actuators
Article
Data-Based Security Fault Tolerant Iterative Learning Control
Under Denial-of-Service Attacks
Zengwei Li 1 , Changren Zhou 1 , Weiwei Che 1, *, Chao Deng 2 and Xiaozheng Jin 3
Citation: Li, Z.; Zhou, C.; Che, W.;
Deng, C.; Jin, X. Data-Based Security
Fault Tolerant Iterative Learning
Control Under Denial-of-Service
Attacks. Actuators 2022, 11, 178.
https://doi.org/10.3390/
act11070178
Academic Editors: Guanghong Yang,
Bin Jiang and Dan Ye
Received: 7 June 2022
Accepted: 24 June 2022
Published: 26 June 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Actuators 2022, 11, 178. https://doi.org/10.3390/act11070178
1 Shandong Key Laboratory of Industrial Control Technology, School of Automation, Qingdao University,
Qingdao 266071, China; 2020020545@qdu.edu.cn (Z.L.); 2019025912@qdu.edu.cn (C.Z.)
2 Institute of Advanced Technology, Nanjing University of Posts and Telecommunications,
Nanjing 210000, China; dengchao@njupt.edu.cn
3 School of Computer Science and Technology, Qilu University of Technology (Shandong Academy of Sciences),
Jinan 250353, China; xzjin@qlu.edu.cn
* Correspondence: cwwemail1980@qdu.edu.cn;
Abstract: This paper mainly studies the data-based security fault tolerant iterative learning control
(SFTILC) problem of nonlinear networked control systems (NCSs) under sensor failures and denial-of-
service (DoS) attacks. Firstly, the radial basis function neural network (RBFNN) is used to approximate
the sensor failure function and a DoS attack compensation mechanism is proposed in the iterative
domain to lessen the impact of DoS attacks. Then, using the dynamic linearization technology, the
nonlinear system considering failures and network attacks is transformed into a linear data model.
Further, based on the designed linearization model, a new data-based SFTILC algorithm is designed
to ensure the satisfactory tracking performance of the system. This process only uses the input
and output data of the system, and the stability of the system is proved by using the compression
mapping principle. Finally, a digital simulation is used to demonstrate the effectiveness of the
proposed SFTILC algorithm.
Keywords: iterative learning control; data-based; fault tolerant control; DoS attacks
1. Introduction
As the network technology develops, traditional local control systems (LCSs) have
been gradually replaced by network control systems (NCSs) [1–4]. NCSs are integrated by
controllers, sensors, actuators, and network transmission channels. NCSs receive sensor
signals of equipment through networks and send control signals to control the equipment,
so that remote control tasks can be completed accordingly. In NCSs, the separation of
controllers and equipment is one of the most important characteristics of NCSs, but it
requires a large amount of data transmission, and the existence of data also stimulates the
research of data-based control (DDC) [5–7].
It is more and more difficult to identify the model of systems, and the control problem
becomes more difficult and unrealistic by using the traditional model-based control theory.
Therefore, DDC theory began to be researched. There have had many DDC methods in the
existing literature, such as PID control, model free adaptive control (MFAC) [8], iterative
learning control (ILC) [9–12], iterative feedback tuning (IFT) [13] and so on. MFAC and ILC
are proposed to solve the control problem of nonlinear systems by directly applying the I/O
data of the controlled systems. The paper [14] proposed a new MFAC method for discrete-
time single input single output nonlinear systems combined with radial basis function
neural networks. In [15], the bounded input–output stability and tracking performance
of full format dynamic linearized MFAC were strictly proved by using the compression
mapping principle. In [16], aiming at the problem of driverless vehicle ground heading
control, an MFAC method based on the adaptive forgetting factor was proposed. When
the control task is repeated, the system shows the same behavior. For the ILC method
https://www.mdpi.com/journal/actuators
Actuators 2022, 11, 178 2 of 11

of the repeated process, when constructing the current algorithm, the information of the
last iteration time stored by storage devices is used to correct the current control input,
so as to improve the error accuracy of the system cycle process. In [17], the contraction
mapping (CM)-based method was suitable for some robust ILCs that are non-competitive
with uncertainties, especially for power system models. In order to ensure the robust
convergence of iterative learning control designs, the H∞ norm condition was proposed,
and the learning gain matrix was determined by solving norm conditions. In [18], a class of
multi-input and multi-output discrete linear systems was studied. Under the assumption
that all iterative uncertainties were bounded, conditions to ensure the boundedness of all
system trajectories and to ensure the convergence of tracking errors were given in [18].
In the presence of data quantization and DoS attacks, a quantized data-based iterative
learning tracking control algorithm was proposed for a class of repetitive nonlinear systems
in [19].
It is well-known that NCSs involve the transmission of data that are required by the
control system through the network, while many network failures or network attacks exist
in networks. Network attacks can be roughly divided into denial-of-service (DoS) [20–22]
attacks, false data injection (FDI ) [23] attacks, and replay attacks. The basic goal of a
DoS attack is to prevent the target network from providing regular service or resource
access, as well as to cause the target system to become unresponsive or crash. In this
way, NCSs cannot achieve the expected control objectives. Therefore, there has had much
research on solving the impacts of DoS attacks in the existing literature. The consensus
problem of linear multi-agent systems (MASs) under DoS attacks was investigated in [24].
The paper [25] studied the output tracking problem of distributed model-free adaptive
control systems under DoS attacks. If the sensor in an NCS fails, it will also affect the
accuracy of data transmission. Inaccurate data measurement leads to that the controller
can not well generate control inputs to control the controlled system, resulting in the
system can not achieve the expected control objectives. For a class of continuous-time
disturbed systems, the paper [26] provided an adaptive compensation technique against
state-dependent and partly bounded actuator failures and disturbances. In [27], the author
studied the distributed fault tolerance and resilient consensus of heterogeneous multi-agent
systems under equipment failures and DoS attacks. Under DoS attacks, the problem of
fault-tolerant MFAC for nonlinear networked control systems was investigated in [28]. The
above research on fault tolerance is to design fault-tolerant control methods and prove the
stability in the time domain. There are few research findings on ILC fault-tolerant control
under the repeated operation NCSs. Therefore, we investigate the nonlinear repetitive
NCSs under DoS attacks and fault-tolerant control in the iteration domain, which are
prompted by the above descriptions. The contributions of this paper can be listed as:
• The nonlinear system is transformed into a linear data model by using the dynamic
linearization technology under failures and DoS attacks. Then, a new data-based
security fault tolerant iterative learning control (SFTILC) algorithm is designed to
ensure the tracking performance of the system under the DoS attack.
• The radial basis function neural network (RBFNN) is used to approximate the sensor
failure function. A DoS attack compensation mechanism is proposed in the iterative
domain to lessen the impacts of DoS attacks.
The structure of this document is as follows. The preparation and explanation of the
problem are described in Section 2. In Section 3, the design of a fault tolerant controller is
presented. Section 4 provides the stability analysis. The simulations and conclusions are
presented in Sections 5 and 6, respectively.
Actuators 2022, 11, 178 3 of 11

2. Preliminaries and Problem Formulations

2.1. System Description
The nonlinear discrete-time single-input single-output system that repeatedly runs at
finite time intervals is given as follows

yi (k + 1) = f (yi (k), . . . , yi (k − ny ), ui (k ), . . . , ui (k − nu )), (1)

where k ∈ {0, 1, · · · , T } is the sampling time, i ∈ {1, 2, · · · } is the number of iterations,

yi (k) and ui (k ) are the output and input signals of the system, respectively, ny and nu are
two unknown positive integers, and f (· · · ) is an unknown nonlinear scalar function.

2.2. Dynamic Linearization

To facilitate the later controller design, the following assumptions are made for the
system (1).

Assumption 1. The partial derivative of f (· · · ) to ui (k) is continuous.

Assumption 2. System (1) satisfies the generalized Lipschitz condition along the iteration axis,
namely, ∀k ∈ {1, · · · T }, if |∆ui (k )| 6= 0, the formula below applies

|∆yi (k + 1)| ≤ a|∆ui (k)|, ∀i = 1, 2, · · · (2)

where ∆yi (k + 1) = yi (k + 1) − yi−1 (k + 1), ∆ui (k ) = ui (k ) − ui−1 (k) and a > 0 is a constant.

Remark 1. Assumptions 1 and 2 are reasonable in many actual situations. In the control system
design, Assumption 1 is a typical constraint condition for generic nonlinear systems. Assumption 2
limits the system output change rate produced by the change of the control input along the iteration
axis. Namely, a bounded input energy change should result in a finite output energy change.

Lemma 1. For the nonlinear system (1) that meets both Assumptions 1 and 2, when |∆ui (k)| 6= 0,
there is a time-varying pseudo partial derivative (PPD) φi (k ) so that system (1) can be transformed
into the following linear data model

∆yi (k + 1) = φi (k)∆ui (k), ∀k ∈ {0, 1, · · · , N }, i = 1, 2, · · · (3)

where |φi (k)| ≤ a. The proof of Lemma 1 is the same as that in [29].

3. Design Fault Tolerant Controller

3.1. Model Free Adaptive Iterative Learning Control
Considering the input cost function as follows

ζ (ui (k)) = |yd (k + 1) − yi (k + 1)|2 + λ(∆ui (k))2 , (4)

where λ is a weight coefficient and yd (k + 1) is the time-invariant desired output signal.

The optimal control rate is obtained by minimizing the cost function ζ (ui (k)). In
order to make the algorithm more general, the step factor ρ ∈ (0, 1] is introduced and the
following optimal control algorithm is obtained

ρφi (k)
u i ( k ) = u i −1 ( k ) + (y (k + 1) − yi−1 (k + 1)). (5)
λ + φi (k)2 d

Because φi (k) are difficult to be measured in real systems, we need to estimate them
by introducing the following performance function

ζ (φ̂i (k)) = |∆yi−1 (k + 1) − φ̂i (k)∆ui−1 (k)|2 + µ|φ̂i (k) − φ̂i−1 (k )|2 (6)
Actuators 2022, 11, 178 4 of 11

where φ̂i (k ) is an estimate of φi (k), and µ is a weight factor.

Minimize the performance function ζ (φ̂i (k)) to get φ̂i (k), similarly, in order to make
the algorithm more general the step-size factor µ ∈ (0, 1] is introduced, then one can obtain

η∆ui−1 (k)
φ̂i (k) = φ̂i−1 (k) + (∆yi−1 (k + 1) − ∆ui−1 (k)φ̂i−1 (k)). (7)
µ + ∆u2i−1 (k)
The above control algorithm is applied to the nominal system. However, in the
practical networked control system, on the one hand, sensor failures will result in some
deviations in the measurement data. On the other hand, the emergence of the network
may bring cyber attacks. So, the above algorithm may not complete the tracking task well
when these problems exist. Therefore, it seems necessary to design the secure fault-tolerant
iterative learning control (SFTILC) algorithm under sensor failures and cyber attacks.
Figure 1 depicts the nonlinear NCS architecture under DoS attacks and sensor fail-
ures consideration.

Figure 1. System block diagram.

3.2. Sensor Failure

According to Figure 1, when a fault occurs, the sensor measurement signal becomes to
the following form
yiF (k) = yi (k ) + Fs,i (k ) (8)
where Fs,i (k) is a fault function, which can be approximated using the radial basis function
neural network (RBFNN), the details are given as

Fs,i (k) = viT (k )ψ(z(k)) (9)

where vi (k) ∈ R p is the ideal weight vector in the output layer, p is the RBFNN’s node
number, and ψ(z(k)) = [ψ1 (z(k )), ψ2 (z(k)) · · · , ψ p (z(k))] T ∈ R p is the hidden layer’s basis
function, which is a Gaussian function, and it can be represented as

k z ( k ) − $ l k2
ψl (z(k)) = exp(− ) (10)
τl2

where $l denotes the RBFNN’s center, τl denotes the RBFNN’s breadth for l ∈ 1, 2, · · · , p,
z(k ) = [ur (k ), Fs,i (k − 1), yi (k)] T is the RBFNN input.
In most cases, vi (k ) is not available, therefore, the following learning process is intro-
duced to obtain it.
vi (k) = v {∆ui (k)ψ T (k )φ̂i (k)} (11)
where v stands for the projection operator, ∆ui (k ) can be calculated by (18), φ̂i (k) can be
defined in (17). Then v is given as following, which has been employed in [30].
Actuators 2022, 11, 178 5 of 11



 0, i f vi (k) = v and ∆ui (k)ψ T (k)φ̂i (k) < 0

v (·) = 0, i f vi (k) = v and ∆ui (k)ψ T (k)φ̂i (k) > 0 (12)

 ∆u (k)ψ T (k)φ̂ (k),

otherwise.
i i

where v and v are the lower and the upper bounds of vi (k), respectively.
In the case of sensor failures, the nonlinear repetitive system is linearized dynami-
cally by
yiF (k + 1) = yi−1 (k + 1) + φi (k)∆ui (k ) + Fs,i (k + 1) (13)
√
where | Fs,i (k + 1)| ≤ pv̄ with p being the node number of RBFNN and v is the upper
bounds of vi (k).

3.3. Attack Compensation

When a network attacker launches DoS attacks, the output signal is expressed as follows
F
y ai (k) = αi (k)yiF (k), (14)

where αi (k) obeys the Bernoulli distribution, αi (k) = 1 denotes a failed DoS attack, whereas
αi (k) = 0 denotes a successful DoS attack. As a result, the probabilities are expressed as
(
Pr [αi (k) = 1] = ᾱ,
Pr (15)
Pr [αi (k) = 0] = 1 − ᾱ.
To compensate for the impact of DoS attacks, the mechanism is designed as follows

yi (k) = αi (k)yiF (k) + (1 − αi (k))yiF−1 (k). (16)

Consider sensor failure and DoS attacks, the SFTILC algorithm is proposed as follows

if ||φ̂i (k)|| ≤ e or |∆ui (k)| ≤ e



 φ̂1 (k),



φ̂i (k ) = η∆u (k) (17)

 φ̂i−1 (k) + µ+∆ui−2 1 (k) yi−1 (k + 1)

 i −1
−yi−2 (k + 1) − ∆ui−1 (k)φ̂i−1 (k) , otherwise.
 F

φi (k )
u i ( k ) = u i −1 ( k ) + (y (k + 1) − yi−1 (k)). (18)
λ + φi (k)2 d
For the convenience, the tracking error ei (k) is defined as follows:

ei ( k ) = y d ( k ) − y i ( k ). (19)

Problem 1. This study aims to design the SFTILC algorithm to realize the mean-square bounded
output tracking task of system (1) under the influence of sensor failures and DoS attacks, namely,

{ei (k) | E{|ei (k)|} ≤ b},

where b is a constant, where E{·} denotes the expectation.

4. Stability Analysis
ρ2 a2
Theorem 1. If there are parameters µ > 0, ρ ∈ (0, 1], λ > 4 , and η ∈ (0, 1] such that the
SFTILC algorithm constructed of (17) and (18) holds for system (1) subject to sensor failures and
DoS attacks, then Problem 1 is solved, where a is the same as described in (2).

Proof. The proof of Theorem 1 consists of two parts: the bounded PPD parameter estima-
tion and the bounded result of the output tracking error.
Actuators 2022, 11, 178 6 of 11

Part I: Boundedness of the PPD parameter estimation

(i) When |φ̂i (k)| ≤ ε, φ̂i (k) = φ̂1 (k ), so φ̂i (k ) is bounded.
(ii) When |φ̂i (k)| > ε, the PPD estimation error is defined as

eφi (k) = φ̂i (k) − φi (k). (20)

Bring (17) into (20) to obtain the following formula

η∆ui−1 (k)  
eφi (k ) = φ̂i−1 (k) − φi (k ) + y a,i−1 (k + 1) − yiF−2 (k + 1) − ∆ui−1 (k)φ̂i−1 (k)
µ + ∆ui−1 (k )
2

η∆ui−1 (k) 
= eφi−1 (k)− ∆φi (k) + αi−1 (k + 1)yiF−1 (k + 1) − yiF−2 (k + 1)
µ + ∆u2i−1 (k)

+ (1 − αi−1 (k + 1))yiF−2 (k + 1) − ∆ui−1 (k)φ̂i−1 (k)
η∆u2i−1 (k) (21)
= (1 − )eφi−1 (k)− ∆φi (k)
µ + ∆u2i−1 (k )
η∆ui−1 (k)
+ (αi−1 (k + 1))∆Fs,i−1 (k + 1)
µ + ∆u2i−1 (k )
η∆u2i−1 (k)
+ (αi−1 (k + 1)− 1)φi−1 (k).
µ + ∆u2i−1 (k )

Taking the absolute value and expectation on both sides of (21), we can get

η∆u2i−1 (k) ᾱη∆ui−1 (k)

E{|eφi (k)|} ≤ |1 − | E{|eφi−1 (k)|}+|∆φi (k)| + | ∆F (k + 1)|
µ + ∆u2i−1 (k ) µ + ∆ui−1 (k) s,i−1
(22)
η∆u2i−1 (k)
+| (ᾱ − 1)φi−1 (k)|
µ + ∆u2i−1 (k)

where |φi (k)| ≤ a, 0 < ᾱ < 1 and µ + ∆u2i−1 (k) > ∆u2i−1 (k) > 0, so

η∆ui−1 (k) η
0< < √ (23)
µ + ∆u2i−1 (k) 2 µ

η (ᾱ − 1)∆u2i−1 (k ) η∆u2i−1 (k )

0<| |< ≤ m1 < 1 (24)
µ + ∆u2i−1 (k) µ + ∆u2i−1 (k)

η∆u2i−1 (k)
0 < 1− ≤ m2 < 1 (25)
µ + ∆u2i−1 (k)
where m1 and m2 are constants.
From (23)–(25), we can reduce (22) to the following inequality

E{|eφi (k)|} ≤ m2 E{|eφi−1 (k)|} + m3 , (26)

√
ᾱη pv̄
where m3 = + 3a.
√
µ
By using the compressed mapping principle, we can get

E{|eφi (k)|} ≤ m22 E{|eφi−2 (k )|} + m2 m3 + m3

≤ ···
≤ m2i−1 E{|eφ1 (k)|} + m2i−2 m3 + · · · + m3 (27)
m3 (1 − m2i−1 )
= m2i−1 E{|eφ1 (k)|} + ,
1 − m2
Actuators 2022, 11, 178 7 of 11

so, eφi (k) converges to the following set


 m3
{eφi (k) E{|eφi (k)|} ≤ }. (28)
1 − m2

So far, the proof of the boundedness of eφi (k ) is completed because φi (k) is bounded,
leading to φ̂i (k) also being bounded.
Part II: Tracking performance analysis
The boundedness of the PPD parameter estimation proved above can be used as a
condition to prove the boundedness of the system tracking error.
By taking (3) and (18) into (19), we can get

ei (k + 1) = yd (k + 1) − yi−1 (k + 1) − φi (k)∆ui (k)

ρφi (k)φ̂i (k)
= ei−1 (k + 1)−
λ + φ̂i2 (k)

× yd (k + 1)+ yiF−1 (k + 1)− yiF−1 (k + 1)− αi−1 (k + 1)yiF−1 (k + 1)
(29)

−(1 − αi−1 (k + 1))yiF−2 (k + 1)
ρφi (k)φ̂i (k) ρφ (k)φi−1 (k)φ̂i (k)
= (1 − ) ei −1 ( k + 1 ) − i (1 − αi−1 (k + 1))∆ui−1 (k)
λ + φ̂i2 (k) λ + φ̂i2 (k)
ρφi (k)φ̂i (k)
− ((1 − αi−1 (k + 1))∆Fs,i−1 (k + 1) + Fs,i−1 (k + 1)).
λ + φ̂i2 (k)

Taking the absolute value and expectation on both sides of (29), we can get

ρφi (k)φ̂i (k)

E{|ei (k + 1)|} ≤ |1 − | E{|ei−1 (k + 1)|}
λ + φ̂i2 (k )
ρφi (k)φi−1 (k )φ̂i (k)
+| (1 − ᾱ)∆ui−1 (k)| (30)
λ + φ̂i2 (k)
ρφi (k)φ̂i (k )
+| ((1 − ᾱ)∆Fs,i−1 (k + 1) + Fs,i−1 (k + 1))|.
λ + φ̂i2 (k)

According to Theorem 1, (3) and (28), we can get

ρφi (k )φ̂i (k) ρa

0 < m4 ≤ 2
≤ √ = m5 < 1, (31)
λ + φ̂i (k) 2 λ

ρφi (k)φ̂i (k )
0 < 1− ≤ m6 < 1, (32)
λ + φ̂i2 (k)
ρφi (k)φi−1 (k)φ̂i (k )
0< ≤ am5 < a, (33)
λ + φ̂i2 (k )
where m4 , m5 , and m6 are constants.
In addition, generally because the energy of the equipment cannot be unlimited,
the input increment ∆ui−1 (k ) is similarly bounded with |∆ui−1 (k )| ≤ σ. From (31)–(33)
and (13), we can reduce (30) to the following inequality
√
E{|ei (k + 1)|} ≤ m6 | E{|ei−1 (k + 1)|} + | a(1 − ᾱ)∆ui−1 (k )| + (2 − ᾱ) pv̄, (34)

E{|ei (k + 1)|} ≤ m6 | E{|ei−1 (k + 1)|} + m7 , (35)

Actuators 2022, 11, 178 8 of 11

√
where m7 = a(1 − ᾱ)σ + (2 − ᾱ) pv̄. Further, by using the principle of the compressed
mapping, we can get

E{|ei (k + 1)|} ≤ m26 | E{|ei−2 (k + 1)|} + m6 m7 + m7

≤ ···
≤ m6i−1 | E{|e1 (k + 1)|} + m6i−2 m7 + · · · + m7 (36)
m7 (1 − m6i−1 )
= m6i−1 | E{|e1 (k + 1)|} +
1 − m6

so, ei (k + 1) converges to the following set


 m7
{ei (k + 1) E{|ei (k + 1)|} ≤ }. (37)
1 − m6

Now, the proof of the boundedness of the tracking error is completed.

5. Simulation Results
The nonlinear repetitive operation system described below is used

yi ( k ) k
y i ( k + 1) = 2
+ (ui (k) + 0.01round( )sin(yi (k)))3 + Fs,i (k).
1 + yi ( k ) 500
The desired output is given as
 t
0.5(−1)round( 100 ) , 0 ≤ t ≤ 500,
yd (k ) =
0.5sin( tπ ) + 0.3cos( tπ ), 501 ≤ t ≤ 1000.
100 50
The control input and control output have their initial values with u1 (1) = u1 (2) = 0.5
and y1 (1) = y1 (2) = y1 (3) = 0.5, respectively, and the estimated values of the PPD
parameters are set to φ̂1 (1) = φ̂1 (2) = 0.5. The controller parameters are selected as
η = 1, µ = 1, ρ = 0.6, λ = 20 and ε = 10−5 . The successful transmission rate is selected to
be ᾱ = 0.5.
Consider the actual sensor fault function as
Fs,i (k − 1)
Fs,i (k) = 2u2r (k) + 2 ( k − 1)
,
1 + Fs,i

k
where ur (k) = sin( 1000 ). In addition, the initial value of the hazard function is set to
Fs,i (1) = 0.
In this paper, the learning rate of the RBFNN is set to χ = 0.05, the RBFNN input
defined as z(k) = [ur (k ), Fs,i (k − 1), yi (k)] T . The initial value of the Gaussian function is set
 
0 0 0 0 0  T
to $0 = and τ0 = 5 5 5 5 5 .
0 0 0 0 0
 T
The input layer weight vector is set to v̄i = 0.6294 0.8116 − 0.7460 0.8268 0.4574 .
The Bernoulli distribution is shown in Figure 2 for a random sequence of DoS attacks.
Figure 3 shows the simulation diagram approaching the sensor fault function by RBFNN.
From Figure 3, we can clearly see the superiority of the approximation performance of the
RBFNN. In Figure 4a–c, it shows the performance of for different iterations.
Figure 5 shows the comparison of the maximum tracking error with the number of
times of iterations by using the SFTILC algorithm and the traditional MFAILC [28] under
DoS attacks. In Figure 5, the SFTILC can make the tracking error converge to a smaller
bound. Comparative results of tracking the performance of the system are depicted in
Figure 6. From Figure 6, we can see that the tracking of the desired trajectory with the
SFTILC algorithm is obviously better than that of the traditional MFAILC.
Actuators 2022, 11, 178 9 of 11

According to simulation results, the SFTILC algorithm can handle the output tracking
tasks well when taking sensor failure and DoS attacks into account.

DoS attack sequence for the 100th iteration

0
0 100 200 300 400 500 600 700 800 900 1000
k

Figure 2. DoS attack sequence for the 100th iteration.

2
Approximation of F (k)
s
F (k)
s

1.5
Approximation performance

0.5

-0.5
0 100 200 300 400 500 600 700 800 900 1000
Sampling time k

Figure 3. Sensor fault estimation.

0.6 1.5 0.6

yd (k) yd (k) yd (k)
y(k) y(k) y(k)
0.4 0.4
1
iterative tracking performance

iterative tracking performance

iterative tracking performance

0.2 0.2

0.5
0 0

-0.2 -0.2
0

-0.4 -0.4

-0.5
-0.6 -0.6

-0.8 -1 -0.8
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
sampling time sampling time sampling time

(a) (b) (c)

Figure 4. Performance for different iterations. (a) The tracking performance for the 10th iteration. (b) The
tracking performance for the 50th iteration. (c) The tracking performance for the 100th iteration.

1.5
SFTILC

1
max error update

0.5

0
0 20 40 60 80 100 120 140 160 180 200
iteration times

Figure 5. Maximum tracking error.

Actuators 2022, 11, 178 10 of 11

0.8
y (k)
d
SFTILC
0.6

(k)
0.4

200
iterative tracking performance y
0.2

-0.2

-0.4

-0.6

-0.8
0 100 200 300 400 500 600 700 800 900 1000
sampling time

Figure 6. Tracking performance.

6. Conclusions
The data-based SFTILC problem of nonlinear networked control systems has been
studied in this paper with sensor failures and DoS attacks. Firstly, the radial basis function
neural network has been used to approximate the sensor failure function and a DoS attack
compensation mechanism has been proposed in iterative domain to lessen the impact of
the DoS attack. Then, using the dynamic linearization technology, the nonlinear system
considering fault and network attack has been transformed into a linear data model. Further,
a new data-based SFTILC algorithm has been designed to ensure the tracking performance
of the system. Finally, a digital simulation has been used to demonstrate the effectiveness
of the proposed SFTILC algorithm. In the future, we will consider the partial form dynamic
linearization-based model-free adaptive algorithm that uses more input information to
improve the control performance of the system.

Author Contributions: Writing—original draft preparation and software, Z.L.; writing—review and
editing, C.Z.; writing—review and editing, supervision, W.C.; investigation and supervision, C.D.;
methodology, validation and software, X.J. All authors have read and agreed to the published version
of the manuscript.
Funding: This work is supported by the National Natural Science Foundation of China (Grant
No. 61873338), the Taishan Scholars (No. tsqn201812052), and the Natural Science Foundation of
Shandong Province (ZR2020KF034).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Gaid, M.; Cela, A.; Hamam, Y. Optimal integrated control and scheduling of networked control systems with communication
constraints: Application to a car suspension system. IEEE Trans. Control Syst. Technol. 2006, 14, 776–787. [CrossRef]
2. Shi, Y.; Yu, B. Output feedback stabilization of networked control systems with random delays modeled by markov chains. IEEE
Trans. Autom. Control 2009, 54, 1668–1674.
3. Xu, H.; Jagannathan, S. Neural network-based finite horizon stochastic optimal control design for nonlinear networked control
systems. IEEE Trans. Neural Netw. Learn. Syst. 2015, 26, 472–485. [CrossRef] [PubMed]
4. Qiu, J.; Gao, H.; Ding, S. Recent advances on fuzzy-model-based nonlinear networked control systems: A survey. IEEE Trans. Ind.
Electron. 2016, 63, 1207–1217. [CrossRef]
5. Guardabassi, G.; Savaresi, S. Virtual reference direct design method: An off-line approach to data-based control system design.
IEEE Trans. Autom. Control 2000, 45, 954–959. [CrossRef]
6. Mahish, P.; Pradhan, A. Mitigating subsynchronous resonance using synchrophasor data based control of wind farms. IEEE Trans.
Power Deliv. 2020, 35, 364–376. [CrossRef]
7. Deng, C.; Wen, C.; Huang, J.; Zhang, X.; Zou, Y. Distributed observer-based cooperative control approach for uncertain nonlinear
MASs under event-triggered communication. IEEE Trans. Autom. Control 2021, 67, 2669–2676. [CrossRef]
Actuators 2022, 11, 178 11 of 11

8. Bu, X.; Hou, Z.; Zhang, H. Data-driven multiagent systems consensus tracking using model free adaptive control. IEEE Trans.
Neural Netw. Learn. Syst. 2017, 29, 1514–1524. [CrossRef]
9. Meng, D.; Zhang, J. Design and analysis of data-driven learning control: An optimization-bBased approach. systems and control.
IEEE Trans. Neural Netw. Learn. Syst. 2021, 1–15. [CrossRef]
10. Jin, X. Nonrepetitive trajectory tracking for nonlinear autonomous agents with asymmetric output constraints using parametric
iterative learning control. Int. J. Robust Nonlinear Control 2019, 29, 1941–1955. [CrossRef]
11. Radac, M.; Lala, T. Hierarchical Cognitive Control for Unknown Dynamic Systems Tracking. Mathematics 2021, 9, 2752. [CrossRef]
12. Memon, F.; Shao, C. Data-driven optimal PID type ILC for a class of nonlinear batch process. Int. J. Syst. Sci. 2020, 52, 263–276.
[CrossRef]
13. Hildebrand, R.; Lecchini, A.; Solari, G.; Gevers, M. Optimal prefiltering in iterative feedback tuning. IEEE Trans. Autom. Control
2005, 50, 1196–1200. [CrossRef]
14. Zhu, Y.; Hou, Z. Data-driven MFAC for a class of discrete-time nonlinear systems with RBFNN. IEEE Trans. Neural Netw. Learn.
Syst. 2014, 25, 1013–1020. [CrossRef]
15. Hou, Z.; Chi, R.; Gao, H. An overview of dynamic-linearization-based data-driven control and applications. IEEE Trans. Ind.
Electron. 2017, 64, 4076–4090. [CrossRef]
16. Liao, Y.; Du, T.; Jiang, Q. Model-free adaptive control method with variable forgetting factor for unmanned surface vehicle control.
Appl. Ocean. Res. 2019, 93, 1141–1187. [CrossRef]
17. Meng, D. Convergence conditions for solving robust iterative learning control problems under nonrepetitive model uncertainties.
IEEE Trans. Neural Netw. Learn. Syst. 2019, 30, 1908–1919. [CrossRef]
18. Meng, D.; Moore, K. Robust iterative learning control for nonrepetitive uncertain systems. IEEE Trans. Autom. Control 2017, 62,
907–913. [CrossRef]
19. Zhou, C.; Che, W. Quantized data-based iterative learning control under denial-of-service attacks. Optim. Control. Appl. Methods
2021. [CrossRef]
20. Zhang, X.; Han, Q.; Ge, X.; Ding, L. Resilient control design based on a sampled-data model for a class of networked control
systems under denial-of-service attacks. IEEE Trans. Cybern. 2020, 50, 3616–3626. [CrossRef]
21. Liu, J.; Suo, W.; Zha, L.; Tian, E.; Xie, X. Security distributed state estimation for nonlinear networked systems against DoS attacks.
Int. J. Robust Nonlinear Control 2020, 30, 1156–1180. [CrossRef]
22. Li, J.; Yang, Z.; Mu, X.; Wu, X. Passivity-based Event-triggered Fault Tolerant Control for Nonlinear Networked Control System
with Actuator Failures and DoS Jamming Attacks. J. Frankl. Inst. 2020, 357, 9288–9307. [CrossRef]
23. Hu, L.; Wang, Z.; Han, Q.; Liu, X. State estimation under false data injection attacks: Security analysis and system protection.
Automatica 2018, 87, 176–183. [CrossRef]
24. Ma, Y.; Che, W.; Deng, C.; Wu, Z. Observer-based event-triggered containment control for MASs under DoS attacks. IEEE Trans.
Cybern. 2021, 1–12. [CrossRef] [PubMed]
25. Ma, Y.; Che, W.; Deng, C.; Wu, Z. Distributed model-free adaptive control for learning nonlinear MASs under DoS attacks. IEEE
Trans. Neural Netw. Learn. Syst. 2021, 1–10. [CrossRef]
26. Jin, X.; Che, W.; Wu, Z.; Wang, H. Analog control circuit designs for a class of continuous-time adaptive fault-tolerant control
systems. IEEE Trans. Cybern. 2020, 52 , 4209–4220. [CrossRef]
27. Deng, C.; Jin, X.; Che, W.; Wang, H. Learning-based distributed resilient fault-tolerant control method for heterogeneous MASs
under unknown leader dynamic. IEEE Trans. Neural Netw. Learn. Syst. 2021, 1–10. [CrossRef]
28. Su, M.; Che, W. Fault-Tolerant control for model-free networked control systems under DoS attacks. J. Frankl. Inst. 2021, 358,
9023–9033. [CrossRef]
29. Xing, J.; Li, N.; Chi, R.; Huang, B.; Hou, Z. Data-driven nonlinear ILC with varying trial lengths. J. Frankl. Inst. 2020, 357,
10262–10287. [CrossRef]
30. Lian, J.; Hu, J.; Żak, S. Variable Neural Adaptive Robust Control: A Switched System Approach. IEEE Trans. Neural Netw. Learn.
Syst. 2015, 26, 903–915. [CrossRef]