1 q=1
βx (k)
− xT (k)P2 x(k)
≤ xT (k)Ψx(k) − βx 0 eT
x (k)Ψex (k). (31)
⎡ T  s ⎤
We set κhq <| eh |< ϑhq , κ > 0, ϑ > 0 to represent the 
s 
asynchronization between the membership functions of the con- = ξ T (k) ⎣ hp Ω2 P2 hp Ω2 ⎦ ξ(k)
p=1 p=1
troller and that of the system, where κ is a sufficiently small
constant. Consider ω T (k)Rω(k) > d, one has − xT (k)P2 x(k). (37)
ΔVβh (k) = Vβh (k + 1) − Vβh (k) In the light of inequality (37), it yields

s
ΔVα0 (k) − π2 Vα0 (k) + z T (k)z(k) − γ 2 ω T (k)ω(k)
≤ (1 − κβh0 )hq (σ(k)) ⎡ T  s 
q=1 s 

s 
s   = ξ (k) ⎣
T
hp Ω2i P2 hp Ω2i
1 − κβh0
≤ hp hq ω T (k)Rω(k). (32) p=1 p=1
d
p=1 q=1  T  ⎤

s 
s
Besides, from the switched closed-loop fuzzy DPS (17), we + hp Ω3i hp Ω3i ⎦ ξ(k)
can further obtain p=1 p=1
T 2 T
ΔVα1 (k) + π1 Vα1 (k) + z (k)z(k) − γ ω (k)ω(k) 
s

⎡ T  s s  + hp xT (k) (− (1 + π2 ) P2 + Ψ) x(k)

s s  p=1
≤ ξ T (k) ⎣ ϑ̄ hp hq Γ2pq P1 ϑ̄ hp hq Γ2pq 
p=1 q=1 p=1 q=1 −βx0 eT (k)Ψe(k) − γ 2 ω T (k)ω(k) . (38)

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
 618 IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, VOL. 10, NO. 2, JUNE 2023


From condition (22), it can be verified that −z T (k − Dnon − 1) z (k − Dnon − 1)

s k−D on −1  2 T
hp Ωp < 0. (39) + π̄2 n γ ω (k − Dnon ) ω (k − Dnon )
 k−D on −2
p=1
−z T (k − Dnon ) z (k − Dnon ) + π̄2 n
It yields from inequalities (38) and (39) that 
× γ 2 ω T (k − Dnon + 1) ω (k − Dnon + 1)
ΔVα0 (k) ≤ π2 Vα0 (k) − z T (k)z(k) + γ 2 ω T (k)ω(k). (40) 
−z T (k − Dnon + 1) z (k − Dnon + 1)
Let V (k) = Vα1 (k) when there is no DoS attacks; otherwise,
V (k) = Vα0 (k). Then, we set out to analyze the relationships + · · · + γ 2 ω T (k − 1)ω(k − 1)
between V (k) and V (0) for any k ≥ 0. First, inequalities (35) − z T (k − 1)z(k − 1)
and (40) are reconstructed as
..
Vα1 (k + 1) ≤ π̄1 Vα1 (k) − z T (k)z(k) + γ 2 ω T (k)ω(k) (41) .

Vα0 (k + 1) ≤ π̄2 Vα0 (k) − z T (k)z(k) + γ 2 ω T (k)ω(k) (42) k−(0,k) (0,k) (0,k)

k−1
≤ π̄1 π̄2  V (0) + π̄1k−1−a
where π̄1 = 1 − π1 and π̄2 = 1 + π2 . a=0
Conditions (23) and (24) can derive −(a,k−1) (a,k−1)
× π̄1 π̄2 (a,k−1)
Vα1 (k) ≤ Vα0 (k) (43)  
× γ 2 ω T (a)ω(a) − z T (a)z(a) . (45)
Vα0 (k) ≤ Vα1 (k). (44)
/ [Dnon , Dnoff ), the identical result as
On the other hand, for k ∈
In the presence of DoS attacks, inequalities (41)–(44) can inequality (45) can be derived, with the specifics omitted here.
deduce that Then, from inequality (45), we get
V (k) ≤ π̄2 Vα0 (k − 1) + γ 2 ω T (k − 1)ω(k − 1) 
k−1
k−(0,k) (0,k) (0,k)
T V (k) ≤ π̄1 π̄2  V (0) + π̄1k−1−a
− z (k − 1)z(k − 1)
a=0
≤ π̄22 Vα0 (k − 2) + π̄2 γ 2 ω T (k − 2)ω(k − 2) −(a,k−1) (a,k−1) (a,k−1)
× π̄1 π̄2 
T 2
− π̄2 z (k − 2)z(k − 2) + γ ω (k − 1) T  
× γ 2 ω T (a)ω(a) − z T (a)z(a) . (46)
T
× ω(k − 1) − z (k − 1)z(k − 1)
Besides, recalling Assumption 2, the following relationship
.. is tenable:
.
on on
−1  (k − (0, k)) ln π̄1 + (0, k) ln π̄2
Vα0 (Dnon ) + π̄2 γ 2 ω T (k − Dnon )
k−Dn k−Dn
≤ π̄2
  
 1 1
×ω (k − Dnon ) − z T (k − Dnon ) z (k − Dnon ) ≤ λ (ln π̄2 − ln π̄1 ) + k 1 − ln π̄1 + ln π̄2 . (47)
Q Q
k−D on −2  2 T
+ π̄2 n γ ω (k − Dnon + 1)
Condition (27) implies that
× ω (k − Dnon + 1) − z T (k − Dnon + 1)  
1 1
1− ln π̄1 + ln π̄2 ≤ ln π̄3 . (48)
×z (k − Dnon + 1)] + · · · + γ 2 ω T (k − 1) Q Q
× ω(k − 1) − z T (k − 1)z(k − 1) Keeping inequalities (47) and (48) in mind, one has
k−D on k−D on −1  2 T
≤ π̄2 n Vα1 (Dnon ) + π̄2 n γ ω (k − Dnon ) (k − (0, k)) ln π̄1 + (0, k) ln π̄2

×ω (k − Dnon ) − z T (k − Dnon ) z (k − Dnon ) ≤ k ln π̄3 + λ (ln π̄2 − ln π̄1 ) . (49)
k−D on −2  2 T
+ π̄2 n γ ω (k − Dnon + 1) Combining Assumption 2 and (27), it can be concluded that

× ω (k − Dnon + 1) − z T (k − Dnon + 1) (k̂ − a − (a, k̂)) ln π̄1 + (a, k̂) ln π̄2

   
×z (k − Dnon + 1)] + · · · + γ 2 ω T (k − 1) k̂ − a k̂ − a
≤ ln π̄1 k̂ − a − λ − + ln π̄2 λ +
× ω(k − 1) − z T (k − 1)z(k − 1) Q Q
on on
−1 
≤ π̄2
k−Dn
π̄1 Vα1 (Dnon − 1) + π̄2
k−Dn
 γ2 ≤ (k̂ − a) ln π̄3 + λ (ln π̄2 − ln π̄1 ) (50)

× ω T (k − Dnon − 1) ω (k − Dnon − 1) where k̂ = k − 1.

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
CHENG et al.: NETWORK-BASED ADAPTIVE MULTIEVENT-TRIGGERED FUZZY DYNAMIC POSITIONING CONTROLLER DESIGN 619

For k̄ ∈ [0, N − 1], inequality (50) and Assumption 1 can Due to π̄1 < 1, π̄2 ≥ 1,  ≥ 1, (a, k̂) ≥ 0, and (a, k̂) ≥
figure out 0, one has


k̂ 
k̂

e(k̂−a−(a,k̂)) ln π̄1 +(a,k̂) ln π̄2 +(a,k̂) ln ω T (a)ω(a) e(k̂−a−(a,k̂)) ln π̄1 +(a,k̂) ln π̄2 +(a,k̂) ln z T (a)z(a)
a=0
a=0


k̂ 
k̂

≤ e e
ln 
[λ(ln π̄2 −ln π̄1 )+ρln ] (ln π̄3 + φ )(k̂−a)
ω T (a)ω(a) ≥ e(k̂−a) ln π̄1 z T (a)z(a)
a=0
a=0

π̄ λ ρ (1 − π̄3  φ )N −1
1

k̂
(k̂−a) T
≤ 2λ 1
2
. (51) = π̄1 z (a)z(a). (56)
π̄1 1 − π̄3  φ a=0

From inequalities (51) and (56), it yields

Taking inequalities (49) and (51) into inequality (46) yields

k̂
(k̂−a) T
V (k) π̄1 z (a)z(a)
1 1 a=0
π̄2λ ρ π̄3  φ π̄ λ ρ (1 − π̄3  φ )N −1
≤ V (0) + 2 λ 2
π̄2λ ρ   (k̂−a)
λ 1 . (52) k̂
π̄1 π̄1 1 − π̄3  φ ≤ γ2 π̄ 3 
1
φ ω T (a)ω(a). (57)
π̄1λ a=0
1 1
Define λmin (P̄s ) = λmin (R− 2 Ps R− 2 ) and λmax (P̄s ) =
1 1
λmax (R− 2 Ps R− 2 ). Following inequality (52) renders Performing summation for k̂ on both sides of inequality (57)
from 0 to N − 1 and exchanging the summation order, one has
1 π̄2λ ρ 1   
N −1 N
 −1
xT (k)Rx(k) < λ
(π̄3  φ )k λmax (P̄s )c1 + 1 (k̂−a) T
λmin (P̄s ) π̄1 π̄1 z (a)z(a)
1
 a=0 k̂=a
π̄2λ ρ (1 − π̄3  φ )N −1 2
π̄2λ ρ    (k̂−a)
+ (53) N −1 N −1
π̄1λ 1 − π̄3  φ
1
≤ γ2
1
π̄3  φ ω T (a)ω(a). (58)
λ
π̄1 a=0
k̂=a
which indicates xT (k)Rx(k) < c2 on the basis of conditions
By means of condition (26), we have
(25) and (28). Therefore, the switched closed-loop fuzzy DPS
(17) is FTB. 
N −1 
N −1

Next, the H∞ noise attenuation performance under the zero z T (k)z(k) ≤ γ̃ 2 ω T (k)ω(k) (59)
boundary condition will be discussed. It can be deduced from k=0 k=0
1
inequality (46) that π̄2λ ρ 1−(π̄3 φ )N
where γ̃ 2 = γ 2 π̄1λ 1 .
1−π̄3 φ

k−1
k−1−a−(a,k−1) (a,k−1) (a,k−1) The relationships (53) and (59) with Definition 2 can ensure
V (k) ≤ π̄1 π̄2  that the switched closed-loop fuzzy DPS (17) is FTB with an
a=0
  H∞ noise attenuation level γ̃. With this, the proof is finished.
× γ 2 ω T (a)ω(a) − z T (a)z(a) . (54) Theorem 1 establishes some rigorous restraints to guarantee
the switched closed-loop fuzzy DPS (17) is FTB with an H∞
Based on V (k) ≥ 0, inequality (54) means noise attenuation level γ̃; however, some nonlinearities existing
in Theorem 1 will be obstacles to the control strategy. Therefore,

k̂
k̂−a−(a,k̂) (a,k̂) (a,k̂) T
the following theorem is present to linearize the restraints in
π̄1 π̄2  z (a)z(a) Theorem 1.
a=0 Theorem 2: For given scalars ρ ≥ 0, λ ≥ 0, φ > 0, Q > 1,

k̂ c1 > 0, ε1 > 0, ε2 > 0, N > 0, κ > 0, d > 0, and 2 > 0, if,
= e(k̂−a−(a,k̂)) ln π̄1 +(a,k̂) ln π̄2 +(a,k̂) ln z T (a)z(a) for ∀p, q ∈ S1 , there exist γ > 0,  ≥ 1, 0 < π1 < 1, π2 ≥ 0,
a=0 π̄3 > 0, βx0 > 0, βh0 > 0, and R > 0, symmetric matrices
X1 > 0 and X2 > 0, and matrices K̂q , Ψ  1 , and Ψ
 2 , such that

k̂
k̂−a−(a,k̂) (a,k̂) (a,k̂) T linear matrix inequalities (LMIs) (26)–(28) and the following
≤ γ2 π̄1 π̄2  ω (a)ω(a)
LMIs hold:
a=0
Πpp < 0 (60)

k̂
2 (k̂−a−(a,k̂)) ln π̄1 +(a,k̂) ln π̄2 +(a,k̂) ln T
=γ e ω (a)ω(a). Πpq + Πqp < 0, p < q (61)
a=0
(55) Ξp < 0 (62)

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
 620 IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, VOL. 10, NO. 2, JUNE 2023

 
−X2 X2 Plant rule 1: IF x2 (k) is 0.5 and x3 (k) is 0.2, THEN
<0 (63)
∗ −X1
  x(k + 1) = A1 x(k) + B1 u(k) + D1 ω(k)
−X1 X1 z(k) = C1 x(k).
<0 (64)
∗ −X2
Plant rule 2: IF x2 (k) is 0.5 and x3 (k) is −0.2, THEN
R < Xs < ε−1
s R, s = 1, 2 (65)
x(k + 1) = A2 x(k) + B2 u(k) + D2 ω(k)
where   z(k) = C2 x(k).
Π1 Π2
Πpq = Plant rule 3: IF x2 (k) is −0.5 and x3 (k) is 0.2, THEN
∗ Π3
⎡ ⎤
−2π̄1 X1 + Ψ 1 0 0 x(k + 1) = A3 x(k) + B3 u(k) + D3 ω(k)
Π1 = ⎣ ∗ −βx0 Ψ1 0 ⎦
z(k) = C3 x(k).
2 1−κβh0
∗ ∗ −2γ I + R
⎡ T T T T
⎤ d
Plant rule 4: IF x2 (k) is −0.5 and x3 (k) is −0.2, THEN
2X1 Ap + 2K̂q Bp 2X1 Cp
Π2 = ⎣ 2K̂qT BpT 0 ⎦ x(k + 1) = A4 x(k) + B4 u(k) + D4 ω(k)
T
2Dp 0 z(k) = C4 x(k)
   
−X1 0 Ξ1 Ξ2
Π3 = , Ξp = with
∗ −I ∗ Ξ3
⎡ ⎤
−π̄2 X2 + Ψ 2 0 0 ⎡ ⎤
Ξ1 = ⎣ 2 0 ⎦ 0.4440 0.0037 0 0 0 0
∗ −βx0 Ψ ⎢ −0.0389 0.3564 0
⎢ 0 0 0⎥ ⎥
∗ ∗ −γ 2 I ⎢ 0.0564 0.0002 0.6320
⎡ ⎤ ⎢ 0 0 0⎥ ⎥
X2 AT T
X2 C p   A1 = ⎢ 0.0684 0.0013 0
p
−X2 0 ⎢ 0.9996 0.0300 0 ⎥⎥
Ξ2 = ⎣ 0 0 ⎦ , Ξ3 = ⎣ −0.0039 0.0624 0 −0.0300 0.9996 0 ⎦
∗ −I
DpT 0 0.0035 0 0.0802 0 0 1
then the switched closed-loop fuzzy DPS (17) is FTB with ⎡ ⎤
π̄2λ ρ 1−(π̄3
1
φ )N
0.4440 −0.0037 0 0 0 0
H∞ noise attenuation performance γ̃ = π̄1λ 1 γ. ⎢ 0.0389 0.3564 0 0 0 0⎥
1−π̄3 φ ⎢ ⎥
⎢ −0.0564 0.0002 0.6320 0 0 0⎥
Besides, the controller gain and the weight matrix of AMET A2 = ⎢ ⎥
⎢ 0.0684 −0.0013 0 0.9996 −0.0300 0 ⎥
mechanism are figured out by ⎢ ⎥
⎣ 0.0039 0.0624 0 0.0300 0.9996 0 ⎦
 1 X −1 = X −1 Ψ
Kq = K̂q X1−1 , Ψ = X1−1 Ψ  2 X −1 . (66) −0.0035 0 0.0802 0 0 1
1 2 2
⎡ ⎤
Proof: By definining X1 = P1−1 and X2 = P2−1 and premul- 0.4440 0.0037 0 0 0 0
tiplying and postmultiplying both sides of inequalities (20) and ⎢ −0.0389 0.3564 0 0 0 0⎥
⎢ ⎥
(21) with diag{X1 , X1 , I, I, I}, and both sides of inequality ⎢ −0.0564 −0.0002 0.6320 0 0 0⎥
A3 = ⎢ ⎥
(22) with diag{X2 , X2 , I, I, I} and their transposes, respec- ⎢ 0.0684 0.0013 0 0.9996 0.0300 0 ⎥
⎢ ⎥
tively, conditions (60)–(62) can then be guaranteed. ⎣ −0.0039 0.0624 0 −0.0300 0.9996 0 ⎦
On the other hand, inequalities (21)–(24) are tenable by −0.0035 0 0.0802 0 0 1
inequalities (63)–(65). This ends the proof.  ⎡ ⎤
0.4440 −0.0037 0 0 0 0
⎢ 0.0389 0.3564 0 0 0 0⎥
V. SIMULATION RESULTS ⎢ ⎥
⎢ 0.0564 −0.0002 0.6320 0 0 0⎥
A4 = ⎢ ⎥
In this part, the availability of the proposed dynamic po- ⎢ 0.0684 −0.0013 0 0.9996 −0.0300 0 ⎥
⎢ ⎥
sitioning controller for the switched closed-loop fuzzy DPS ⎣ 0.0039 0.0624 0 0.0300 0.9996 0 ⎦
(17) will be represented. The specific parameters of the 0.0035 0 0.0802 0 0 1
USV are m11 = 5.2 kg, m22 = 1.6 kg, m33 = 1.7 kg · m2 , ⎡ ⎤ ⎡ ⎤
d11 = 42.8 kg/s, d22 = 16.5 kg/s, and d33 = 7.8 kg · m2 /s. 0.0132 0 0.1923 0 0
From the actual operation situations of the USV, we assume sway ⎢ −0.0005 0 ⎥ ⎢ 0 0.6250 0 ⎥
⎢ ⎥ ⎢ ⎥
speed x2 (k) ∈ [−0.5, 0.5] and the yaw rate x3 (k) ∈ [−0.2, 0.2]. ⎢ 0.0007 0.0472 ⎥ ⎢ 0 0.5882 ⎥
Bp = ⎢ ⎥ Dp = ⎢ 0 ⎥,
By utilizing the fuzzy technology proposed in [30] to determine ⎢ 0.0007 0 ⎥ ⎢ 0 0 0 ⎥
⎢ ⎥ ⎢ ⎥
maximum and minimum values of x2 (k) and x3 (k), we obtain ⎣ −0.0000 0 ⎦ ⎣ 0 0 0 ⎦
the fuzzy sets as 0.0000 0.0025 0 0 0
M1 (x2 (k)) = 0.5 + x2 (k), M2 (x2 (k)) = 0.5 − x2 (k) p = 1, 2, 3, 4.
Besides, we give ρ = 5, λ = 5, φ = 6, and Q = 3 to build
N1 (x3 (k)) = 0.5 + 2.5x3 (k), N2 (x3 (k)) = 0.5 − 2.5x3 (k).
DoS attacks. Conditions (27) and (28) can be ensured by setting
Set sampling period Th = 0.05 s, then the pth rule of the plant  = 1, π1 = 0.01, π2 = 0.08, and π̄3 = 0.02. Other parameters
is described as as follows. are provided as ϑ = 1, κ = 0.02, d = 0.2, = 2.9, c1 = 402,

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
CHENG et al.: NETWORK-BASED ADAPTIVE MULTIEVENT-TRIGGERED FUZZY DYNAMIC POSITIONING CONTROLLER DESIGN 621

Fig. 5. State trajectories (position and velocity) in the closed-loop case

Fig. 4. State trajectories (position and velocity) in the open-loop case. with DoS attack.

N = 30, and R = I. Via figuring out LMIs (26)–(28) and (60)–

(65) by MATLAB LMI toolbox, one has the H∞ noise attenu-
ation performance index γ = 18.1563, c2 = 1.7833 × 103 , and
the controller parameter matrices as

⎡ ⎤T
−54.6846 2.3772
⎢ 461.7760 −90.1992 ⎥
⎢ ⎥
⎢ −0.190.2 −13.6156 ⎥
K1 = ⎢ ⎥
⎢ −171.1667 −21.6620 ⎥
⎢ ⎥
⎣ 4767.3473 −929.3688 ⎦
−8.6895 −73.0923
⎡ ⎤T
−101.5312 11.0057
⎢ −540.3068 87.9651 ⎥ Fig. 6. State trajectories (position and velocity) in the closed-loop case
⎢ ⎥ without DoS attack.
⎢ 2.7612 −13.8672 ⎥
K2 = ⎢ ⎥
⎢ −508.5127 26.6260 ⎥
⎢ ⎥
⎣ −5566.5 910.0995 ⎦
25.6533 −76.1095
⎡ ⎤T
−57.5108 3.6244
⎢ 543.7766 −79.8642 ⎥
⎢ ⎥
⎢ 1.4527 −13.8752 ⎥
K3 = ⎢ ⎥
⎢ −150.1461 −26.8549 ⎥
⎢ ⎥
⎣ 5609.3236 −8224.9715 ⎦
8.8311 −74.9059
⎡ ⎤T
−93.6458 10.6403
⎢ −471.6240 99.2396 ⎥
⎢ ⎥
⎢ 1.0805 −13.9980 ⎥
K4 = ⎢ ⎥ .
⎢ −480.9800 37.4743 ⎥
⎢ ⎥ Fig. 7. Release instance and release intervals.
⎣ 4852.5740 1025.3948 ⎦
5.8228 −76.8687
(17) still has a satisfactory H∞ noise attenuation performance,
The initial state of the switched closed-loop fuzzy DPS (17) although DoS attacks appear. Fig. 6 reveals the state trajectories
is given as x(0) = [2m/s 0m/s 0rad/s 18m 0m 1.6rad]T . of the switched closed-loop fuzzy DPS (17) without DoS attacks.
The disturbances of surge, sway, and yaw motions ω1 (k), ω2 (k), Comparing Fig. 6 with Fig. 5, we find that the system without
and ω3 (k) are set as ω1 (k) = 1.3 sin(k), ω2 (k) = 0.22 cos(k), DoS attack has superior performance than that with DoS attacks.
and ω3 (k) = cos(1.58 k). Fig. 7 describes the ET release instants and release intervals
The simulation results are shown in Figs. 4–10. As shown in by the designed AMET mechanism. Only “1” indicates that
Fig. 4, the T-S fuzzy DPS (8) with DoS attacks is unstable. Fig. 5 there exists a triggering event. It is apparent from Fig. 7 that
depicts the state trajectories of the switched closed-loop fuzzy the communication resource is saved markedly. Besides, the
DPS (17) with DoS attacks. The switched closed-loop fuzzy DPS dynamically regulated βx (k) and βh (k) are embodied in Fig. 8.

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
 622 IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, VOL. 10, NO. 2, JUNE 2023

Fig. 8. Thresholds βx (k) and βh (k).

Fig. 10. State trajectories (position and velocity) in the closed-loop

case with four different DoS OFF/ON transitions. (a) (0,30) = 3.
(b) (0,30) = 5. (c) (0,30) = 7. (d) (0,30) = 8.

TABLE I
COMPARING RESULTS FOR FOUR DIFFERENT DOS ATTACKS

TABLE II
COMPARING RESULTS FOR THREE DIFFERENT MECHANISMS

Fig. 9. Evolution of four different DoS attacks. (a) (0,30) = 3.

(b) (0,30) = 5. (c) (0,30) = 7. (d) (0,30) = 8.

A. Effects of DoS OFF/ON Transitions on the System B. Seeking a Compromise Between the System
Performance and Triggering Times Performance Level and Economy of Communication
Here, we employ SUM = 30 Resources
k=0 x(k) to reflect the H∞
noise attenuation performance level, and T count to signify From Table II, we find that the TT mechanism in [31] furnishes
the triggering times. Obviously, the smaller value SUM sug- the best H∞ noise attenuation performance at the cost of the most
gests the better H∞ noise attenuation performance within the communication space. By contrast, the ET mechanism in [32]
time interval [0,30]. Fig. 10 exhibits the state trajectories of occupies the least amount of communication resources subject
the switched closed-loop fuzzy DPS (17) with four differ- to the worse H∞ noise attenuation performance. Note that the
ent DoS OFF/ON transitions. It is worth noting that the DoS ATEM proposed in this work can seek a compromise between
OFF/ON transitions increase while retaining the original attacks the system performance level and economy of communication
for the control experiment. It can be concluded that the DoS resources, which promotes the system performance prominently
attacks destroy the stability of dynamics severely. Besides, just by leading to fewer triggering times.
from Table I, we also find that as the DoS OFF/ON transi- Remark 7: As seen in Figs. 4 and 5, the state trajectories of
tions increase, the H∞ noise attenuation performance level the T-S fuzzy DPS (8) with/without control actions are subjected
will get worse and more events will be triggered, which im- to DoS attacks plotted in Fig. 10(b). We can find (0, 30) = 5
plies that malicious attacks can damage the normal running of and (0, 30) = 8. Thus, Assumptions 1 and 2 are tenable with
dynamics. given parameters (ρ, λ, φ, Q).

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
CHENG et al.: NETWORK-BASED ADAPTIVE MULTIEVENT-TRIGGERED FUZZY DYNAMIC POSITIONING CONTROLLER DESIGN 623

VI. CONCLUSION [15] Q. Jia and W. K. Tang, “Event-triggered protocol for the consensus
of multi-agent systems with state-dependent nonlinear coupling,” IEEE
The objective of this article is to design an AMET fuzzy Trans. Circuits Syst. I, Reg. Papers, vol. 65, no. 2, pp. 723–732, Feb. 2018.
dynamic positioning controller for USVs against DoS attacks [16] P. Cheng, S. He, V. Stojanovic, X. Luan, and F. Liu, “Fuzzy fault de-
tection for Markov jump systems with partly accessible hidden informa-
within a finite-time interval. The proposed AMET mechanism tion: An event-triggered approach,” IEEE Trans. Cybern., vol. 52, no. 8,
can improve the data transmission efficiency by dynamically ad- pp. 7352–7361, Aug. 2022.
justing the ET thresholds. Note that the AMET mechanism takes [17] M. Ghaderi, K. Gheitasi, and W. Lucia, “A blended active detection
strategy for false data injection attacks in cyber-physical systems,” IEEE
into account the asynchrony between the membership functions Trans. Control Netw. Syst., vol. 8, no. 1, pp. 168–176, Mar. 2020.
of the controller and that of the system is embedded into the [18] C. Wu, W. Yao, W. Pan, G. Sun, J. Liu, and L. Wu, “Secure control for
AMET mechanism, which can reduce the conservatism of the cyber-physical systems under malicious attacks,” IEEE Trans. Control
Netw. Syst., vol. 9, no. 2, pp. 775–788, Jun. 2022.
control scheme due to mismatched membership functions. Then, [19] Z. Feng and G. Hu, “Secure cooperative event-triggered control of linear
through the use of LMI technology, the criteria of the switched multiagent systems under DoS attacks,” IEEE Trans. Control Syst. Tech-
closed-loop fuzzy DPS are FTB with an H∞ noise attenuation nol., vol. 28, no. 3, pp. 741–752, May 2020.
[20] P.-B. Wang, X.-M. Ren, and D.-D. Zheng, “Event-triggered resilient
performance is achieved. To conclude, we analyze the dynamic control for cyber-physical systems under periodic DoS jamming attacks,”
positioning performance and verify the correctness and validity Inf. Sci., vol. 577, pp. 541–556, 2021.
of the proposed control scheme under the AMET mechanism [21] L. Gao, F. Li, and J. Fu, “Output-based decentralised event-triggered
dissipative control of NCSS under aperiodic dos attacks,” Int. J. Syst. Sci.,
and DoS attacks. In this article, DoS attacks are assumed to be vol. 52, no. 10, pp. 2003–2019, 2021.
of finite energy by Assumptions 1 and 2. However, how do obtain [22] W.-J. Chang, W. Chang, and H.-H. Liu, “Model-based fuzzy modeling
critical conditions under which DoS attacks cannot stabilize the and control for autonomous underwater vehicles in the horizontal plane,”
J. Mar. Sci. Technol., vol. 11, pp. 155–163, 2003.
system? This meaningful topic will be discussed in our further [23] C. De Persis and P. Tesi, “Input-to-state stabilizing control under denial-
research. of-service,” IEEE Trans. Autom. Control, vol. 60, no. 11, pp. 2930–2944,
Nov. 2015.
[24] P. Tague, M. Li, and R. Poovendran, “Mitigation of control channel
jamming under node capture attacks,” IEEE Trans. Mobile Comput., vol. 8,
REFERENCES no. 9, pp. 1221–1234, Sep. 2009.
[25] L.-Y. Hao, H. Zhang, H. Li, and T.-S. Li, “Sliding mode fault-tolerant
[1] Z. Peng, J. Wang, D. Wang, and Q.-L. Han, “An overview of recent
control for unmanned marine vehicles with signal quantization and time-
advances in coordinated control of multiple autonomous surface vehicles,”
delay,” Ocean Eng., vol. 215, 2020, Art. no. 107882.
IEEE Trans. Ind. Informat., vol. 17, no. 2, pp. 732–745, Feb. 2021.
[26] S. He and F. Liu, “Finite-time H∞ fuzzy control of nonlinear jump systems
[2] J. Almeida, C. Silvestre, and A. Pascoal, “Cooperative control of multiple
with time delays via dynamic observer-based state feedback,” IEEE Trans.
surface vessels in the presence of ocean currents and parametric model un-
Fuzzy Syst., vol. 20, no. 4, pp. 605–614, Aug. 2012.
certainty,” Int. J. Robust Nonlinear Control, vol. 20, no. 14, pp. 1549–1565,
[27] P. Cheng, S. He, X. Luan, and F. Liu, “Finite-region asynchronous H∞
2010.
control for 2D Markov jump systems,” Automatica, vol. 129, 2021,
[3] L.-Y. Hao, H. Zhang, T.-S. Li, B. Lin, and C. P. Chen, “Fault tolerant
Art. no. 109590.
control for dynamic positioning of unmanned marine vehicles based on
[28] P. Cheng, S. He, J. Cheng, X. Luan, and F. Liu, “Asynchronous output
TS fuzzy model with unknown membership functions,” IEEE Trans. Veh.
feedback control for a class of conic-type nonlinear hidden Markov jump
Technol., vol. 70, no. 1, pp. 146–157, Jan. 2021.
systems within a finite-time interval,” IEEE Trans. Syst., Man, Cybern.
[4] L. Liu, W. Zhang, D. Wang, and Z. Peng, “Event-triggered extended state
Syst., vol. 51, no. 12, pp. 7644–7651, Dec. 2021.
observers design for dynamic positioning vessels subject to unknown sea
[29] C. Possieri and A. R. Teel, “Stochastic robust simulation and stability
loads,” Ocean Eng., vol. 209, 2020, Art. no. 107242.
properties of chemical reaction networks,” IEEE Trans. Control Netw.
[5] L. Ma, Y.-L. Wang, and Q.-L. Han, “Event-triggered dynamic positioning
Syst., vol. 6, no. 1, pp. 2–12, Mar. 2019.
for mass-switched unmanned marine vehicles in network environments,”
[30] H. O. Wang, K. Tanaka, and M. F. Griffin, “An approach to fuzzy control of
IEEE Trans. Cybern., vol. 52, no. 5, pp. 3159–3171, May 2022.
nonlinear systems: Stability and design issues,” IEEE Trans. Fuzzy Syst.,
[6] T. Takagi and M. Sugeno, “Fuzzy identification of systems and its ap-
vol. 4, no. 1, pp. 14–23, Feb. 1996.
plications to modeling and control,” IEEE Trans. Syst., Man, Cybern.,
[31] Z.-G. Wu, Y. Shen, P. Shi, Z. Shu, and H. Su, “H∞ control for 2-D Markov
vol. SMC-15, no. 1, pp. 116–132, Jan./Feb. 1985.
jump systems in Roesser model,” IEEE Trans. Autom. Control, vol. 64,
[7] J. Cheng, Y. Shan, J. Cao, and J. H. Park, “Nonstationary control for t–S
no. 1, pp. 427–432, Jan. 2019.
fuzzy Markovian switching systems with variable quantization density,”
[32] H. Shen, L. Su, Z.-G. Wu, and J. H. Park, “Reliable dissipative control
IEEE Trans. Fuzzy Syst., vol. 29, no. 6, pp. 1375–1385, Jun. 2020.
for Markov jump systems using an event-triggered sampling information
[8] H. Liu, Y. Pan, J. Cao, Y. Zhou, and H. Wang, “Positivity and stability anal-
scheme,” Nonlinear Anal., Hybrid Syst., vol. 25, pp. 41–59, 2017.
ysis for fractional-order delayed systems: A TS fuzzy model approach,”
IEEE Trans. Fuzzy Syst., vol. 29, no. 4, pp. 927–939, Apr. 2021. Peng Cheng received the M.E. degree in con-
[9] Y. Xue, B.-C. Zheng, and X. Yu, “Robust sliding mode control for TS fuzzy trol engineering from the School of Electrical
systems via quantized state feedback,” IEEE Trans. Fuzzy Syst., vol. 26, Engineering and Automation, Anhui University,
no. 4, pp. 2261–2272, Aug. 2018. Hefei, China, in 2021. He is currently working
[10] X. Liang, J. Xu, and H. Zhang, “Optimal control and stabilization for toward the Ph.D. degree in control science and
networked control systems with asymmetric information,” IEEE Trans. engineering with the Department of Automa-
Control Netw. Syst., vol. 7, no. 3, pp. 1355–1365, Sep. 2020. tion, Shanghai Jiao Tong University, Shanghai,
[11] E. Semsar-Kazerooni and K. Khorasani, “Team consensus for a network China.
of unmanned vehicles in presence of actuator faults,” IEEE Trans. Control His current research interests include un-
Syst. Technol., vol. 18, no. 5, pp. 1155–1161, Sep. 2010. manned marine vehicles, 2-D systems, adaptive
[12] Y.-L. Wang and Q.-L. Han, “Network-based modelling and dynamic output control, stochastic control, asynchronous con-
feedback control for unmanned marine vehicles in network environments,” trol, and fault detection.
Automatica, vol. 91, pp. 43–53, 2018. Dr. Cheng is a Reviewer of many international journals, such as IEEE
[13] N. Meskin and K. Khorasani, “Actuator fault detection and isolation for TRANSACTIONS ON AUTOMATIC CONTROL, IEEE TRANSACTIONS ON CY-
a network of unmanned vehicles,” IEEE Trans. Autom. Control, vol. 54, BERNETICS, IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS:
no. 4, pp. 835–840, Apr. 2009. SYSTEMS, IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: EXPRESS
[14] S. Liet al., “Event-trigger heterogeneous nonlinear filter for wide-area BRIEFS, International Journal of Robust and Nonlinear Control, Me-
measurement systems in power grid,” IEEE Trans. Smart Grid, vol. 10, chanical Systems and Signal Processing, IEEE SYSTEMS JOURNAL, IET
no. 3, pp. 2752–2764, May 2018. Cyber-Physical Systems: Theory Applications, and Nonlinear Dynamics.

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.
 624 IEEE TRANSACTIONS ON CONTROL OF NETWORK SYSTEMS, VOL. 10, NO. 2, JUNE 2023

Chengming Zhang received the B.E. degree in Shuping He (Senior Member, IEEE) received
naval architecture and ocean engineering and the B.S. degree in automation and the Ph.D.
the M.E. degree in design and manufacture of degree in control theory and control engineering
ships and marine structures from Harbin Engi- from Jiangnan University, Wuxi, China, in 2005
neering University, Harbin, China, in 2017 and and 2011, respectively.
2020, respectively. He is currently working to- From 2010 to 2011, he was a Visiting Scholar
ward the Ph.D. degree in control science and with the Control Systems Center, School of
engineering with the Department of Automa- Electrical and Electronic Engineering, Univer-
tion, Shanghai Jiao Tong University, Shanghai, sity of Manchester, Manchester, U.K. He is cur-
China. rently a Professor with the School of Electrical
His research interests include reinforcement Engineering and Automation, Anhui University,
learning, ship motion control, and decision-making in multiagent envi- Hefei, China. He has authored or coauthored more than 90 papers in
ronments. professional journals, conference proceedings, and technical reports in
the given areas, and coauthored a book about stochastic systems. His
current research interests include stochastic systems and stochastic
control, finite-time control theory, reinforcement learning and adaptive
optimal control, system modeling methods and applications, signal pro-
Wei Xie (Member, IEEE) received the M.Sc. de- cessing, and artificial intelligence methods.
gree in electromechanical engineering and the Dr. He was the recipient of the Distinguished Youth Funds Award of
Ph.D. degree in electrical and computer engi- Anhui Province and the Baosteel Excellent Teacher Award. He is an
neering from the University of Macau, Macau, Associate Editor for IEEE ACCESS and an Youth Editor of the Journal of
China, in 2016 and 2021, respectively. Central South University.
He is currently an Assistant Professor with
the Department of Automation, Shanghai Jiao
Tong University, Shanghai, China. His research
interests include nonlinear control of marine and
aerial vehicles, and coordinated control of multi-
vehicle systems.

Weidong Zhang (Senior Member, IEEE) re-

ceived the B.S., M.S., and Ph.D. degrees from
Zhejiang University, Hangzhou, China, in 1990,
1993, and 1996, respectively.
In 1998, he joined Shanghai Jiao Tong Uni-
versity, Shanghai, China, as an Associate Pro-
fessor and has been a Full Professor since
1999. From 2003 to 2004, he was with the Uni-
versity of Stuttgart, Stuttgart, Germany, as an
Alexander von Humboldt Fellow. In 2011, he
was appointed as a Chair Professor with Shang-
hai Jiao Tong University. He is currently Director of the Engineering
Research Center of Marine Automation, Shanghai Municipal Education
Commission, and the Deputy Dean of the Department of Automation,
Shanghai Jiao Tong University. He is the author of one book and more
than 300 refereed papers, and holds 51 patents. His research interests
include control theory and information processing theory and their appli-
cations in several fields, including power/chemical processes, USV/ROV
and aerocraft.
Dr. Zhang was the recipient of the National Science Fund for Distin-
guished Young Scholars of China and Shanghai Subject Chief Scientist.

Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on December 17,2024 at 03:02:49 UTC from IEEE Xplore. Restrictions apply.