This paper studies the finite-time synchronization problem for a class of complex
dynamical networks by means of periodically intermittent control. Based
on some analysis techniques and finite-time stability theory, some novel and
effective finite-time synchronization criteria are given in terms of a set of
linear matrix inequalities. Particularly, the previous synchronization problem
by using periodically intermittent control has been extended in this paper.
Finally, numerical simulations are presented to verify the theoretical results.
1. Introduction
A complex dynamical network consists of a number of nodes, which are dynamic systems, and links between the nodes. Complex networks exist in various fields of science, engineering, and society and have been deeply investigated in recent years [1]. As the major collective behavior, synchronization is one of the key issues that has been extensively addressed. Several books and reviews [2–6] have appeared which deal with this topic.
Up till now, the synchronization for nonlinear systems especially dynamical networks [7, 8] has been one of the extensive research subjects and many important and fundamental results have been reported on the synchronization and control of nonlinear systems. Meanwhile, lots of control approaches have been developed to synchronize nonlinear systems such as adaptive control [9, 10], feedback control [11, 12], observer control [13, 14], impulsive control [15–19], and intermittent control [20–26]. Among these control approaches and other control methods, the discontinuous control methods, such as impulsive control and intermittent control, have received much interest because they are practical and easily implemented in engineering fields such as transportation and communication. Though the two control methods are discontinuous control, the intermittent control is different from the impulsive control since impulsive control is activated only at some isolated instants, while intermittent control has a nonzero control width. Under some circumstances, using intermittent control is more effective and robust [27]. Hence, some synchronization criteria for nonlinear systems with or without time delays via intermittent control have been presented in recent years; see [21, 22, 28, 29].
Nevertheless, to our best knowledge, the previous results only focus on asymptotical or exponential synchronization of networks through intermittent control; there are few results concerned with finite-time synchronization via intermittent control. In view of this, the purpose of this paper is to study the synchronization of a class of systems by designing reasonable intermittent control. In addition, some previous work views the finite-time synchronization via intermittent control in [30], which will be extended in this paper. Besides, many superiority in finite-time stability has no emphasis in this paper (see [30–32]).
The main contribution of this paper lies in the following aspects. Firstly, a new central lemma is proved by using analysis method. Additionally, an intermittent controller is designed to synchronize the addressed complex networks and some new and useful finite-time synchronization criteria are obtained. Besides, the derivative of the Lyapunov function V(t) is smaller than -αVη(t)+βV(t), which enriches the previous results in [30], when controllers are added into the network. Finally, numerical examples are given to show the effectiveness of the theoretical results.
The paper is organized as follows. In Section 2, the problem statement and synchronization scheme to be studied are formulated, and some useful lemmas and preliminaries are presented. In Section 3, some finite-time synchronization criteria for the complex dynamical networks are rigorously derived. In Section 4, the effectiveness of the developed methods is shown by numerical examples. Conclusions are finally drawn in Section 5.
2. Preliminaries
Consider a complex dynamical network consisting of N linearly and diffusively coupled identical nodes, with each node being an n-dimensional dynamical system. The state equation of the entire network is designed as follows:(1)x˙it=fxit+c∑j=1NaijΓxjt,i=1,2,…,N,where xi(t)=(xi1(t),xi2(t),…,xin(t))T∈Rn is the state vector of the ith dynamical node, f:Rn→Rn is a smooth nonlinear vector-value function, and the constant c>0 is a coupling strength. Γ∈Rn×n is the inner-coupling matrix of the network. Matrix A=(aij)∈RN×N represents the coupling configuration of the network, in which aij is defined as follows: if there is a connection from the nodes j to i, then aij≠0; otherwise, aij=0, and the diagonal elements of matrices A are defined as (2)aii=-∑j=1,j≠iNaij.
To achieve the aim of this paper, the following assumptions and some lemmas are necessary.
Assumption 1.
Assume that there exists a positive definite diagonal matrix P=diag(p1,…,pn) and a diagonal matrix Θ=diag(θ1,…,θn), such that f(·) satisfies the following inequality: (3)y-xTPfy-fx-Θy-x≤-ξy-xTy-x,for some ξ>0, all x,y∈Rn, and t>0.
Lemma 2 (see [33]).
Assume that a continuous, positive-definite function V(t) satisfies the following differential inequality:(4)V˙t≤-αVηt,∀t≥t0,Vt0≥0,where α>0 and 0<η<1 are all constants. Then, for any given t0,V(t) satisfies the following inequality:(5)V1-ηt≤V1-ηt0-α1-ηt-t0,t0≤t≤t1,Vt≡0,∀t≥t1,with t1 given by(6)t1=t0+V1-ηt0α1-η.
Lemma 3.
Assume that a continuous, positive-definite function V(t) satisfies the following differential inequality:(7)V˙t≤-αVηt+pVt,∀t≥t0,V1-ηt0≤αp,where α>0, 0<η<1, and p>0 are three constants. Then, for any given t0,V(t) satisfies the following inequality: (8)V1-ηtexp-1-ηpt≤V1-ηt0exp-1-ηpt0+αpexp-1-ηpt-exp-1-ηpt0,t0≤t≤t1,Vt≡0,∀t≥t1,with t1 given by (9)t1=ln1-p/αV1-η0pη-1for t0=0.
Proof.
Consider the following differential equation: (10)X˙t=-αXηt+pXt,Xt0=Vt0.By multiplying exp{-pt}, we have (11)dexp-ptXtdt=-αexp-ptXtηexp-1-ηpt.Although this differential equation does not satisfy the global Lipschitz condition, the unique solution to this equation can be found as (12)X1-ηtexp-1-ηpt=X1-ηt0exp-1-ηpt0+αpexp-1-ηpt-exp-1-ηpt0,Xt≡0,∀t≥t1.It is direct to prove that x(t) is differential for t>t0. From the comparison lemma, one obtains(13)V1-ηtexp-1-ηpt≤V1-ηt0exp-1-ηpt0+αpexp-1-ηpt-exp-1-ηpt0,t0≤t≤t1,Vt≡0,∀t≥t1with t1 given in (9) with t0=0.
Remark 4.
Lemma 3 is similar to Lemma 2, but our result can enrich the famous differential inequality [33] to general differential inequality, and give a direction to proof the following Lemma 5.
Lemma 5.
Suppose that function V(t) is continuous and nonnegative when t∈[0,∞) and satisfies the following conditions: (14)V˙t≤-αVηt+pVt,lT≤t<lT+θT,V˙t≤0,lT+θT≤t<l+1T,where α>0, T>0, p>0, 0<η, θ<1, l∈ι={1,2,…,ς}, and ς is a nature number; then the following inequality holds: (15)V1-ηtexp-1-ηpt≤V1-η0+αpexp-1-ηpθt-1,0≤t≤T~.
Proof.
Denote M0=V1-η(0)-α/p and W(t)=V1-ηtexp{-(1-η)pt}, where t≥0. Let Q(t)=W(t)-M0-α/pexp{-(1-η)pt}. It is easy to see that (16)Qt<0,for t=0.
In the following, we will prove that(17)Qt<0,∀t∈0,θT.Otherwise, there exists a t0∈[0,θT) such that (18)Qt0=0,Q˙t0>0,(19)Qt<0,0≤t<t0.From (16), (18), and (19), we obtain (20)Q˙t0=1-ηV-ηt0V˙t0exp-1-ηpt0+V1-ηt0-1-ηpexp-1-ηpt0-αpexp-1-ηpt0-1-ηp≤1-η·V-ηt0-αVηt0+pVt0·exp-1-ηpt0-p1-ηV1-ηt0·exp-1-ηpt0+α1-η·exp-1-ηpt0=-α1-η·exp-1-ηpt0+α1-η·exp-1-ηpt0+p1-ηV1-ηt0·exp-1-ηpt0-p1-ηV1-ηt0·exp-1-ηpt0=0,which leads to a contradiction with (18). Hence inequality (17) holds.
Now, we prove that for t∈[θT,T)(21)Ht=Wt-M0-αpexp-1-ηptexp1-ηpt-θT<0.Otherwise, there exists a t1∈[θT,T) such that (22)Ht1=0,H˙t1>0,(23)Ht<0,θT≤t<t1.By (22) and (23), we have (24)H˙t1=W˙t1-αpexp-1-ηpt1-1-ηp·exp1-ηpt1-θT-αpexp-1-ηpt1·exp1-ηpt1-θT1-ηp≤α1-η·exp-1-ηpt1exp1-ηpt1-θT-α1-ηexp-1-ηpt1·exp1-ηpt1-θT=0,which contradicts (22). Hence (21) holds.
Consequently, on the one hand, for t∈[θT,T), (25)Wt<M0+αpexp-1-ηptexp1-ηpt-θT≤M0+αpexp-1-ηptexp1-ηp1-θT.On the other hand, it follows from (16) and (17) that for t∈[0,θT)(26)Wt<M0+αpexp-1-ηpt≤M0+αpexp-1-ηptexp1-ηp1-θT.So (27)Wt<M0+αpexp-1-ηptexp1-ηp1-θT,∀t∈0,T.Similarly, we can prove the following results for t∈[T,(1+θ)T), (28)Wt<M0+αpexp-1-ηptexp1-ηp1-θT,and for t∈[(1+θ)T,2T)(29)Wt<M0+αpexp-1-ηpt·exp1-ηp1-θT·exp1-ηpt-θT-T=M0+αp·exp-1-ηptexp1-ηpt-2θT.
Now, using mathematical induction method, suppose that the following statements are true; for any integers m, we can obtain W(t).
For mT≤t<(m+θ)T,(30)Wt<M0+αpexp-1-ηpt·exp1-ηp1-θmT,and for (m+θ)T≤t<(m+1)T, (31)Wt<M0+αpexp-1-ηpt·exp1-ηpt-m+1θT.Since, for any t≥0, there exists a positive integer k, such that kT≤t<(k+1)T, we can conclude the following estimation of W(t) by (30) and (31).
For kT≤t<(k+θ)T, (32)Wt<M0+αpexp-1-ηptexp1-ηp1-θkT≤M0+αpexp-1-ηptexp1-ηp1-θt,and for (k+θ)T≤t<(k+1)T, (33)Wt<M0+αpexp-1-ηpt·exp1-ηpt-k+1θT≤M0+αp·exp-1-ηptexp1-ηp1-θt.From the previous definition of W(t), we have (34)V1-ηtexp-1-ηpt≤V1-η0-αp+αpexp-1-ηptexp1-ηp1-θt=V1-η0-αp+αpexp-1-ηθpt=V1-η0+αpexp-1-ηpθt-1,t≥0.The proof of Lemma 3 is completed.
Remark 6.
Lemmas 3 and 5 played an important role in the finite-time synchronization analysis of dynamical networks via intermittent control in this paper, because it shows the utilization of finite-time intermittent control.
Lemma 7 (see [30]).
Let x1,x2,…,xn∈Rm be any vectors and 0<q<2 is a real number satisfying (35)x1q+x2q+⋯+xnq≥x12+x22+⋯+xn2q/2.
3. Criteria for Finite-Time Synchronization
In this section, we study finite-time synchronization of system (1) with system (36) under the following intermittent controller (37).
In order to drive system (1) to achieve finite-time synchronization by means of periodically intermittent control, the corresponding response system is designed as follows: (36)y˙it=fyit+c∑j=1NaijΓyjt+uit,i=1,2,…,N,where yi(t)=(yi1(t),yi2,…,yin(t))T∈Rn, i=1,2,…,N, denote the response state vector of the node i of system (36). u(t)=(u1(t),u2(t),…,uN(t))T is an intermittent controller defined as follows:(37)uit=-ηieit-k¯λmaxPλminPsignei,lT≤t<lT+δ,uit=0,ei=0 or lT+δ≤t<l+1T,where ηi>0 is a positive constant control gain and k¯>0 is a tunable constant. Denote λmax(P)(λmin(P)) as the maximum (minimum) eigenvalue of the matrix P. T>0 is the control period, δ>0 is called the control width (control duration), and θ=δ/T is the ratio of the control width δ to the control period T called control rate. ι={1,2,…,ς} is a finite natural number set and l∈ι.
Let ei(t)=yi(t)-xi(t)(1≤i≤N) be synchronization errors between the states of drive system (1) and response system (37); then the following error system can be obtained:(38)e˙it=fyit-fxit+c∑j=1NaijΓejt+uit,lT≤t<lT+θT,i=1,2,…,N,e˙it=fyit-fxit+c∑j=1NaijΓejt,lT+θT≤t<l+1T,i=1,2,…,N.The main results are stated as follows.
Theorem 8.
Let Assumption 1 hold. Suppose that positive constants η1,η2,…,ηN, ξ, β, and a positive defined diagonal matrix P>0 satisfy (39)θjIN-Ξ+cγjA-βIN≤0,j=1,2,…,n,(40)θjIN-ξλmaxPIN+cγjA≤0,j=1,2,…,n,where Γ=diag(γ1,γ2,…,γn), Ξ=diag(η1,η2,…,ηn), Θ=diag(θ1,θ2,…,θn), and IN is the N×N identity matrix. Then under the periodically intermittent controllers (37), the error system (38) is synchronized in a finite time: (41)T1=-2ln1-β/2k¯V1/20θβ,where V(0)=1/2∑i=1NeiT(0)Pei(0) and ei(0) is the initial condition of ei(t).
Proof.
Consider the following Lyapunov function: (42)Vt=12∑i=1NeiTtPeit.Then the time derivative of (42) along the trajectories of the first subsystem of system (38) is calculated and estimated as follows.
When lT≤t<(l+θ)T, for l∈ι,(43)V˙t=∑i=1NeiTtPe˙it=∑i=1NeiTtPfyit-fxit+c∑j=1NaijΓejt+uit=∑i=1NeiTtPfyit-fxit-Θeit+eiTtPΘeit+eiTtPuit+ceiTtPΓ∑j=1Naijejt≤-ξ∑i=1NeiTteit+∑i=1NeiTtΘP-ηiPeit+c∑i=1NeiTΓ∑j=1NaijPejt-k¯∑i=1NλmaxPλminPeiTtPsignei-β∑i=1NeiTt·eit+β∑i=1NeiTteit≤∑j=1npje~jTtθjIN-Ξ+cγjA-βINe~jt+β∑i=1NeiTteit-k¯∑i=1NλmaxPλminPeiTtPsignei=∑j=1npje~jTt·θjIN-Ξ+cγjA-βINe~jt+β∑i=1NeiTteit-k¯∑i=1NλmaxPλminPeiTtPsignei.Defining sign(ei)=(signei1,signei2,…,sign(ein))T, we have (44)V˙t≤β∑i=1NeiTtPeit+∑j=1npje~jTtZje~jTt-k¯∑i=1NλmaxPei,where e~j(t)=[e~j1,e~j2,…,e~jN] is a column vector of ej(t) and Zj is defined as (45)Zj=θjIN-Ξ+cγjA-βIN,j=1,2,…,n.Using Lemma 7, we have (46)V˙t≤β∑i=1NeiTtPeit+∑j=1npje~jTtZje~jTt-2k¯12∑i=1NeiTtPeit1/2.It follows from inequality (39) that(47)Zj≤0,which shows that V˙(t)≤-2k¯V1/2(t)+βV(t).
When (l+θ)T≤t<(l+1)T, for l∈ι, we have (48)V˙t=∑i=1NeiTtPe˙it=∑i=1NeiTtPfyit-fxit-Θeit+eiTtPΘeit+c∑j=1NaijeiTtPΓejt≤-ξ∑i=1NeiTteit+∑j=1npje~jTtθjINe~jt+c∑j=1npje~jTtγjAe~jt≤-∑j=1npje~jTt·ξλmaxPINe~jt+∑j=1npje~jTtθjINe~jt+c∑j=1npje~jTtγjAe~jt=∑j=1npje~jTtSje~jTt,where e~j(t)=[e~j1,e~j2,…,e~jN] is a column vector of ej(t) and Sj is defined as (49)Sj=θjIN-ξλmaxPIN+cγjA,j=1,2,…,n.It follows from inequality (40) that (50)Sj≤0,which shows that V˙(t)≤0.
Namely, defining α=2k¯, η=1/2, and p=β, we get(51)D+Vt≤-αVηt+pVt,lT≤t≤lT+θT,D+Vt≤0,lT+θT≤t<l+1T.According to Lemma 5, we have (52)V1-ηtexp-1-ηpt≤V1-η0+αpexp-1-ηpθt-1,0≤t≤T1.By Lemma 3, we have(53)t≤ln1-p/αV1-η0pθη-1=-2ln1-β/2k¯V1/20θβ,0≤t≤T1.The proof of Theorem 8 is completed.
Remark 9.
Obviously, when θ=1, the intermittent control (37) is degenerated to a continuous control input which has been extensively proposed in previous work (see [34, 35]) and focuses on [13]. However, this trivial case is not to be discussed in this paper.
If the Lyapunov function V˙(t)≤-αV(t)(p=0) when controllers are added into the network, then it is easy to see that Theorem 8 can be restated as the following corollary.
Corollary 10.
Let Assumption 1 hold. Suppose that positive constants η1,η2,…,ηN, ξ, and a positive define diagonal matrix P>0 satisfy (54)θjIN-Ξ+cγjA≤0,j=1,2,…,n,(55)θjIN-ξλmaxPIN+cγjA≤0,j=1,2,…,n,where Γ=diag(γ1,γ2,…,γn), Ξ=diag(η1,η2,…,ηn), Θ=diag(θ1,θ2,…,θn), and IN is the N×N identity matrix. Then under the periodically intermittent controllers (37), the error system (38) is synchronized in a finite time: (56)T2=V1-η02k¯θ,where V(0)=1/2∑i=1NeiT(0)Pei(0) and ei(0) is the initial condition of ei(t).
Remark 11.
Corollary 10 in this paper is the main result of Theorem 2 in [30] and the main results in [36].
Remark 12.
According to (41) and (56) and the convergence time T1,T2, we can conclude that the convergence time satisfies T2≤T1. We can analyse that the term βV(t) should impede the convergence time. But compared with the control gain matrix Ξ in (54), it is easy to seek an appropriate control gain matrix Ξ in (39) for which synchronization happens.
Remark 13.
It is clear to see that inequality (39) is more easily satisfied compared with inequality (54) under the same controllers and conditions via LMI Toolbox, which reveals a very interesting phenomenon; that is, the control gain under condition (39) is more easily designed than condition (54), though it can impede the convergence time.
Remark 14.
We can find that if condition (55) is satisfied, condition (54) easily holds when the positive definite control gain matrix Ξ is anything. Then we have the following corollary.
Corollary 15.
Let Assumption 1 hold. Suppose that positive constants η1,η2,…,ηN, ξ, and a positive defined diagonal matrix P>0 satisfy(57)θjIN-ξλmaxPIN+cγjA≤0,j=1,2,…,n,where Γ=diag(γ1,γ2,…,γn), Θ=diag(θ1,θ2,…,θn), and IN is the N×N identity matrix. Then under the periodically intermittent controllers (37), the error system (38) is synchronized in a finite time: (58)T3=V1-η02k¯θ,where V(0)=1/2∑i=1NeiT(0)Pei(0) and ei(0) is the initial condition of ei(t).
4. Numerical Examples
In this section, we give some numerical examples to show the validity and effectiveness of the derived results for finite-time synchronization via periodically intermittent control.
In this case, to demonstrate the results above, we consider general complex dynamical networks, in which each subsystem is a Lorenz system. The dynamics of Lorenz system is described as follows:(59)s˙=fs=s˙1s˙2s˙3=-aa0c-1000-bs1s2s3+0-s1s3s1s2≜Cs+ψs,where the parameters are selected as a=10, c=30, and b=8/3; then the Lorenz system has a chaotic attractor (see Figure 1). Moreover, it is known that |xi1|≤18.6360=r1, |xi2|≤25.3679=r2, and |xi3|≤45.9792=r3. Now we will show that there exists a positive definite diagonal matrix P that satisfies Assumption 1. Let x=[x1,x2,x3]T, y=[y1,y2,y3]T, δ=x-y=[δ1,δ2,δ3]T=[x1-y1,x2-y2,x3-y3]T, and |ϵ|=[|ϵ1|,|ϵ2|,|ϵ3|]T; then (60)ψx-ψy=0-x1x3+y1y3x1x2-y1y2=0-x1ϵ3-y3ϵ1x1ϵ2+y2ϵ1.Hence, (61)x-yTψx-ψy=-y3ϵ1ϵ2+y2ϵ1ϵ3≤12ϵT0r3r2r300r200ϵ≜12ϵTRϵ.Then (62)x-yTPfx-fy-Θx-y≤λmaxC+1/2λmaxR-λminΘλminPx-yT·x-y,where λmax(C),λmax(R),λmin(Θ) are the maximum eigenvalues of C,R,Θ, respectively. Let ξ=-λmax(C)+1/2λmax(R)-λmin(Θ)/λmin(P)=43.48; therefore Assumption 1 is satisfied with Θ=diag(50,50,50) and P=diag(0.5,0.4,0.2).
The chaotic trajectories of the Lorenz system.
Consider the complex dynamical network (1) consisting of 50 identical Lorenz oscillators nodes, which is described by(63)x˙it=fxit+c∑j=150aijΓxjt,i=1,2,…,50,where Γ=diag(1,1,1) and A=(aij)50×50 is a symmetrically diffusive coupling matrix with aij=0 or 1(j≠i) and the coupling strength c=1.
The corresponding controlled response system of (36) is of the form (64)y˙it=fyit+c∑j=150aijΓyjt+uit,i=1,2,…,50,where aij and fi(·) are the same as (63) and(65)uit=-ηieit-k¯λmaxPλminPsignei,lT≤t<lT+δ,uit=0,ei=0 or lT+δ≤t<l+1T,where k¯=5.
The initial states of the numerical simulations in the master and slave systems are as follows: xi(0)=(2+0.2i,0.2+0.3i,0.3+0.1i)T and yi(0)=(-8+0.7i,-5+0.8i,-10+0.6i)T. The initial conditions of the error system are ei(0)=(-10+0.5i,-5.2+0.5i,-10.3+0.5i)T, where 1≤i≤50. In addition, the values of the parameters for the controllers (37) are selected as T=1 and δ=0.8. Choosing Θ=diag(50,50,50), P=diag(0.5,0.4,0.2), and λmax(A)=-1.6414, one can easily obtain that inequality (40) holds. Besides, we can obtain the parameters of the intermittent controllers as Ξ=diag[57.9071,…,57.9071]50×50 and β=1.304 by using LMI toolbox in Matlab to solve (39) and (40). Therefore from Theorem 8 it can be obtained that the response system (36) will synchronize the drive system (1) under the periodically intermittent controllers (37) within the time T=1.234 (the average time). The synchronous error ei(t) is illustrated in Figures 2–4.
Synchronization errors ei1, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.8.
Synchronization errors ei2, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.8.
Synchronization errors ei3, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.8.
Take k¯=5, T=0.5, and θ=0.4. By calculating (41), we can obtain the convergence time t=2.242. The time responses of the error variables eij, 1≤i≤50, j=1,2,3, are illustrated in Figures 5–7, with k¯=5, T=0.5, and θ=0.4. Let k¯=5, T=0.5, and θ=0.1. By calculating (41), we can obtain the convergence time t=3.278. Figures 8–10 describe the time responses or the error variables eij, 1≤i≤50, j=1,2,3, respectively, with different parameters of k¯,T,θ.
Synchronization errors ei1, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.4.
Synchronization errors ei2, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.4.
Synchronization errors ei3, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.4.
Synchronization errors ei1, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.1.
Synchronization errors ei2, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.1.
Synchronization errors ei3, 1≤i≤50, with control parameters k¯=5, T=0.5, and θ=0.1.
Remark 16.
From the above analysis and figures, we can conclude that the control rate influences the convergence time; in addition, the tunable k¯ can also influence the convergence time, which is discussed in [34]; we should omit it here. And then, it can be seen that the continuous controller can synchronize the network at t=0.984. Therefore, the convergence time is shorter than the periodically intermittent control.
To further verify the effectiveness of the proposed control design, we take the control rate θ=1, namely, general continuous full control. Figures 11–13 demonstrate time response of the error variables eij, 1≤i≤50, j=1,2,3, respectively, with k¯=5, T=0.5, and θ=1. When the control variables ui=0, 1≤i≤50, namely no controlled is added to the system. The time response of the error variables eij, 1≤i≤50, j=1,2,3, is displayed in Figures 14–16, respectively, with no control input ui.
Synchronization errors ei1, 1≤i≤50, with full control parameters k¯=5, T=0.5, and θ=1.
Synchronization errors ei2, 1≤i≤50, with full control parameters k¯=5, T=0.5, and θ=1.
Synchronization errors ei3, 1≤i≤50, with full control parameters k¯=5, T=0.5, and θ=1.
Synchronization errors ei1, 1≤i≤50, with no control input.
Synchronization errors ei2, 1≤i≤50, with no control input.
Synchronization errors ei3, 1≤i≤50, with no control input.
5. Conclusion
This paper has dealt with the finite-time synchronization problem for a class of complex dynamical networks by means of periodically intermittent control. Some novel and useful synchronization criteria, given in terms of a set of linear matrix inequalities, have been obtained by some analysis techniques and finite-time stability theory. Our results reduce the previous works on the controllers that the derivative of the Lyapunov function V(t) is smaller than -αVη(t). Several simulations are presented to verify the effectiveness of the proposed synchronization criteria finally. The future work in this endeavor will focus on the global problem.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
The authors would like to thank the editor and the anonymous reviewers for their valuable comments and constructive suggestions. This project is supported by the National Science Foundation of China (11371162 and 1171129) and the Education Department of Hubei Province (T201103).
BoccalettiS.LatoraV.MorenoY.ChavezM.HwangD.-U.Complex networks: structure and dynamics20064244-517530810.1016/j.physrep.2005.10.009MR21936212-s2.0-31344474880ArenasA.Díaz-GuileraA.KurthsJ.MorenoY.ZhouC.Synchronization in complex networks200846939315310.1016/j.physrep.2008.09.002MR24770972-s2.0-56549084695PikovskyA.RosenblumM.KurthsJ.200112Cambridge, UKCambridge University PressCambridge Nonlinear Science Series10.1017/cbo9780511755743MR1869044StrogatzS. H.Exploring complex networks2001410682526827610.1038/350657252-s2.0-0035826155WangX. F.ChenG.Synchronization in small-world dynamical networks200212118719210.1142/s02181274020042922-s2.0-0036340313WuC.2007SingaporeWorld ScientificWangX.ChenG.Synchronization in scale-free dynamical networks: robustness and fragility20024915462MeiJ.JiangM.HuangZ.Outer synchronization between two complex networks with identical and nonidentical topological structuresProceedings of the 4th International Workshop on Advanced Computational Intelligence (IWACI '11)October 201175776210.1109/iwaci.2011.61601082-s2.0-84858790218YuH.XiaX.Adaptive consensus of multi-agents in networks with jointly connected topologies20124881783179010.1016/j.automatica.2012.05.068MR29504302-s2.0-84864486368YuH.ShenY.XiaX.Adaptive finite-time consensus in multi-agent networks2013621088088910.1016/j.sysconle.2013.06.011MR31051272-s2.0-84881279630JiangM.WangS.MeiJ.ShenY.Finite-time synchronization control of a class of memristor-based recurrent neural networks20156313314010.1016/j.neunet.2014.11.005JiangM.MeiJ.HuJ.New results on exponential synchronization of memristor-based chaotic neural networks2015156606710.1016/j.neucom.2014.12.085ShenY.HuangY.GuJ.Global finite-time observers for lipschitz nonlinear systems201156241842410.1109/tac.2010.2088610MR27611142-s2.0-79951662588ShenY.HuangY.Uniformly observable and globally Lipschitzian nonlinear systems admit global finite-time observers200954112621262510.1109/tac.2009.2029298MR25719252-s2.0-79952900805YangX.CaoJ.LuJ.Synchronization of delayed complex dynamical networks with impulsive and stochastic effects20111242252226610.1016/j.nonrwa.2011.01.007MR28010172-s2.0-79955522542ZhangQ.LuJ.-a.ZhaoJ.Impulsive synchronization of general continuous and discrete-time complex dynamical networks20101541063107010.1016/j.cnsns.2009.05.048MR25570142-s2.0-70350367337SunM.ZengC.TaoY.TianL.Adaptive-impulsive synchronization in drive-response networks of continuous systems and its application2009373343041304610.1016/j.physleta.2009.06.0472-s2.0-67651122957ZhengS.DongG.BiQ.Impulsive synchronization of complex networks with non-delayed and delayed coupling2009373464255425910.1016/j.physleta.2009.09.0432-s2.0-74249104644CaiS.ZhouJ.XiangL.LiuZ.Robust impulsive synchronization of complex delayed dynamical networks2008372304990499510.1016/j.physleta.2008.05.0772-s2.0-46449094194XiaW.CaoJ.Pinning synchronization of delayed dynamical networks via periodically intermittent control200919101312010.1063/1.3071933MR25137642-s2.0-63849123030CaiS.LiuZ.XuF.ShenJ.Periodically intermittent controlling complex dynamical networks with time-varying delays to a desired orbit2009373423846385410.1016/j.physleta.2009.07.081MR25525422-s2.0-70349132609CaiS.HaoJ.HeQ.LiuZ.Exponential synchronization of complex delayed dynamical networks via pinning periodically intermittent control2011375191965197110.1016/j.physleta.2011.03.0522-s2.0-79955475409MeiJ.JiangM.WuZ.WangX.Periodically intermittent controlling for finite-time synchronization of complex dynamical networks201579129530510.1007/s11071-014-1664-yYuJ.HuC.JiangH.TengZ.Synchronization of nonlinear systems with delays via periodically nonlinear intermittent control20121772978298910.1016/j.cnsns.2011.11.028MR28804672-s2.0-84856425757ZhuH.CuiB.Stabilization and synchronization of chaotic systems via intermittent control201015113577358610.1016/j.cnsns.2009.12.029MR2646187ZBL1222.931942-s2.0-77952237294MeiJ.JiangM.WangB.LiuQ.XuW.LiaoT.Exponential psynchronization of non-autonomous Cohen-Grossberg neural networks with reaction-diffusion terms via periodically intermittent control201440103126LiC.FengG.LiaoX.Stabilization of nonlinear systems via periodically intermittent control200754111019102310.1109/TCSII.2007.9032052-s2.0-58349087625CaiS.HeQ.HaoJ.LiuZ.Exponential synchronization of complex networks with nonidentical time-delayed dynamical nodes2010374252539255010.1016/j.physleta.2010.04.023MR26400292-s2.0-77953291894HuC.YuJ.JiangH.TengZ.Exponential synchronization of complex networks with finite distributed delays coupling201122121999201010.1109/tnn.2011.21677592-s2.0-83855160852MeiJ.JiangM.XuW.WangB.Finite-time synchronization control of complex dynamical networks with time delay20131892462247810.1016/j.cnsns.2012.11.009MR30420522-s2.0-84878013103MeiJ.JiangM.WangB.LongB.Finite-time parameter identification and adaptive synchronization between two chaotic neural networks201335061617163310.1016/j.jfranklin.2013.04.005MR30675772-s2.0-84878234569MeiJ.JiangM.WangJ.Finite-time structure identification and synchronization of drive-response systems with uncertain parameter2013184999101510.1016/j.cnsns.2012.08.039MR29966122-s2.0-84868212737TangY.Terminal sliding mode control for rigid robots1998341515610.1016/s0005-1098(97)00174-xMR1614164YangX.CaoJ.Finite-time stochastic synchronization of complex networks201034113631364110.1016/j.apm.2010.03.012MR26517952-s2.0-77952957846LiuX.JiangN.CaoJ.WangS.WangZ.Finite-time stochastic stabilization for BAM neural networks with uncertainties201335082109212310.1016/j.jfranklin.2013.05.027MR30840612-s2.0-84880717606MeiJ.JiangM.WangX.HanJ.WangS.Finite-time synchronization of drive-response systems via periodically intermittent adaptive control201435152691271010.1016/j.jfranklin.2014.01.008MR31919152-s2.0-84898801293