Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

Adaptive Fuzzy Event-Triggered Tracking Control


for Nonstrict Nonlinear Systems
Yingkang Xie, Qian Ma, and Zhen Wang

Abstract—This paper addresses adaptive fuzzy event-triggered nonstrict feedback nonlinear systems in recent years (see, e.g.
controller design for nonstrict nonlinear systems. Firstly, for [19]–[22]).
estimating the unmeasurable states, the high-gain fuzzy state With the deepening of research, the authors in [23] realized
observer is designed. By using backstepping technique, the
adaptive fuzzy controller is designed. A new switching threshold that the backstepping technique would cause the problem of
event-triggered mechanism is given to decide when the controller “complexity explosion” and firstly proposed dynamic surface
needs to be updated and the Zeno behavior is avoided. Stability control (DSC) to deal with this problem. Since then, many
analysis shows that the tracking error can be arbitrarily small researchers have made efforts to develop DSC in a large
and all variables of the closed-loop system remain bounded. At number of literature [24]–[26]. For instance, in [27], finite-
last, the effectiveness of our control strategy is illustrated through
simulation. time adaptive fuzzy DSC was considered. The DSC technique
was adopted to reduce the computational complexity for strict-
Index Terms—Fuzzy state observer, event-triggered strategy, feedback nonlinear systems in [28]. In [29], the DSC technique
nonlinear systems, output feedback control.
was used in the underactuated quadrotor control system.
On the other hand, communication and computation abil-
I. I NTRODUCTION ities are always limited in the real applications. Therefore,
from the perspective of energy saving, it is necessary to
D URING the past decades, nonlinear systems have been
widely studied (see, e.g. [1]–[8]). With the increasing
complexity of the control system, the unknown function in
reduce redundant and useless transmission data. The event-
triggered scheme has attracted wide attention, since it ef-
fectively reduces the waste of communication resources and
the system is inevitable. Many results are obtained under the
heavy communication burden. For example, several event-
assumption that the unknown functions satisfy the growth
triggered control (ETC) methods were proposed in [30], [31],
conditions or can be linearly parameterized [9], [10]. But in
with the input-to-state stability (ISS) assumption. In [32],
practical applications, the assumption is difficult to verify.
the authors proposed a new ETC mechanism and proved
Since fuzzy logic systems can effectively approximate the
that the proposed mechanism was Zeno free. In [33], the
unknown functions, fuzzy control has been widely used in
adaptive event-triggered global practical tracking control was
practical systems (see, e.g. [11]–[14]).
investigated and two distinct event-triggered mechanisms were
However, it is difficult to establish the exact model of
proposed. In [34], the switching ETC method was used to
plant in practical applications. Thus, the uncertainty of the
achieve global stabilization. In [35]–[38], the ETC methods
system should be considered. Based on backstepping tech-
were applied to delayed nonlinear systems. We should notice
niques, adaptive fuzzy control is proposed to deal with these
the fact that ETC saves communication resources at the
problems due to its strong robustness and low conservation.
expense of system performance. Although great achievements
Great achievements about uncertain nonlinear systems have
about ETC have been obtained, few literature considers both
been obtained (see, e.g. [15]–[18]). As is well known, the
system performance and communication constraints [39]. It
adaptive backstepping method is often used in strict nonlinear
is worth pointing out that how to balance communication
systems. Nevertheless, numerous actual systems, for instance,
constraints and system performance is still a problem.
biological systems, chemical systems and physical systems,
In this paper, the adaptive fuzzy event-triggered tracking
are nonstrict nonlinear systems. Generally, the control methods
control for uncertain nonlinear system with external distur-
of strict feedback systems cannot be directly used in nonstrict
bance is considered. Since the system is nonstrict form, it
feedback systems, since it may result in “algebraic loop
is difficult to use backstepping method to design the virtual
problem”. To solve this problem, a few meaningful adaptive
control laws and adaptive laws. Besides, it is challenging
fuzzy control approaches have been proposed for uncertain
to design a controller which can balance the communication
This work was supported in part by the NSFC under Grant 61773207, constraints and system performance as well as achieve good
62173183, and the Natural Science Fund for Distinguished Young Scholars tracking effect. The main contributions of the paper are as
of Jiangsu Province under Grant BK20190020. (Corresponding author: Qian follows:
Ma.)
Y. Xie and Q. Ma are with the School of Automation, Nanjing University of 1) The novel adaptive switching threshold event-triggered
Science and Technology, Nanjing 210094, China (e-mail: xyk150@163.com; controller is designed in this paper. Compared with the con-
qianmashine@gmail.com). troller in [32], [38], [40], the proposed switching threshold
Z. Wang is with College of Mathematics and Systems Science, Shandong
University of Science and Technology, Qingdao, 266590, China (e-mail: event-triggered controller can balance system performance and
wangzhen@sdust.edu.cn). communication constraints.

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

2) The nonlinear functions can be totally unknown, which Lemma 1 [43]: f (x) is a continuous function defined on a
means that the growth conditions and the global Lipschitz compact set Ω, and for any positive constant , there is an FLS
conditions proposed in [38], [40] are removed completely so that
sup f (x) − θT ϕ(x) ≤ .

by using the high-gain fuzzy state observer. Moreover, the (5)
high-gain fuzzy state observer can also handle the external x∈Ω
disturbances well.
3) The ISS condition required in [30], [41], [42] is no III. S TATE O BSERVER
longer needed by combining the adaptive controller design In this section, the high-gain fuzzy state observer will be
and the event-triggered design. The proposed design scheme designed to observe the unavailable states.
can enhance robustness and reduce conservatism. Rewrite system (1) as follows:
n
X
II. P ROBLEM S TATEMENT ẋ = Ax + Bi fi (x) + Cu + D, (6)
A. Model Descriptions i=1

The following nonlinear system is considered: where  


 0
 ..
 ẋi (t) = xi+1 (t) + fi (x(t)) + di (t), A =  . In−1 ,
 
ẋn (t) = u(t) + fn (x(t)) + dn (t), (1) 0 ... 0


y = x1 , i (7)
z }| {
Bi = [0, . . . , 0, 1, 0 . . . , 0]T ,
where i = 1, . . . , n − 1 and x = [x1 , . . . , xn ]T represents the
C = [0, . . . , 1]T ,
system state vector. u and y denote system input and con-
D = [d1 , . . . , dn ].
trol output, respectively. di (t) represent external disturbances.
fi (x) are unknown nonlinear continuous functions and local By Lemma 1, using FLSs to estimate functions fi (x), we get
Lipschitz in the argument. Moreover, fi (x) satisfy fi (0) = 0.
The output y is the only measurable variable. fi (x | θi ) = θiT ϕi (x) , (8)
Assumption 1: The external disturbances di (t) satisfy
¯ which satisfies
|di (t)| ≤ d.
Assumption 2: The reference signal yr is known and yr , ẏr , fi (x) = fi (x | θi ) + δi (x), |δi (x)| < . (9)
and ÿr are bounded.
The purpose of this paper is to design an adaptive fuzzy Then, the following state observer is constructed:
output feedback event-triggered controller for systems (1),  
which can ensure the output y tracks yr in arbitrarily small x̂˙ i = x̂i+1 + fˆi x̂ | θ̂i − k i li (x̂1 − x1 ),
  (10)
error and all variables of the closed-loop system remain x̂˙ n = uτ + fˆn x̂ | θ̂i − k n ln (x̂1 − x1 ),
bounded.
where i = 1, . . . , n − 1. k and li are the design parameters
satisfying k > 1 and li > 0. x̂i represents the estimation of
B. Fuzzy Logic Systems (FLSs) T
xi . Let x̂ = [x̂1 , x̂2 , . . . , x̂n ] . Then the observer error can
T T
The knowledge base consists of a series of fuzzy IF- be defined as e = [x1 − x̂1 , . . . , xn − x̂n ] , [e1 , . . . , en ] .
THEN rules, as R l : IF x1 is Φl1 and . . . and xn is Φln , According to (6) and (10), the following error system can be
THEN y is Θl , where l = 1, . . . , N and N represents the obtained:
number of fuzzy rules. xi and y represent the input and the n   
output of FLSs. Φli and Θl are fuzzy sets, and their fuzzy
X
ė = Ak e + Bi fi (x) − fˆi x̂ | θ̂i + D, (11)
membership functions are µΦli (xi ) and µΘl (y), respectively. i=1
The FLSs can be described as
PN Qn where  
−l1 k
l=1 ȳl i=1 µΦli (xi )
y(x) = P Q , (2) Ak =  ..
. (12)
 
N n . In−1
l=1 i=1 µΦli (xi )
−ln k n 0 ... 0
where ȳl = maxy∈R µΘl (y). Define fuzzy basis functions as
Based on (9) and (10), (11) can be written as
Qn
µΦl (xi )
i=1 n n
ϕl =  i .
(3)
X X
PN
l=1
Qn
i=1 µΦl (xi ) ė = Ak e + Bi θ̃iT ϕi (x̂) + Bi ∆fi (x) + δ + D, (13)
i
i=1 i=1
T T
Let ϕ(x) = [ϕ1 (x), . . . , ϕN (x)] , θ = [ȳ1 , . . . , ȳN ] = T
T T where ∆fi (x) = θiT (ϕi (x) − ϕi (x̂)), and δ = [δ1 , . . . , δn ] .
[θ1 , . . . , θN ] and x = [x1 , . . . , xn ] , then (2) can be de-
Now, applying the coordinate transformation
scribed as:
y(x) = θT ϕ(x). K = diag 1, k −1 , . . . , k 1−n .

(4) ξ = Ke, (14)

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

Based on (13) and (14), we get where αi−1 and si are the input and output of the first-order
n
X n
X filter, and χi stand for the output error of it.
ξ˙ =kAo e + K Bi θ̃iT ϕi (x̂) + K Bi ∆fi (x) Step 1: From (1) and (21), we obtain
i=1 i=1
(15)
ż1 =x2 + f1 (x) + d1 − ẏr
+ K(δ + D),
=z2 + α1 + χ2 + e2 + θ1T ϕ1 (x) + δ1 (x) + d1 − ẏr .
where   (22)
−l1 Choose Lyapunov function candidate as
 ..
Ao =  . . (16)

In−1 1 1 T
V1 = Vξ + z12 + θ̃ θ̃1 , (23)
−ln 0 ... 0 2 2a1 1
Obviously, Ao is a strict Hurwitz matrix. Thus, for a matrix where a1 > 0 is the design parameter. Based on (22), we get
Q > 0, there is a matrix P > 0 such that 1 T˙
V̇1 =V̇ξ + z1 ż1 − θ̃ θ̂1
a1 1
ATo P + P Ao = −Q. (17)
=V̇ξ + z1 (z2 + α1 + χ2 + e2 + δ1 (x) + d1 − ẏr
Further, construct a Lyapunov candidate Vξ = ξ T P ξ. Accord-
+ θ1T ϕ1 (x) − θ1T ϕ1 (x̂1 ) + θ̂1T ϕ1 (x̂1 ) + θ̃1T ϕ1 (x̂1 ))
ing to (15), we get
1 ˙
n n − θ̃1T θ̂1 .
X X a1
V̇ξ =2ξ T P [kAo e + K Bi θ̃iT ϕi (x̂) + K Bi ∆fi (x) (24)
i=1 i=1 By using Youngs inequality, we obtain
+ K(δ + D)]. 1
(18) z1 (z2 + χ2 + d1 ) ≤ (3z 2 + z22 + χ22 + d¯2 ), (25)
According to 0 < ϕTi ϕi < 1, k ≥ 1 and Young’s inequality, 2 1
1
we obtain z1 (e2 + δ1 (x)) ≤ (2z 2 + 2 + e22 ), (26)
2 1
(kθ̃i kkP k)2
Pn
Pn
2ξ T P K i=1 Bi θ̃iT ϕi (x̂) ≤ kξ T ξ + Pi=1 k , 1 T
z1 (θ1T ϕ1 (x) − θ1T ϕ1 (x̂1 )) ≤ υz12 + θ θ1 , (27)
T
Pn T 4 n i=1 (kθi kkP k)
2
υ 1
2ξ P K i=1 Bi ∆Fi (x) ≤ kξ ξ + k ,
2 where υ > 0 is the design parameter. Let
2ξ T P (δ + D) ≤ 2kξ T ξ + kP k k(D̄+¯) ,
(19) 5
α1 = −κ1 z1 − ( + υ)z1 − θ̂1T ϕ1 (x̂1 ) + ẏr , (28)
where D̄ > 0 and ¯ > 0 satisfy kDk2 ≤ D̄ and kδk2 ≤ ¯. 2
Then, substituting (19) into (18) yields ˙
θ̂1 = a1 z1 ϕ1 (x̂1 ) − ρ1 θ̂1 . (29)
Pn 2

i=1 θ̃i kP k2 Then, we get
V̇ξ ≤ − (λmin (Q) − 4) kξ T ξ + + β0 , 2
k Pn 2
i=1 θ̃i kP k

(20) 1
Pn  V̇1 ≤ − h1 ξ T ξ + − κ1 z12 + χ22
where β0 = 4 i=1 (kθi k kP k)2 + kP k2 (D̄ + ¯) /k. k 2 (30)
Remark 1: Compared with the fuzzy state observer used in 1 ρ1
+ z22 + θ̃1T θ̂1 + β1 ,
[27], [40], in the high-gain fuzzy state observer, the design 2 a1
parameter k is introduced. k appears in (20) and if k is
where h1 = λmin (Q)k − 4k − 21 , β1 = β0 + 12 (2 + d¯2 ) +

sufficiently large, the influence of external disturbance will 1 T
be small. Thus, the high-gain fuzzy state observer can handle υ θ1 θ1 . Based on [23], the new state variable s2 is introduced.
Let α1 pass the first-order filter with a time constant λ2 , and
external disturbances well.
get s2 as follows:

IV. C ONTROLLER D ESIGN AND S TABILITY A NALYSIS λ2 ṡ2 + s2 = α1 , s2 (0) = α1 (0). (31)
In this section, the virtual control law and the event- Thus, from (31), we can obtain
triggered controller will be designed. The stability analysis χ2
of the closed loop system will be studied. ṡ2 = − . (32)
λ2
According to (21) and (32), we get
A. Virtual Control Law Design χ2
χ̇2 = ṡ2 − α̇1 = − − α̇1 . (33)
The virtual control laws will be constructed on the basis λ2
of adaptive control and backstepping techniques. Since back- Step i: From (10) and (21), we obtain
stepping technique needs n steps, first we give the coordinate   χi
transformation of each step as follows: żi =x̂i+1 + fˆi x̂ | θ̂i − k i li (x̂1 − x1 ) + . (34)
λi
z1 = y − yr , Choose Lyapunov function candidate as
zi = x̂i − si , 2 ≤ i ≤ n, (21) 1 1 T 1
χi = si − αi−1 , Vi = Vi−1 + zi2 + θ̃i θ̃i + χ2i , (35)
2 2ai 2

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

where ai is a design parameter. Then, we obtain where υ is a design parameter. By using Youngs inequality,
χi we obtain
V̇i =V̇ξ + zi (zi+1 + αi + χi+1 + k i li e1 + (zn k n ln e1 )2
λ zn k n ln e1 ≤ + r12 ,
T T
 T
 i h21
(48)
ˆ
+ θi ϕi (x̂) − θi ϕi x̄i + θ̂i ϕi x̄i ˆ (36)
1 χn (zn χn )2
ˆi ) − θ̃iT θ̂˙i + χi χ̇i ,
+ θ̃iT ϕi x̄ zn ≤ + r22 , (49)

ai λn h22 λ2n
(zn θ̂nT ϕn (x̂))2
ˆi = (x̂1 , x̂2 , · · · , x̂i ). By using Youngs inequality, we
with x̄ zn θ̂nT ϕn (x̂) ≤ + r32 , (50)
h23
obtain
where r1 , r2 and r3 are design parameters. Let
1
zi (zi+1 + χi+1 ) ≤ (2z 2 + zi+1
2
+ χ2i+1 ), (37)
2 i (θ̂nT ϕn (x̂))2 (k n ln e1 )2 χ2
1 αn = − zn ( 2 + 2 + 2 n2
zi (θiT ϕi (x̂) − θiT ϕi x̄
ˆi ) ≤ υzi2 + θiT θi ,

(38) h3 h1 h2 λn
υ + κn + 1), (51)
where υ is a design parameter. Let ˙
θ̂n =an zn ϕn (x̂) − ρθ̂n . (52)
χi
αi = −κi zi − (2 + υ)zi − θ̂iT ϕi x̄ − k i li e1 − , (39)

ˆi Then, it yields
λi
2
˙ 
ˆi − ρθ̂i .
Pn
θ̂i = ai zi ϕi x̄ (40) θ̃i kP k2 n

i=1 X
V̇n ≤ − h1 ξ T ξ + − κj zj2
Then, we get k j=1
n
2 X ρj (53)
Pn
θ̃i kP k2
i + θ̃jT θ̂j + βn + zn (u − αn )
T i=1 X aj
V̇i ≤ − h1 ξ ξ + − κj zj2 j=1
k n
j=1 X χj 1
i i − χj ( + α̇j−1 − χj ),
X ρj X χj 1 (41) λj 2
+ θ̃jT θ̂j − χj ( + α̇j−1 − χj ) j=2
aj λj 2
j=1 j=2 where βn = βn−1 + r12 + r22 + r32 .
1 2 1
+ zi+1 + χ2i+1 + βi ,
2 2 B. Event-Triggered Controller Design
Pi
where βi = β0 + 12 (2 + d¯2 )+ j=1 υ1 θjT θj . The state variable Compared with the fixed threshold event-triggered strate-
si+1 can be introduced. Let αi pass the first-order filter with gy (FTETS), the relative threshold event-triggered strategy
a time constant λi+1 , and get si+1 as follows: (RTETS) can reduce the number of trigger times. However,
for RTETS, large measurement errors of the control input
λi+1 ṡi+1 + si+1 = αi , si+1 (0) = αi (0). (42) are unavoidable, if the amplitude of the control input is
sufficiently large. In this case, the large impulse will certainly
Thus, from (42), we can obtain degrade system performance. On the contrary, for FTETS, the
−si+1 + αi χi+1 measurement error is bounded. Therefore, based on the above
ṡi+1 = − . (43) observations, we proposed the following switching threshold
λi+1 λi+1
event-triggered strategy (STETS):
Thus, it yields 
χi+1  v1 (t) = (1 + δ)αn − sign(zn )σ1 ,
χ̇i+1 = −α̇i − . (44) u1 (t) = v1 (tk ) , for ∀t ∈ [tk , tk+1 ) , (54)
λi+1
tk+1 = inf{t > tk ||ẽ1 (t)| > δ|u(t)| + σ1 },

Step n: From (10) and (21), we obtain

 v2 (t) = αn − sign(zn )σ2 ,
  χn u2 (t) = v2 (tk ) , for ∀t ∈ [tk , tk+1 ) , (55)
żn =u(t) + fˆn x̂ | θ̂i − k n ln (x̂1 − x1 ) + . (45) 
tk+1 = inf{t > tk ||ẽ2 (t)| > σ2 },
λn
Choose Lyapunov function candidate as where 0 ≤ δ < 1 and σ1 , σ2 > 0. ẽ1 (t) = v1 (t) − u1 (t) and
ẽ2 (t) = v2 (t) − u2 (t). The controller u is as follows:
1 1 T 1
Vn = Vn−1 + zn2 + θ̃n θ̃n + χ2n , |u(t+ )| < m,

(46) u1 (t),
2 2an 2 u(t) = (56)
u2 (t), |u(t+ )| ≥ m.
where an is a design parameter. According to (45), we obtain The design procedures is in Fig .1.
χn Remark 2: STETS consists of FTETS and RTETS. If the
V̇n =V̇n−1 + zn (u(t) + k n ln e1 + + θnT ϕn (x̂) control signal meets the condition |u(t+ )| < m, RTETS is
λn
1 T˙ (47) adopted to obtain accurate control; Otherwise, when |u(t+ )| ≥
− θ̂nT ϕn (x̂) + θ̂nT ϕn (x̂)) − θ̃ θ̂n + χn χ̇n , m, RTETC is switched to FTETS, so that the measurement
an n

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

(39), we obtain
xi (t )  xi 1 (t )  f i ( x(t ))  di (t ) Pn 2
θ̃i kP k2 n

xn (t )  u (t )  f n ( x(t ))  d n (t ) i=1 X
V̇ ≤ − h1 ξ T ξ + − κj zj2
k
y j=1
n
ρj
 
X
θ̃jT θ̂j (58)
xˆi  xˆi 1  fˆi xˆ | ˆi  k ili (xˆ1  x1 ) + βn + zn (u − αn ) +
aj
j=1
xˆn  u  fˆ  xˆ | ˆ   k l (xˆ  x )
 n i
n
n 1 1
n
X χj 1
− χj ( + α̇j−1 − χj ).
j=2
λj 2

 
ˆi  ai zii xˆi  ˆi By using Youngs inequality, we get
n n n
X ρj X ρj 2 X ρj T
θ̃jT θ̂j ≤ θj − θ̃ θ̃j , (59)
5 aj 2aj 2aj j
1  1 z1  (   ) z1  ˆ1T1  xˆ1   y r
j=1 j=1 j=1
2 Xn Xn
1 2 2
 − χj α̇j−1 ≤ χj α̇j−1 + 2(n − 1)τ. (60)
 
 i   i zi  (2   ) zi  ˆiTi xˆi  k ili e1  i
i j=2 j=2

n (ˆn n  xˆ )
n 2 2 T 2 Substituting (59) and (60) into (58) yields
(k ln e1 )
 n   zn ( n  1    )
h12 h22 n2 h32 Xn
ρj kP k2 T Xn
V̇ ≤ − h1 ξ T ξ − ( − )θ̃j θ̃j − κj zj2
j=1
2a j k j=1
v1 (t )  (1   ) n  sign( z n )1
n
1
X 1 2 1
+ zn (u − αn ) − − χ2j (
α̇j−1 − ) + β,
v2 (t )   n  sign( z n ) 2 λ j 2τ 2
j=2
Pn ρ (61)
switching threshold where β = βn + 2(n − 1)τ + j=1 2ajj θj2 . For |u(t+ )| <
u m, based on (56), we get that there are two time-varying
event-triggered strategy
parameters η1 (t), η2 (t) with η1 (ti ) = η2 (ti ) = 0,
η1 (ti+1 ) , η2 (ti+1 ) = ±1, and |η1 (t)| , |η2 (t)| ≤ 1, ∀t ∈
[ti , ti+1 ), satisfying v1 (t) = (1 + η1 (t)δ) u(t)+η2 (t)σ1 . Thus,
Fig 1. Design procedures. we get
v1 (t) σ1 η2 (t)
u(t) = − . (62)
1 + η1 (t)δ 1 + η1 (t)δ
error will remain within a certain range to ensure system Therefore, we obtain
performance. Clearly, compared with the RTETS which is
v1 (t) σ1 η2 (t)
proposed in [32], STETS can balance the system performance zn (u − αn ) = zn ( − − αn ). (63)
and the communication constraints. 1 + η1 (t)δ 1 + η1 (t)δ
Remark 3: As can be seen, when m is sufficiently large, Easy to get that zn αn ≤ 0, thus we obtain
STETS can be seen as the relative one and when m = 0,
the STETS can be seen as the fixed one. Thus, our proposed zn (u − αn )
strategy is more general. (1 + δ)αn − sign(zn )σ1 − σ1 η2 (t)
=zn ( − αn )
1 + η1 (t)δ (64)
C. Stability Analysis −sign(zn )σ1 − σ1 η2 (t)
≤zn ( ).
1 + δη1 (t)
Theorem 1: Considering system (1) with any bounded initial
conditions, our proposed control strategy consisting of the state For zn = 0, easy to get that
estimator (10) with update laws (29), (40), (52), virtual control
zn (u − αn ) = 0. (65)
laws (28) and (39) and the controller (56) ensures that the
output y tracks yr in a bounded error and all variables of For zn 6= 0, we get
the closed-loop system remain bounded. Moreover, the Zeno
behavior can be avoided. zn (u − αn ) ≤ 0. (66)
Proof: Let
For |u(t+ )| ≥ m, based on (56), we get that there are two time-
n n T n
X zi2 θ˜i θ˜i
X χ2i X varying parameters η3 with η3 (ti ) = 0, η3 (ti+1 ) = ±1, and
V = ξT P ξ + + + . (57) |η3 (t)| ≤ 1, ∀t ∈ [ti , ti+1 ), satisfying v2 (t) = u(t) + η3 (t)σ2 .
i=1
2 i=1
2ai i=2
2
Thus, we get
By update laws (29), (40), (52), the virtual control (28) and u(t) = v2 (t) − η3 (t)σ2 . (67)

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

Therefore, we obtain |v2 (ti+1 ) − v2 (ti )| > σ2 , which is a contradiction. Therefore,


zn (u − αn ) =zn (αn − sign(zn )σ2 − η3 (t)σ2 − αn ) we conclude that the Zeno behavior can be avoided.
(68) Remark 4: From (73), it is easy to get that the tracking error
≤zn (−sign(zn )σ2 − η3 (t)σ2 ).
is related to β and c. Both of the two parameters are related
For zn = 0, easy to get that to the design parameters proposed in observer (10), Lyapunov
function (46), the virtual control law (28), (39), the adaptive
zn (u − αn ) = 0. (69)
law (40), and the controller (56). Thus, in order to obtain
For zn 6= 0, we get the better tracking effect, we can select the design parameters
appropriately to reduce β and increase c.
zn (u − αn ) ≤ 0. (70)
Remark 5: A novel switching threshold adaptive event-
Based on the above discussion, we get triggered controller is designed by using the sign function,
n n which compensates for the measurement error. Thus, the ISS
X ρj kP k2 T X
condition is no longer needed.
V̇ ≤ − h1 ξ T ξ − ( − )θ̃j θ̃j − κj zj2
j=1
2a j k j=1 Remark 6: By using FLSs to approximate nonlinear func-
n (71) tions, the nonlinear functions can be totally unknown. But the
X 1 1 2 1
− ( − α̇j−1 − )χ2j + β. approximation error will be introduced. Thus, it is difficult to
λj 2τ 2
j=2 achieve asymptotic stability of the closed loop system.
h1 2kP k2 a Based on the above discussion, the controller design rules
Let c = min{ λmax (P ) , ρj − k
j
, 2κj , 2( λ1i − 2τ
1 2
α̇i−1 − 12 )}, and procedures are given as follows, from which we can find
then we obtain that the computational complexity is low due to just the simple
V̇ ≤ −cV + β. (72) algebra computation is involved.
By directly integrating (72), we have Step 1: Define the IF-THEN fuzzy rules and membership
  functions of fuzzy sets.
β β
0 ≤ V (t) ≤ + Vn (0) − e−ct , (73) Step 2: Select the design parameters k and li and design
c c the state observer as (10).
which means that the output y tracks yr in a bounded error Step 3: Design the Lyapunov function Vi as (23), (35), (46),
and all variables of the closed-loop system remain bounded. and select the design parameters ai .
Next, we will exempt Zeno behavior. Assume that ∆i = Step 4: Design the virtual control law αi (39), the adaptive
ti+1 − ti → 0. For |u(t+ )| < m, we get laws θ̂i (40), and select the design parameters κi , λi and v.
lim |v1 (ti+1 ) − v1 (ti )| Step 5: Design the adaptive laws θ̂n , v1 and v2 . Select
∆i →0 the design parameters δ, σ1 and σ2 . Then, select the design
=|(1 + δ)αn (ti+1 ) − sign(zn (ti + 1))σ1 parameter m and design the STETS to get the controller u.
− ((1 + δ)αn (ti ) − sign(zn (ti ))σ1 )| (74)
≤ lim |(1 + δ)(αn (ti+1 ) − αn (ti ))| V. S IMULATION EXAMPLE
∆i →0
+ lim |σ1 (sign(zn (ti+1 )) − sign(zn (ti )))|. In this section, two simulation examples are given to support
∆i →0 our theoretical results. The corresponding programs can be
Since αn is continuous, we get lim∆i →0 (αn (ti+1 ) − found at https://www.researchgate.net/publication/355034007
αn (ti )) → 0. Thus TFS programs.
Example 1: The following one-link manipulator with a
lim |v1 (ti+1 ) − v1 (ti )|
∆i →0 motor proposed in [20] is adopted to support our theoretical
(75)
≤ lim |σ1 (sign(zn (ti+1 )) − sign(zn (ti )))|. results:
∆i →0 M q̈ + C q̇ + G sin(q) = sin(t) + τ,
(77)
Since zn is continuous, for zn = 0, we get lim∆i →0 |v(ti+1 )− B τ̇ + Hτ = cos(t) + u − Km q̇.
v(ti )| ≤ σ1 . For zn 6= 0, we get lim∆i →0 |v(ti+1 )−v(ti )| = 0.
Thus, lim∆i →0 |v(ti+1 ) − v(ti )| ≤ σ1 is always satisfied. But where q is the output of the system. The values of these
by the triggered condition, we get |v(ti+1 ) − v(ti )| > δ|u| + parameters are M = 1, C = 1, G = 10, B = 0.1, H = 1,
σ1 ≥ σ1 , which is a contradiction. and Km = 0.2. Then (77) can be rewritten as
For |u(t+ )| ≥ m, we get ẋ1 = x2 ,
lim |v2 (ti+1 ) − v2 (ti )| ẋ2 = x3 − x2 − 10 sin (x1 ) + sin(t), (78)
∆i →0
=|αn (ti+1 ) − sign(zn (ti + 1))σ2 (76) ẋ3 = 10u − 2x2 − 10x3 + cos(t).
− (αn (ti ) − sign(zn (ti ))σ2 )|. Choose the reference signal as yr = sin(t) and the fuzzy
membership functions as
With the same method above, we get that for zn =
0, lim∆i →0 |v2 (ti+1 ) − v2 (ti )| ≤ σ2 and for zn 6= 0, −(x̂i −4)2 −(x̂i −2)2 −(x̂i −0)2

lim∆i →0 |v(ti+1 ) − v(ti )| = 0. Thus, limt∆i →0 |v2 (ti+1 ) − µi1 = e 8 , µi2 = e 8 , µi3 = e 8 ,
−(x̂i +2)2 −(x̂i +4)2
v2 (ti )| ≤ σ2 satisfies. But by the triggered condition, we get µi4 = e 8 , µi5 = e 8 , i = 1, 2, 3

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

and
20
u1j
ϕj (x1 ) = P5 ,
j=1 u1j 15
u1j u2j
ϕj (x1 , x2 ) = P5 ,
j=1 u1j u2j
10

u1j u2j u3j


ϕj (x1 , x2 , x3 ) = P5 . 5
j=1 u1j u2j u3j
0
Choose the design parameters as k = 6, l1 = 7, l2 = 8,
l3 = 10, a1 = a2 = a3 = 5, p1 = p2 = p3 = 0.5, κ1 = κ2 = -5
κ3 = 5, v = 30, λ2 = λ3 = 0.1, δ = σ1 = σ2 = 0.5 and
m = 10. Set the initial values as x1 (0) = 0.2, x2 (0) = 0.2, -10
x3 (0) = 0.2 and all the other values as zero to get Figs. 2-5.
Fig. 2 shows time trajectories of output y and reference signal -15
0 1 2 3 4 5 6 7 8 9 10
yr . Fig. 3 gives the time trajectories of the states x1 − x3 . t(sec)
The time trajectory of controller u is shown in Fig. 4 and the
time trajectories of kθ1 k2 , kθ2 k2 , kθ3 k2 are shown in Fig. 5. Fig 3. Trajectories of states x1 − x3 .
The simulation results show that the tracking error is small
and all variables of the closed-loop system remain bounded,
which shows the effectiveness of our controller.
150

1 100

0.8
50
0.6

0.4 0

0.2
-50
0
-100
-0.2

-0.4 -150

-0.6
-200
-0.8 0 1 2 3 4 5 6 7 8 9 10
t(sec)
-1
0 1 2 3 4 5 6 7 8 9 10
Fig 4. Trajectories of u.
t(sec)

Fig 2. Trajectories of y and yr .

0.8
Example 2: To compare our STETS with RTETS [32], [44]
and FTETS [45], consider the following second-order nonstrict 0.7
nonlinear system:
0.6
2

 ẋ1 = x2 (t) + 0.2x1 (t) + 0.1sin(x2 (t)) + 0.2sin(t),

0.5
ẋ2 = u(t) + 0.1x2 (t) + 0.2cos(x1 (t)) + 0.2cos(t), (79)
 0.4
y = x1 .

0.3
Choose the reference signal as yr = sin(t) and the fuzzy
membership functions as 0.2

−(x̂i −4)2 −(x̂i −2)2 −(x̂i −0)2


0.1
µi1 = e 8 , µi2 = e 8 , µi3 = e 8 ,
−(x̂i +2)2 −(x̂i +4)2
µi4 = e 8 , µi5 = e 8 , i = 1, 2, 0
0 1 2 3 4 5 6 7 8 9 10
t(sec)
and
Fig 5. Trajectories of kθ1 k2 , kθ2 k2 , kθ3 k2 .
u u u
ϕj (x1 ) = P5 1j , ϕj (x1 , x2 ) = P5 1j 2j .
j=1 u1j j=1 u1j u2j

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

1.5
40

1
20

0.5 0

-20
0

5
-40 0
-0.5
-5

-60 -10
-15
-1
7 8 9 10
-80

-1.5
0 2 4 6 8 10 12 14 16 18 20 -100
t(sec) 0 2 4 6 8 10 12 14 16 18 20
t(sec)
Fig 6. Trajectories of x1 and x̂1 with STETS.
Fig 8. Trajectory of controller u with STETS.

2.5

2 10-3
3
1.5

1 2.5

0.5
2
0

-0.5
1.5
-1

-1.5 1

-2
0.5
-2.5
0 2 4 6 8 10 12 14 16 18 20
t(sec)
0
0 2 4 6 8 10 12 14 16 18 20
Fig 7. Trajectories of x2 and x̂2 with STETS.
t(sec)

Fig 9. Trajectories of kθ1 k2 and kθ2 k2 with STETS.


Let ϕ(x1 ) = [ϕ1 (x1 ), · · · , ϕ5 (x1 )] and ϕ(x1 , x2 ) =
[ϕ1 (x1 , x2 ), · · · , ϕ5 (x1 , x2 )]. Then, design fuzzy state observ-
er as
0.45
x̂˙ 1 = x̂i+1 + θ̂1 ϕ(x̂1 , x̂2 ) − kl1 (x̂1 − x1 ),
(80)
x̂˙ 2 = u + θ̂2 ϕ(x̂1 , x̂2 ) − k 2 l2 (x̂1 − x1 ), 0.4

0.35
where the design parameters are chosen as k = 3, l1 = 3 and
l2 = 5. 0.3
Let a1 = a2 = 1, ρ1 = ρ2 = 0.5, κ1 = 2, κ2 = 5, λ2 = 0.1,
0.25
δ = 0.9, v = 10, σ1 = 0.1, σ2 = 0.1, r1 = r2 = r3 = 1 and
m = 10. Set the step size as t = 0.001s. Choose the initial 0.2

values as x1 (0) = 0.4, x2 (0) = 0.8 and all the other values 0.15
as zero.
0.1
TABLE I
C OMPARISON OF THREE CONTROL STRATEGIES 0.05

Event-triggered strategy STETS RTETS [32], [44] FTETS [45] 0


0 2 4 6 8 10 12 14 16 18 20
Control object Achieve Fail Achieve
t(sec)
Trigger times 653 - 884
Fig 10. Time intervals with STETS.
The time trajectories of the states x1 , x̂1 and x2 , x̂2 are

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

10
104 the unmeasurable states, the high-gain fuzzy state observer
is designed. The virtual control laws are obtained by using
8 adaptive backstepping techniques. Moreover, a novel switching
6 threshold event-triggered controller is designed. The results
show that our proposed controller ensures that all signals of
4
the closed-loop system remain bounded. Besides, the Zeno
2 behavior is avoided. Finally, the validity of the results is
verified by simulation examples.
0

-2 R EFERENCES
-4 [1] M. Chen, S. Shao and B. Jiang, “Adaptive neural control of uncertain
nonlinear systems using disturbance observer,” IEEE Trans. Cybern.,
-6 vol. 47, no. 10, pp. 3110-3123, Oct. 2017.
[2] Y. Li, T. Yang and S. Tong, “Adaptive neural networks finite-time
-8 optimal control for a class of nonlinear systems,” IEEE Trans. Neural
0 0.005 0.01 0.015
t(sec)
Netw. Learn. Syst., vol. 31, no. 11, pp. 4451-4460, Nov. 2020.
[3] W. Sun, S. F. Su, Y. Wu, J. Xia and V. T. Nguyen, “Adaptive fuzzy con-
Fig 11. Trajectory of u with RTETS used in [32]. trol with high-order barrier Lyapunov functions for high-order uncertain
nonlinear systems with full-state constraints,” IEEE Trans. Cybern., vol.
50, no. 8, pp. 3424-3432, Aug. 2020.
[4] Z. Chen, B. Zhang, Y. Zhang, Q. Ma and Z. Zhang, ”Event-based control
1
for networked TCS fuzzy systems via auxiliary random series approach,”
y
IEEE Trans. Cybern., vol. 50, no. 5, pp. 2166-2175, May 2020.
r
0.8 [5] L. Liu, Y. Liu, S. Tong and C. L. P. Chen, “Integral barrier Lyapunov
y with STETS
y with FTETS function based adaptive control for switched nonlinear systems,” Sci.
0.6 China Inf. Sci., 63, no. 3, pp. 1-14, Mar. 2020.
0.4
[6] R. E. Precup, S. Preitl and G. Faur, “PI predictive fuzzy controllers
for electrical drive speed control: methods and software for stable
0.2 development,” Comput. Ind., vol.52, no. 3, pp. 253-270, Dec. 2003.
[7] Y. Xie and Q. Ma, “Adaptive event-triggered neural network control
0 for switching nonlinear systems with time delays,” IEEE Trans. Neural
Netw. Learn. Syst., doi: 10.1109/TNNLS.2021.3100533.
-0.2
[8] A. Turnip and J. H. Panggabean, “Hybrid controller design based
-0.4 magneto-rheological damper lookup table for quarter car suspension,”
Int. J. Artif. Intell, vol. 18, no. 1, pp. 193-206, 2020.
-0.6 [9] Y. Li, “Finite time command filtered adaptive fault tolerant control for a
class of uncertain nonlinear systems,” Automatica, vol. 106, pp. 117-123,
-0.8
Aug. 2019.
-1 [10] Y. Man and Y. Liu, “Global adaptive stabilization and practical tracking
0 2 4 6 8 10 12 14 16 18 20 for nonlinear systems with unknown powers,” Automatica, vol. 100: pp.
t(sec) 171-181, Feb. 2019.
[11] R. C. Roman, R. E. Precup and E. M. Petriu, “Hybrid data-driven fuzzy
Fig 12. Trajectories of y and yr . active disturbance rejection control for tower crane systems,” Eur. J.
Control, vol. 58, pp. 373-387, Mar. 2021.
[12] T. Furuhashi, “Fuzzy control stability analysis using a generalized fuzzy
petri net model,” J. Adv. Comput. Intell, vol. 3, no. 2, pp. 99-105, Apr.
displayed in Fig. 6 and Fig. 7, respectively. The time trajectory 1999.
[13] W. Pedrycz and G. Vukovich, “Granular computing in the development
of controller u is shown in Fig. 8. The time trajectories of fuzzy controllers,” Int. J. Intell Syst., vol. 14, no. 4, pp. 419-447, Apr.
of kθ1 k2 and kθ2 k2 are shown in Fig. 9. As can be seen 1999.
from Figs. 6-9, all signals of the closed-loop system remain [14] R. E. Precup and S. Preitl, “Optimisation criteria in development of
fuzzy controllers with dynamics,” Eng. Appl. Artif. Intel., vol. 17, no. 6,
bounded, which demonstrates the correctness of our results. pp. 661-674, Sep. 2004.
Fig. 10 shows the time intervals of triggering events. From Fig. [15] J. Zhang and G. Yang, “Fuzzy adaptive output feedback control of
10, it is easy to get that all time intervals of triggering events uncertain nonlinear systems with prescribed performance,” IEEE Trans.
Cybern., vol. 48, no. 5, pp. 1342-1354, May 2018.
are strictly positive, which means that the Zeno behavior is [16] W. Wang and S. Tong, “Observer-based adaptive fuzzy containment
avoided. Fig. 11 shows that RTETS proposed in [32] and control for multiple uncertain nonlinear systems,” IEEE Trans. Fuzzy
[44] is invalid in our proposed system. Fig. 12 reveals time Syst., vol. 27, no. 11, pp. 2079-2089, Nov. 2019.
[17] Y. Li and G. Yang, “Fuzzy adaptive output feedback fault-tolerant
trajectories of output y with FTETS proposed in [45] and tracking control of a class of uncertain nonlinear systems with non-
STETS, respectively. Both of two control methods work well affine nonlinear faults,” IEEE Trans. Fuzzy Syst., vol. 24, no. 1, pp.
and the STETS just triggers 653 times, while FTETS triggers 223-234, Feb. 2016.
[18] Y. Liu, M. Gong, L. Liu, S. Tong and C. Chen, “Fuzzy observer
884 times (see TABLE I). The results show that STETS constraint based on adaptive control for uncertain nonlinear MIMO
proposed in this paper alleviates the burden of communication systems with time-varying state constraints,” IEEE Trans. Cybern., vol.
effectively. 51, no. 3, pp. 1380-1389, Mar. 2021.
[19] Y. Li, K. Li and S. Tong, “Finite-time adaptive fuzzy output feedback
dynamic surface control for MIMO nonstrict feedback systems,” IEEE
VI. C ONCLUSIONS Trans. Fuzzy Syst., vol. 27, no. 1, pp. 96-110, Jan. 2019.
[20] Y. Liang, Y. Li, W. Che and Z. Hou, “Adaptive fuzzy asymptotic tracking
Adaptive fuzzy event-triggered control for nonstrict nonlin- for nonlinear systems with nonstrict-feedback structure,” IEEE Trans.
ear systems is investigated in this paper. Firstly, for estimating Cybern., vol. 51, no. 2, pp. 853-861, Feb. 2021.

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TFUZZ.2021.3118732, IEEE
Transactions on Fuzzy Systems

10

[21] H. Li, L. Bai, Q. Zhou, R. Lu and L. Wang, “Adaptive fuzzy control of [44] W. Sun, J. Zhao, W. Sun, J. Xia, and Z. Sun, “Adaptive event-triggered
stochastic nonstrict-feedback nonlinear systems with input saturation,” global fast finite-time control for a class of uncertain nonlinear systems,”
IEEE Trans. Syst., Man, Cybern., Syst., vol. 47, no. 8, pp. 2185-2197, Int. J. Robust Nonlin. Control, vol. 30, no. 9, pp. 3773-3785, June 2020.
Aug. 2017. [45] C. Zhang, and G. Yang, “Event-triggered practical finite-time output
[22] W. Sun, S. F. Su, G. Dong and W. Bai, “Reduced adaptive fuzzy tracking feedback stabilization of a class of uncertain nonlinear systems,” Int. J.
control for high-order stochastic nonstrict feedback nonlinear system Robust Nonlin. Control, vol. 29, no. 10, pp. 3078-3092, Mar. 2019.
with full-state constraints,” IEEE Trans. Syst., Man, Cybern., Syst., vol.
51, no. 3, pp. 1496-1506, Mar. 2021.
[23] D. Swaroop, J. K. Hedrick, P. P. Yip and J. C. Gerdes, “Dynamic surface
control for a class of nonlinear systems,” IEEE Trans. Autom. Control,
vol. 45, no. 10, pp. 1893-1899, Oct. 2000.
[24] M. Chen, G. Tao and B. Jiang, “Dynamic surface control using neural
networks for a class of uncertain nonlinear systems with input satura-
tion,” IEEE Trans. Neural Netw. Learn. Syst., vol. 26, no. 9, pp. 2086-
2097, Sep. 2015
[25] G. Sun, X. Ren, Q. Chen and D. Li, “A modified dynamic surface
approach for control of nonlinear systems with unknown input dead
zone,” Int. J. Robust Nonlin. Control, vol. 25, no. 8, pp. 1145-1167,
May 2015.
[26] H. Ma, H. Liang, Q. Zhou and C. K. Ahn, “Adaptive dynamic surface
control design for uncertain nonlinear strict-feedback systems with
unknown control direction and disturbances,” IEEE Trans. Syst., Man,
Cybern., Syst., vol. 49, no. 3, pp. 506-515, Mar. 2019.
[27] Y. Li, K. Li and S. Tong, “Finite-time adaptive fuzzy output feedback
dynamic surface control for MIMO nonstrict feedback systems,” IEEE
Trans. Fuzzy Syst., vol. 27, no. 1, pp. 96-110, Jan. 2019.
[28] T. Zhang, M. Xia and Y. Yi, “Adaptive neural dynamic surface control
of strict-feedback nonlinear systems with full state constraints and
unmodeled dynamics,” Automatica, vol. 81, pp. 232-239, July 2017.
[29] K. Sun, L. Liu, J. Qiu and G. Feng, “Fuzzy adaptive finite-time fault-
tolerant control for strict-feedback nonlinear systems,” IEEE Trans.
Fuzzy Syst., vol. 29, no. 4, pp. 786-796, Apr. 2021.
[30] R. Postoyan, P. Tabuada, D. Nešić and A. Anta, “A framework for the
event-triggered stabilization of nonlinear systems,” IEEE Trans. Autom.
Control, vol. 60, no. 4, pp. 982-996, Apr. 2015.
[31] P. Tallapragada and N. Chopra, “Decentralized event-triggering for
control of nonlinear systems,” IEEE Trans. Autom. Control, vol. 59,
no. 12, pp. 3312-3324, Dec. 2014.
[32] C. Zhang and G. Yang, “Event-triggered global finite-time control for a
class of uncertain nonlinear systems,” IEEE Trans. Autom. Control, vol.
65, no. 3, pp. 1340-1347, Mar. 2020.
[33] Y. Huang and Y. Liu, “Practical tracking via adaptive event-triggered
feedback for uncertain nonlinear systems,” IEEE Trans. Autom. Control,
vol. 64, no. 9, pp. 3920-3927, Sep. 2019.
[34] Y. Huang and Y. Liu, “Switching event-triggered control for a class of
uncertain nonlinear systems,” Automatica, vol. 108, pp. 108471, Oct.
2019.
[35] P. Zhang, T. Liu and Z. Jiang, “Event-triggered stabilization of a class
of nonlinear time-delay systems,” IEEE Trans. Autom. Control, vol. 66,
no. 1, pp. 421-428, Jan. 2021.
[36] Q. Zhu, “Stabilization of stochastic nonlinear delay systems with ex-
ogenous disturbances and the event-triggered feedback control,” IEEE
Trans. Autom. Control, vol. 64, no. 9, pp. 3764-3771, Sep. 2019.
[37] K. Zhang, B. Gharesifard and E. Braverman, “Event-triggered control
for nonlinear time-delay systems,” IEEE Trans. Autom. Control, doi:
10.1109/TAC.2021.3062577.
[38] J. Ma, S. Xu, Q. Ma and Z. Zhang, “Event-triggered adaptive neural
network control for nonstrict-feedback nonlinear time-delay systems
with unknown control directions,” IEEE Trans. Neural Netw. Learn.
Syst., vol. 31, no. 10, pp. 4196-4205, Oct. 2020.
[39] L. Xing, C. Wen, Z. Liu, H. Su and J. Cai, “Event-triggered adaptive
control for a class of uncertain nonlinear systems,” IEEE Trans. Autom.
Control, vol. 62, no. 4, pp. 2071-2076, Apr. 2017.
[40] J. Qiu, K. Sun, T. Wang and H. Gao, “Observer-based fuzzy adap-
tive event-triggered control for pure-feedback nonlinear systems with
prescribed performance,” IEEE Trans. Fuzzy Syst., vol. 27, no. 11, pp.
2152-2162, Nov. 2019.
[41] M. Wang, J. Sun and J. Chen “Input-to-state stability of perturbed non-
linear systems with event-triggered receding horizon control scheme,”
IEEE Trans. Ind. Electron., vol. 66, no. 8, pp. 6393-6403, Aug. 2019.
[42] T. Liu and Z. Jiang, “Event-based control of nonlinear systems with
partial state and output feedback,” Automatica, vol. 53, pp. 10-22, Mar.
2015.
[43] L. Wang and J. Mendel, “Fuzzy basis functions, universal approximation,
and orthogonal least-squares learning,” IEEE Trans. Neural Netw., vol.
3, no. 5, pp. 807-814, Sep. 1992.

1063-6706 (c) 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: VIT University- Chennai Campus. Downloaded on December 26,2021 at 02:18:51 UTC from IEEE Xplore. Restrictions apply.

You might also like