Received May 14, 2020, accepted May 26, 2020, date of publication May 28, 2020, date of current

version June 8, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.2998359

Function Projective Synchronization of Complex

Networks With Distributed Delays via Hybrid
Feedback Control
XIULIANG QIU1 , WENSHUI LIN 2,3 , (Member, IEEE), AND YIMING ZHENG3
1 Chengyi University College, Jimei University, Xiamen 361021, China
2 Fujian Key Laboratory of Sensing and Computing for Smart City, Xiamen 361005, China
3 School of Informatics, Xiamen University, Xiamen 361005, China

Corresponding author: Wenshui Lin ([email protected])

This work was supported in part by NSFC under Grant 11771362, and in part by the Scientific Research Fund of Fujian Provincial
Department of Education under Grant JT180870.

ABSTRACT Due to its applications, some types of synchronization of complex networks have been
intensively investigated. In particular, as a more general type of synchronization, function projective
synchronization (FPS) has been investigated for complex networks with time delay or with time-varying
delay. In this paper, we investigate FPS of complex networks with distributed delays. It is proven that, FPS
of such networks can be realized via hybrid feedback control. Finally, two numerical examples are provided
to illustrate the effectiveness of the proposed method.

INDEX TERMS Function projective synchronization, complex networks, distributed delays, hybrid feed-
back control.

I. INTRODUCTION Complex networks usually have propagation delays, which

Complex networks attract a lot of research interests, because have been observed in lasers, neuron models, electronic cir-
they are widely applicable to various fields [1]–[5]. Syn- cuits and so on. Many phenomena in the real world indicate
chronization is one of the most important and interesting that the current state of a node is affected by those of its
collective behaviors in complex networks. Due to its potential neighbors in the previous period. Therefore complex net-
engineering applications (see [6]–[14]), the types of complete works with distributed delays were introduced into the model
synchronization [6]–[8], cluster synchronization [9]–[11], lag system [21], [22]. In order to serve the practical life better,
synchronization [7], projective synchronization [12], [13], complex networks with distributed time delays are worthy of
and function projective synchronization (FPS) [14], [15] were serious investigation. Some types of synchronization of such
intensively investigated in the past few decades. networks have been investigated [23]–[25]. However, to our
FPS means synchronizing the driver and response systems best knowledge, up to now there are no works concerning FPS
up to a scaling function. Hence it is a more general type for these networks.
of chaotic synchronization, and encompasses complete syn- In this paper we investigate the problem of FPS for gen-
chronization and projective synchronization. For a complex eral complex networks with distributed delays. In Section II
network, FPS means with a desired scaling function all the we introduce the network model, and prove that FPS of
nodes are synchronized to an equilibrium point or a periodic such networks can be realized via hybrid feedback control.
orbit. Because the scaling function is hard to predict, FPS In Section III, two numerical examples are provided to illus-
has application in secure communication [16]. Hence FPS trate the effectiveness of our method.
has attracted a lot of research interests [17]–[20]. In partic-
ular, FPS was investigated for complex networks with time
II. FPS OF COMPLEX NETWORKS WITH DISTRIBUTED
delay [19] and those with time-varying delay [20].
DELAYS
Consider a general complex network with N identical nodes
The associate editor coordinating the review of this manuscript and v1 , v2 , · · · , vN , which are linearly coupled. The network
approving it for publication was Hocine Cherifi . topology is represented by the matrix G = (gij ) ∈ RN ×N ,

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
99110 VOLUME 8, 2020
X. Qiu et al.: FPS of Complex Networks With Distributed Delays via Hybrid Feedback Control

where for i 6 = j, gij 6 = 0 if vi is connected to vj , and 0 Construct the Lyapunov function

PN
otherwise; And gii , 1 ≤ i ≤ N , is defined as − gij . N N
1X T 1X 1
j=1,j6=i V (t) = ei (t)ei (t) + (di (t) − d ∗ )2
xi = (xi1 , xi2 , · · · , xin )T ∈ Rn will denote the state of vi , 2 2 ki
i=1  i=1
and the behavior of nodes can be described by a continuously Z t N Z +∞
1 X
differentiable vector function f : Rn → Rn . Let ui (t) ∈ Rn +  K (u)eTj (w)du
be the control input. Then a controlled network with dis- 2 t−u 0
j=1
tributed delays can be described by the following system of Z +∞ 
integro-differential equations: · K (u)ej (w)du dw,
0
N t d∗ R+
Z
X where ∈ is to be determined.
ẋi (t) = f (xi (t)) + gij K (t − s)xj (s)ds + ui (t). (1) Consequently, we have
j=1 −∞

V̇ (t)
Assumption 1: The kernel function K (·) in (1) is a real N N
valued nonnegative 1
R +∞ continuous function defined on [0, +∞)
X X
= eTi (t)ėi (t) + (di (t) − d ∗ )ḋi (t)
and satisfies 0 K (s)ds = 1. ki
i=1 i=1
Definition 1 (FPS): The network (1) is said to achieve func- N Z
1 X +∞
Z +∞
tion projective synchronization if there exists a continuously + K (u)eTj (t)du K (u)ej (t)du
differentiable scaling function α(t) such that 2
j=1 0 0
N Z
1 X +∞
Z +∞
lim kei (t)k = lim kxi (t) − α(t)s(t)k = 0, − T
K (u)ej (t − u)du K (u)ej (t − u)du
t→+∞ t→+∞
2
j=1 0 0
where s(t) ∈ Rn
is an equilibrium point, a periodic orbit, or an N

N Z +∞

orbit of a chaotic attractor, which satisfies ṡ(t) = f (s(t)).
X X
= eTi (t)  gij K (u)ej (t − u)du − di (t)ei (t)
Based on the following result presented in [26], we are able i=1 j=1 0
to prove that, FPS of network (1) can be realized via a hybrid N N Z
feedback control.
X
∗ T 1 X +∞
+ (di (t) − d )ei (t)ei (t) + K (u)eTj (t)du
Lemma 1: For any vectors x, y ∈ Rn and positive definite 2 0
i=1 j=1
matrix Q ∈ Rn×n , the following matrix inequality holds: Z +∞ N Z +∞
1
2x T y ≤ x T Qx + yT Q−1 y.
X
· K (u)ej (t)du − K (u)eTj (t − u)du
Theorem 2: For any given initial conditions xi (0), di (0), and 0 2
j=1 0
s(0), FPS of network (1) can be realized by the control law: Z +∞
ui (t) = ui1 (t) + ui2 (t), i = 1, 2, · · · , N , where · K (u)ej (t − u)du
0
N X
N Z +∞ N
ui1 (t) = α(t)f (s(t)) + α̇(t)s(t) − f (xi (t)), (2)
X X
= eTi (t)gij K (u)ej (t − u)du − d ∗ eTi (t)
ui2 (t) = −di (t)ei (t), (3) i=1 j=1 0 i=1
ḋi (t) = ki eTi (t)ei (t), (4) 1
N Z +∞
X Z +∞
·ei (t) + K (u)eTj (t)du K (u)ej (t)du
and ki > 0 is an any constant. 2 0 0
j=1
Proof: Let ei (t) = xi (t) − α(t)s(t), i = 1, 2, · · · , N . It 1
N Z +∞
X Z +∞
follows from (1) that − K (u)eTj (t − u)du K (u)ej (t − u)du.
2 0 0
j=1
N
X Z t N
ėi (t) = f (xi (t)) + gij K (t − s)ej (s)ds Let Q = G In (the Kronecker product of G and In ),
j=1 −∞ and h(t)R = (h1 (t), h2 (t), · · · , hN (t))T ∈ Rn×N , where
+∞
+ui (t) − α̇(t)s(t) − α(t)f (s(t)). hi (t) = 0 K (u)eTi (t − u)du, i = 1, 2, · · · , N . Let e(t) =
(eT1 (t), eT2 (t), · · · , eTN (t))T ∈ Rn×N . By Lemma 1 we have
Combining with (2) − (4), we have
V̇ (t) ≤ eT (t)Qh(t) − d ∗ eT (t)e(t)
N Z t 1 1
+ eT (t)e(t) − hT (t)h(t)
X
ėi (t) = gij K (t − s)ej (s)ds − di (t)ei (t).
−∞
2 2
j=1 1 1
≤ eT (t)QQT e(t) + hT (t)h(t) − d ∗ eT (t)e(t)
Let s = t − u. Then 2 2
1 T 1 T
+ e (t)e(t) − h (t)h(t)
N +∞  2 2
Z
QQT
X  
ėi (t) = gij K (u)ej (t − u)du − di (t)ei (t). (5) 1
0 ≤ ρ ∗
+ − d eT (t)e(t),
j=1 2 2
VOLUME 8, 2020 99111
 X. Qiu et al.: FPS of Complex Networks With Distributed Delays via Hybrid Feedback Control

FIGURE 1. The chaotic attractor of the Lorenz system.

FIGURE 2. The state phases of the Lorenz system.

where ρ(M ) is the greatest

 Teigenvalue
 of a symmetric matrix
M . Hence, if d ∗ = ρ QQ 2 + 3
2 , then V̇ (t) ≤ −eT (t)e(t).
Therefore, the error system (5) is asymptotically stable from
the Lyapunov stability theory. This completes the proof. 
III. COMPUTER SIMULATIONS
In this section, two simulation examples will be employed
to illustrate the theoretical result obtained in the previous
section.
Simulation 1: Consider the following single Lorenz
FIGURE 3. The synchronization errors of (6).
system:

ẋ1 = a(x2 − x1 )

ẋ2 = (b − x3 )x1 − x2 i = 1, 2, 3. Choose K (x) = e−x as the kernel. By using

ẋ3 = x1 x2 − cx3

Theorem 2 the controllers can be designed as ui (t) = ui1 (t) +
ui2 (t), i = 1, 2, 3, where
where a = 10, b = 28, and c = 38 . The chaotic attractor of
the system is depited in Figure 1.  10(ei1 (t)−ei2 (t))+α̇(t)x1 (t) 
We choose the coupling configuration matrix as ei2 (t)−28ei1 (t)+xi1 (t)xi3 (t)−α(t)x1 (t)x3 (t)+α̇(t)x2 (t) ,
ui1 (t) =
3 ei3 (t)−xi1 (t)xi2 (t)+α(t)x1 (t)x2 (t)+α̇(t)x3 (t)
  8
−1 0 1
0 .
 
G= 1 −1 −di ei1 (t)
0 1 −1 ui2 (t) = −di ei2 (t) , ḋi (t) = ki eTi (t)ei (t).
Then the controlled network with distributed delays is −di ei3 (t)
described as
   10(xi2 (t) − xi1 (t))
 In this numerical simulation, we take the initial states
ẋi1 (t)
ẋi2 (t) = (28 − xi3 (t))xi1 (t) − xi2 (t) as x1 (0) = (3, 4, −4)T , x2 (0) = (4, 1, −4)T , x3 (0) =
 8  (−2, 0, 5)T , and x(0) = (5, −3, 5)T . Set k1 = k2 = k3 = 1,
ẋi3 (t) xi1 (t)xi2 (t) − xi3 (t)
3 d(0) = [5, 0, 3]T , and the scaling function α(t) = sin 2πt
10 +4.
X 3 Z t Then Figures 2 and 3 display the numerical results. In partic-
+ gij K (t − s)xj (s)ds + ui (t), (6) ular, Figure 3 displays that e(t) → 0 when t → +∞. That is,
j=1 −∞ FPS takes place rapidly in network (6).

99112 VOLUME 8, 2020

X. Qiu et al.: FPS of Complex Networks With Distributed Delays via Hybrid Feedback Control

FIGURE 4. The chaotic attractor of the Chen system.

FIGURE 5. The state phases of the Chen system.

Simulation 2: Consider the following single Chen system:


ẋ1 = a(x2 − x1 )

ẋ2 = (c − a)x1 − x1 x3 + cx2

ẋ3 = x1 x2 − bx3


where a = 35, b = 3, and c = 28. Figure 4 depicts the chaotic

attractor of the Chen system.
The coupling configuration matrix is chosen to be
 
−1 0 1 FIGURE 6. The synchronization errors of (7).
G= 0 −1 1 .
1 1 −2
Then the controlled complex network with distributed delays In this numerical simulation, we take the initial states as
can be described as x1 (0) = (17, 14, 13)T , x2 (0) = (24, 33, 15)T , x3 (0) =
(2, 11, 13)T , and x(0) = (6, 3, 8)T . Set k1 = k2 = k3 = 2,
   
ẋi1 (t) 35(xi2 (t) − xi1 (t))
ẋi2 (t) = −7xi1 − xi1 (t)xi3 (t) + 28xi2 (t) d(0) = (10, 1, 13)T , and α(t) = cos πt 6 + 4. Figures 5
ẋi3 (t) xi1 (t)xi2 (t) − 3xi3 (t) and 6 display the state phases of the Chen system and the
3 synchronization errors, respectively. As well, FPS takes place
X Z t
+ gij K (t − s)xj (s)ds + ui (t), (7) rapidly in network (7) as expected.
j=1 −∞

IV. CONCLUSION
i = 1, 2, 3. Choose K (x) = e−x as the kernel. By using
Theorem 2 the controllers can be designed as ui (t) = ui1 (t) + FPS schemes for complex networks with distributed delays
ui2 (t), i = 1, 2, 3, where are investigated in this paper. A hybrid feedback control
method is presented to realize FPS in such networks. Finally,
35(ei1 (t)−ei2 (t))+α̇(t)x1 (t)
 
ui1 (t) = 7ei1 (t)−28ei2 (t))+xi1 (t)xi3 (t)−α(t)x1 (t)x3 (t)+α̇(t)x2 (t) , two numerical simulations are used to demonstrate the effec-
3ei3 (t)−xi1 (t)xi2 (t)+α(t)x1 (t)x2 (t)+α̇(t)x3 (t) tiveness of our method.
 
−di ei1 (t) Recently, quantized techniques were shown to be an
ui2 (t) = −di ei2 (t) . effective technique to realize synchronization of com-
−di ei3 (t) plex networks [27]–[32]. On the other hand, finite-time

VOLUME 8, 2020 99113

 X. Qiu et al.: FPS of Complex Networks With Distributed Delays via Hybrid Feedback Control

control [30]–[32] is more practical for engineering applica- [22] A. D. Kachhvah, ‘‘The effect of distributed time-delays on the synchro-
tions. In future works, we will investigate the problem of FPS nization of neuronal networks,’’ Eur. Phys. J. B, vol. 90, no. 8, Jan. 2017,
10.1140/epjb/e2016-70572-9.
of complex networks with distributed delays via finite-time [23] P. Rao, Z. Wu, and M. Liu, ‘‘Adaptive projective synchronization of
quantized control. dynamical networks with distributed time delays,’’ Nonlinear Dyn., vol. 67,
no. 3, pp. 1729–1736, Feb. 2012.
[24] Y.-J. Zhang, S. Liu, R. Yang, Y.-Y. Tan, and X. Li, ‘‘Global synchronization
ACKNOWLEDGMENT of fractional coupled networks with discrete and distributed delays,’’ Phys.
The authors would like to thank the referees for their valuable A, Stat. Mech. Appl., vol. 514, pp. 830–837, Jan. 2019.
comments. [25] H. Leng and Z. Wu, ‘‘Impulsive synchronization of complex-variable
network with distributed time delays,’’ Phys. A, Stat. Mech. Appl., vol. 536,
Dec. 2019, Art. no. 122602.
REFERENCES [26] J. Lu and J. Cao, ‘‘Synchronization-based approach for parameters identi-
[1] E. Ott, C. Grebogi, and J. A. Yorke, ‘‘Controlling Chaos,’’ Phys. Rev. Lett., fication in delayed chaotic neural networks,’’ Phys. A, Stat. Mech. Appl.,
vol. 11, pp. 1196–1199, Mar. 1993. vol. 382, no. 2, pp. 672–682, Aug. 2007.
[2] H. Gang and Z. Qiu, ‘‘Controlling spatiotemporal chaos in coupled map [27] X. Wan, X. Yang, R. Tang, Z. Cheng, H. M. Fardoun, and F. E. Alsaadi,
lattice systems,’’ Phys. Rev. Lett., vol. 72, pp. 68–73, Jan. 1994. ‘‘Exponential synchronization of semi-Markovian coupled neural net-
[3] D. J. Watts and S. H. Strogatz, ‘‘Collective dynamics of ’small-world’ works with mixed delays via tracker information and quantized output
networks,’’ Nature, vol. 393, no. 6684, pp. 440–442, 1998. controller,’’ Neural Netw., vol. 118, pp. 321–331, Oct. 2019.
[4] M. E. J. Newman and D. J. Watts, ‘‘Renormalization group analysis [28] X. Yang, Z. Cheng, X. Li, and T. Ma, ‘‘Exponential synchronization of cou-
of the small-world network model,’’ Phys. Lett. A, vol. 263, nos. 4–6, pled neutral-type neural networks with mixed delays via quantized output
pp. 341–346, Dec. 1999. control,’’ J. Franklin Inst., vol. 356, no. 15, pp. 8138–8153, Oct. 2019.
[5] X. F. Wang and G. Chen, ‘‘Pinning control of scale-free dynamical net- [29] Y. Feng, X. Yang, Q. Song, and J. Cao, ‘‘Synchronization of memristive
works,’’ Phys. A, Stat. Mech. Appl., vol. 310, nos. 3–4, pp. 521–531, neural networks with mixed delays via quantized intermittent control,’’
Jul. 2002. Appl. Math. Comput., vol. 339, pp. 874–887, Dec. 2018.
[6] C. Zhou and J. Kurths, ‘‘Hierarchical synchronization in complex net- [30] X. Yang, J. Cao, C. Xu, and J. Feng, ‘‘Finite-time stabilization of switched
works with heterogeneous degrees,’’ Chaos, Interdiscipl. J. Nonlinear Sci., dynamical networks with quantized couplings via quantized controller,’’
vol. 16, no. 1, Mar. 2006, Art. no. 015104. Sci. China Technol. Sci., vol. 61, no. 2, pp. 299–308, Feb. 2018.
[7] Z.-M. Ge and G.-H. Lin, ‘‘The complete, lag and anticipated synchroniza- [31] Y. Zhou, X. Wan, C. Huang, and X. Yang, ‘‘Finite-time stochastic syn-
tion of a BLDCM chaotic system,’’ Chaos, Solitons Fractals, vol. 34, no. 3, chronization of dynamic networks with nonlinear coupling strength via
pp. 740–764, Nov. 2007. quantized intermittent control,’’ Appl. Math. Comput., vol. 376, Jul. 2020,
[8] A. N. Pisarchik and F. R. Ruiz-Oliveras, ‘‘Optical chaotic communication Art. no. 125157.
using generalized and complete synchronization,’’ IEEE J. Quantum Elec- [32] C. Xu, X. Yang, J. Lu, J. Feng, F. E. Alsaadi, and T. Hayat, ‘‘Finite-time
tron., vol. 46, no. 3, pp. 279–284, Mar. 2010. synchronization of networks via quantized intermittent pinning control,’’
[9] W. Wu, W. Zhou, and T. Chen, ‘‘Cluster synchronization of linearly cou- IEEE Trans. Cybern., vol. 48, no. 10, pp. 3021–3027, Oct. 2018.
pled complex networks under pinning control,’’ IEEE Trans. Circuits Syst.
I, Reg. Papers, vol. 56, no. 4, pp. 829–839, Apr. 2009.
[10] J. Cao and L. Li, ‘‘Cluster synchronization in an array of hybrid coupled
neural networks with delay,’’ Neural Netw., vol. 22, no. 4, pp. 335–342, XIULIANG QIU received the M.S. degree in
May 2009. applied mathematics from Xiamen University,
[11] W. Lu, B. Liu, and T. Chen, ‘‘Cluster synchronization in networks of cou- Xiamen, China, in 2007. Since 2007, he has been a
pled nonidentical dynamical systems,’’ Chaos, Interdiscipl. J. Nonlinear member of the Chengyi University College, Jimei
Sci., vol. 20, no. 1, Mar. 2010, Art. no. 013120. University, where he is currently an Associate Pro-
[12] R. Mainieri and J. Rehacek, ‘‘Projective synchronization in three- fessor. His research interests include graph theory,
dimensional chaotic systems,’’ Phys. Rev. Lett., vol. 82, no. 15, complex networks, and nonlinear system theory.
pp. 3042–3045, Apr. 1999.
[13] D. Xu and Z. Li, ‘‘Controlled projective synchronization in nonpartially-
linear chaotic systems,’’ Int. J. Bifurcation Chaos, vol. 12, no. 06,
pp. 1395–1402, Jun. 2002.
[14] Y. Chen and X. Li, ‘‘Function projective synchronization between two
identical chaotic systems,’’ Int. J. Modern Phys. C, vol. 18, no. 5, WENSHUI LIN (Member, IEEE) received the
pp. 883–888, May 2007. Ph.D. degree in applied mathematics from Xiamen
[15] X. Tang, J. Lu, and W. Zhang, ‘‘The FPS of chaotic system using backstep- University, Xiamen, China, in 2007. Since 2007,
ping design,’’ China J. Dynam. Control, vol. 5, pp. 216–219, Jan. 2007. he has been a member of the School of Informatics,
[16] X.-J. Wu, H. Wang, and H.-T. Lu, ‘‘Hyperchaotic secure communication
Xiamen University. He is currently an Associate
via generalized function projective synchronization,’’ Nonlinear Anal.,
Professor with Xiamen University, where he is also
Real World Appl., vol. 12, no. 2, pp. 1288–1299, Apr. 2011.
[17] H. Du, Q. Zeng, C. Wang, and M. Ling, ‘‘Function projective synchro- a member of the Fujian Key Laboratory of Sensing
nization in coupled chaotic systems,’’ Nonlinear Anal., Real World Appl., and Computing for Smart City. His research inter-
vol. 11, no. 2, pp. 705–712, Apr. 2010. ests include graph theory, algorithms, and machine
[18] H. Du, Q. Zeng, and C. Wang, ‘‘A general method for function projective learning.
synchronization,’’ Intl. J. Innov. Comput. Info. Control, vol. 5, no. 8,
pp. 2239–2248, Aug. 2009.
[19] H. Du, P. Shi, and N. Lü, ‘‘Function projective synchronization in complex
dynamical networks with time delay via hybrid feedback control,’’ Nonlin- YIMING ZHENG received the B.S. degree in com-
ear Anal., Real World Appl., vol. 14, no. 2, pp. 1182–1190, Apr. 2013. puter science from Xiamen University, Xiamen,
[20] L. Shi, H. Zhu, Y. Zeng, and S. Zhong, ‘‘Function projective synchro- China, in 2017, where he is currently pursuing the
nization of complex dynamical networks with time-varying delay via M.S. degree in computer science with the School
mixed feedback control,’’ in Proc. 6th Int. Symp. Comput. Intell. Design, of Informatics. His research interests include algo-
Oct. 2013, pp. 285–288. rithms and artificial intelligence.
[21] C. Wille, J. Lehnert, and E. Schöll, ‘‘Synchronization-desynchronization
transitions in complex networks: An interplay of distributed time delay
and inhibitory nodes,’’ Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat.
Interdiscip. Top., vol. 90, no. 3, Sep. 2014.

99114 VOLUME 8, 2020