This article was downloaded by: [University of California, San Diego]

On: 07 January 2015, At: 21:12

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Control

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tcon20

Consensus of second-order multi-agent systems

with delayed nonlinear dynamics and intermittent
communications
a a b c a d
Guanghui Wen , Zhisheng Duan , Wenwu Yu & Guanrong Chen
a
Department of Mechanics and Aerospace Engineering, College of Engineering , Peking
University , Beijing 100871 , China
b
Department of Mathematics , Southeast University , Nanjing 210096 , China
c
School of Electrical and Computer Engineering , RMIT University , Melbourne , VIC 3001 ,
Australia
d
Department of Electronic Engineering , City University of Hong Kong , Hong Kong SAR ,
China
Published online: 28 Jan 2013.

To cite this article: Guanghui Wen , Zhisheng Duan , Wenwu Yu & Guanrong Chen (2013) Consensus of second-order multi-
agent systems with delayed nonlinear dynamics and intermittent communications, International Journal of Control, 86:2,
322-331, DOI: 10.1080/00207179.2012.727473

To link to this article: http://dx.doi.org/10.1080/00207179.2012.727473

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 International Journal of Control
Vol. 86, No. 2, February 2013, 322–331

Consensus of second-order multi-agent systems with delayed nonlinear dynamics and

intermittent communications
Guanghui Wena*, Zhisheng Duanb, Wenwu Yuac and Guanrong Chenbd
a
Department of Mathematics, Southeast University, Nanjing 210096, China; bDepartment of Mechanics and
Aerospace Engineering, College of Engineering, Peking University, Beijing 100871, China; cSchool of Electrical
and Computer Engineering, RMIT University, Melbourne, VIC 3001, Australia; dDepartment of Electronic Engineering,
City University of Hong Kong, Hong Kong SAR, China
(Received 18 March 2012; final version received 3 September 2012)
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

This article investigates the second-order consensus problem of multi-agent systems with inherent delayed
nonlinear dynamics and intermittent communications. Each agent is assumed to obtain the measurements of
relative states between its own and the neighbours’ only at a sequence of disconnected time intervals. A new kind
of protocol based only on the intermittent measurements of neighbouring agents is proposed to guarantee the
states of agents to reach second-order consensus under a fixed strongly connected and balanced topology.
By constructing a common Lyapunov function, it is shown that consensus can be reached if the general algebraic
connectivity and the communication time duration are larger than their corresponding threshold values,
respectively. Finally, simulation examples are provided to verify the effectiveness of the theoretical analysis.
Keywords: multi-agent system; consensus; intermittent communication; delayed nonlinear dynamics; directed
graph

1. Introduction neighbours (Vicsek et al. 1995). Some theoretical

Recently, cooperative control of coupled dynamical analysis of the consensus problem on the linearised
systems has received considerable attention partly Vicsek’s mode was provided in Jadbabaie et al. (2003).
owing to its wide applications in multi-agents systems, Then, in Olfati-Saber and Murray (2004), a general
where examples include state consensus seeking of framework of the consensus problem for networks of
multiple mobile vehicles (Beard, McLain, Goodrich, dynamic agents with fixed or switching topology and
and Anderson 2002; Fax and Murray 2004; Ren and communication time delays was suggested and investi-
Atkins 2007), design of distributed sensor networks gated. The consensus conditions derived in Olfati-Saber
(Yu, Chen, Wang, and Yang 2009) and control of and Murray (2004) were further relaxed in Ren and
flocking and rendezvous in natural as well as social Beard (2005), by showing that consensus in first-order
systems (Vicsek, Cziroók, Ben-Jacob, Cohen, and multi-agent systems can be achieved if and only if the
Shochet 1995; Toner and Tu 1998; Topaz and switching topology contains a spanning tree frequently
Bertozzi 2005; Wen, Duan, Su, Chen, and Yu 2012c). enough. In Wang, Tan, and Mareels (2009), the
Among the numerous research topics in cooperative robustness problem of consensus seeking in multi-
control, the consensus problem receives particular agent systems with leader-following topology has been
attention (Jadbabaie, Lin, and Morse 2003; Olfati- introduced and studied. Note that most of the above-
Saber and Murray 2004; Ren and Beard 2005), which mentioned works are concerned with the first-order
can be generally described as how to design an consensus problem, where each agent is governed by
appropriate protocol based only on the local relative first-order dynamics. In reality, however, a large class
information under some communication topologies to of multi-agent systems are modelled by second-order
ensure the multiple agents to reach an agreement on dynamics (Xie and Wang 2005; Hong, Hu, and Gao
certain quantities of interest. 2006; Ren and Atkins 2007; Hong, Chen, and Bushnell
In 1995, Vicsek et al. introduced an interesting 2008; Ren 2008). In Xie and Wang (2005), the second-
discrete-time model of mobile agents, where each order consensus problem for multi-agent systems with
agent’s motion is updated according to a local rule zero initial and finial consensus velocities under undi-
based on its own state as well as the states of its rected communication topology was investigated.

*Corresponding author. Email: wenguanghui@gmail.com

ISSN 0020–7179 print/ISSN 1366–5820 online

ß 2013 Taylor & Francis
http://dx.doi.org/10.1080/00207179.2012.727473
http://www.tandfonline.com
 International Journal of Control 323

Taking into account the general case where information introduced into the model of each agent in this article.
flow may be unidirectional due to sensors with limited By virtue of the Lyapunov control approach, some
sensing ranges or multi-agents with directed communi- sufficient conditions are derived for reaching second-
cation links, a new kind of second-order consensus order consensus with time-varying agent velocities.
problems under directed communication topology was Thus, the main contributions of this work are two-fold.
discussed in Ren and Atkins (2007). Some sufficient First, the agent dynamics are extended to be in a
conditions were obtained for achieving second-order general form, without limiting to integrators. Second, a
consensus, and it was shown that the communication new kind of consensus protocols based only on the
topology having a spanning tree is not a sufficient intermittent relative state information of neighbouring
condition for reaching second-order consensus, which agents is designed, which is more general than the
is different in kind from the first-order consensus common second-order consensus protocols where the
problems (Ren and Atkins 2008). Then, several novel agents need to communicate with their neighbours all
consensus tracking algorithms have been proposed in the time.
Hong et al. (2006) and Hong et al. (2008), respectively. The rest of this article is organised as follows. In
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

Knowing that transmission time delay is a key factor Section 2, some preliminaries in algebraic graph theory
influencing the stability of consensus in linear multi- and the model formulation are given. In Section 3, a
agent systems, a necessary and sufficient condition was second-order consensus problem with delayed non-
obtained for second-order consensus in networked linear dynamics and intermittent measurements is
multi-agent systems with transmission delays in Yu, investigated. In Section 4, numerical simulations are
Chen, and Cao (2010a). In contrast to the aforemen- provided to verify the theoretical results. Conclusions
tioned second-order consensus algorithms, where the are finally drawn in Section 5.
consensus velocity is a constant, a consensus algorithm The following notations are used throughout this
in coupled second-order linear harmonic oscillators article. Let R and N be the sets of real and natural
with asymptotic periodic velocity and directed commu- numbers, respectively. RN is the N-dimensional real
nication topology was considered in Ren (2008). The vector space and k  k denotes the Euclidian norm.
dynamical model studied in Ren (2008) in essence is a RNN is the N  N real matrix space. Let IN (ON) be
second-order multi-agent system with inherent linear the N-dimensional identity (zero) matrix, and 1N (0N)
dynamics. A more general case is that each agent has its be the N-dimensional column vector with each entry
inherent nonlinear dynamics (Li, Wang, and Chen being 1 (0) and 1NN be the N  N matrix with all
2004; Su, Chen, Wang, and Lin 2011; Li, Liu, and Fu entries being 1. Suppose that matrix M 2 RNN has real
2011; Zhang and Lewis 2012). From this perspective, eigenvalues, with i(M) being the ith smallest eigen-
Yu, Chen, Cao, and Kurths (2010b), investigated the value (1  i  N ). Notations  and k  k represent the
second-order consensus problem in multi-agent systems Kronecker product and Euclidean norm, respectively.
with nonlinear dynamics and directed topology by
using tools from algebraic graph theory and Lyapunov
control approach, some sufficient conditions were 2. Preliminaries and formulation of the model
derived for reaching second-order consensus with
In this section, some preliminaries in algebraic graph
time-varying consensus velocities.
theory and model formulation for second-order con-
It should be noticed that most of the aforemen-
sensus in multi-agent systems with delayed nonlinear
tioned works on second-order consensus problems in
dynamics and intermittent measurements are
multi-agent systems were derived based on an assump-
introduced.
tion that information is transmitted continuously
among multi-agents. However, this may not be the
case in reality due to technological limitations or
external disturbances. For example, in some cases, 2.1 Algebraic graph theory
agents can only obtain the measurements of states of Let G(V, E, A) be a directed graph with the set of nodes
its neighbours intermittently due to the limited of V ¼ {v1, v2, . . . , vN}, the set of directed edges E  V  V
sensing abilities (Wen, Duan, Li, and Chen 2012a, and a weighted adjacency matrix A ¼ [aij]NN with
2012b; Wen, Duan, Yu, and Chen 2012d). To deal with non-negative adjacency elements aij. An edge eij in
this challenging situation, a kind of intermittent graph G is denoted by the ordered pair of nodes (vj, vi),
consensus protocol is proposed in this article to where vj and vi are called the parent and child nodes,
guarantee second-order consensus. On the other respectively, and eij 2 E if and only if aij 4 0.
hand, in order to analyse the second-order consensus Furthermore, self-loops are not allowed, i.e. aii ¼ 0
problem in multi-agent systems within a general for all i ¼ 1, 2, . . . , N. For simplicity, denote G(V, E, A)
framework, inherent delayed nonlinear dynamics are by G(A) if no confusion will arise.
 324 G. Wen et al.

A directed path from node vi to vj is a finite-ordered its complete state information for feedback, some real
sequence of edges, ðvi , vk1 Þ, ðvk1 , vk2 Þ, . . . , ðvkl , vj Þ, with multi-agent vehicle systems with complex dynamics,
distinct nodes vkm , m ¼ 1, 2, . . . , l. A directed graph is e.g. holonomic mobile robots, can be feedback line-
called strongly connected if and only if there is a arised as coupled double integrators described by (3).
directed path between any pair of distinct nodes. However, such complete information of agents is hard
Moreover, a directed graph G(A) is called balanced if or impossible to get in the framework of multi-agent
X X systems. It is thus interesting and important to
aij ¼ aji 8 i ¼ 1, 2, . . . , N: ð1Þ
j j
investigate the consensus problem of multi-agent
system models without using feedback linearisation.
The Laplacian matrix L ¼ [lij]NN of G(A) is defined as Note also that a large class of real coupled dynamical
8 systems can be modelled as second-order multi-agent
> a , i 6¼ j,
< ij systems with inherent nonlinear dynamics; examples
lij ¼ X
N
ð2Þ include coupled pendulum systems (Amster and
>
: aik , i ¼ j:
k¼1,k6¼i Mariani 2008) and networks of mass-spring systems
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

with uncertain spring constants (Li, Duan, Xie, and

For a directed graph, the Laplacian matrix L has Liu 2012). Furthermore, it is commonly assumed in
the following properties. most existing works on consensus for nonlinear multi-
Lemma 1 (Ren and Beard 2005): Suppose that a agent systems that all information is transmitted
directed graph G(A) is strongly connected. Then, 0 is a continuously among the multiple dynamic agents.
simple eigenvalue of its Laplacian matrix L, and all the However, in some real situations, agents may only
other eigenvalues of L have positive real parts. communicate with their neighbours over some discon-
nected time intervals due to the unreliability of
Lemma 2 (Olfati-Saber and Murray 2004): A direc- communication channels, failure of physical devices,
ted graph G(A) is balanced if and only if 1N is the left and limitations of sensing ranges, etc. Motivated by
eigenvector of its Laplacian matrix L associated with this observation and based on the above-mentioned
zero eigenvalue, i.e. 1TN L ¼ 0TN . works (Ren and Atkins 2007, 2008), the following
For an undirected graph, its Laplacian matrix L is consensus protocol with time delay and intermittent
positive semi-definite. For a connected undirected measurements is considered:
graph, there is one simple zero eigenvalue of L, and 8
>
> x_ i ðtÞ ¼ vi ðtÞ,
all the other eigenvalues of L are positive and real. >
>
>
> X
N X
N
>
> v_ ðtÞ ¼ f ðv ðt  Þ, v ðtÞ, tÞ   l x ðtÞ   lij vj ðtÞ,
>
< i i i ij j
j¼1 j¼1
2.2 Formulation of the model >
>
> t 2 ½k!, k! þ Þ,
Consider a group of N agents indexed by 1, 2, . . . , N. >
>
>
> v_i ðtÞ ¼ f ðvi ðt  Þ, vi ðtÞ, tÞ, t 2 ½k! þ , ðk þ 1Þ!Þ,
The commonly studied continuous-time second-order >
>
:
protocol of the N agents is described as follows (Ren k 2 N, i ¼ 1, 2, . . . , N,
and Atkins 2007): ð4Þ
8
> x_ ðtÞ ¼ vi ðtÞ, where f : Rn  Rn  [0, þ1) ! Rn is a continuously
< i
X N X
N differentiable vector-valued function representing the
>
: v_i ðtÞ ¼  lij xj ðtÞ   lij vj ðtÞ, i ¼ 1, 2, . . . , N, inherent delayed nonlinear dynamics of agent i,  4 0
j¼1 j¼1 is the time delay, and the communication time duration
ð3Þ  satisfies  5   !. Positive scalars  and  represent
n n the coupling strengths. Furthermore, xi(t) ¼ i(t),
where xi 2 R and vi 2 R are the position and velocity
vi(t) ¼ i(t), for all t 2 [, 0], and the initial functions
states of the ith agent, respectively,  and  represent
i and i are continuous for all t 2 [, 0], where
the coupling strengths, L ¼ [lij]NN is the Laplacian
matrix of the fixed communication topology G(A). i ¼ 1, 2, . . . , N.
When the agents reach second-order P consensus, the Definition 1: Second-order consensus in multi-agent
velocities of all agents converge to N j¼1 j vj ð0Þ, which system (4) is said to be achieved if, for any initial
depends only on the initial velocities of the agents, conditions,
where  ¼ (1, 2, . . . , N)T is the non-negative left
eigenvector of L associated with the eigenvalue 0 lim kxi ðtÞ  xj ðtÞk ¼ 0,
t!1
satisfying T1N ¼ 1 (Ren and Atkins 2007, 2008). Note ð5Þ
lim kvi ðtÞ  vj ðtÞk ¼ 0, 8i, j ¼ 1, 2, . . . , N:
that under the assumption that each agent can obtain t!1
 International Journal of Control 325

Remark 1: It is noted that system (4) covers the 3. Main results

multi-agent systems with second-order integrator-type In this section, the second-order consensus problem in
dynamics. For example, system (4) becomes a second- strongly connected and balanced networks with time
order multi-agent system studied in Yu et al. (2010a), delayed nonlinear dynamics and intermittent measure-
if  ¼ !, and f(vi(t  ), vi(t), t)  0, for all t 0, ments is investigated.
i ¼ 1, 2, . . . , N. Before moving on, the following assumption is
PN introduced, which will be used in deriving the main
Clearly, since j¼1 lij ¼ 0, if consensus can be
achieved, it is natural to require a solution sðtÞ ¼ results.
ðsT1 ðtÞ, sT2 ðtÞÞT 2 R2n of the system (4) be a possible Assumption 1: There exist non-negative constants i,
trajectory of an isolated node satisfying i 2 {1, 2}, such that
(
s_1 ðtÞ ¼ s2 ðtÞ, X
2
ð6Þ k f ðx1 , x2 , tÞ  f ð y1 , y2 , tÞk  i kxi  yi k,
s_2 ðtÞ ¼ f ðs2 ðt  Þ, s2 ðtÞ, tÞ: i¼1
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

Here, s(t) may be an isolated equilibrium point, a where xi, yi 2 Rn, t 0.

periodic orbit or even a chaotic orbit in some P
P Let x~ i ðtÞ ¼ xi ðtÞ  N1 N j¼1 xj ðtÞ and v~i ðtÞ ¼ vi ðtÞ 
applications. 1 N
v j ðtÞ. One has the following error dynamical
N j¼1
Lemma 3 (Schur complement, Boyd, Ghaoui, Ferion, system:
and Balakrishnan 1994): The following linear matrix 8
>
> x_~ i ðtÞ ¼ v~i ðtÞ,
inequality, >
>
>
> X N
! >
> v_~ ðtÞ ¼ f ðv ðt  Þ, v ðtÞ, tÞ  1
>
> i i i f ðvj ðt  Þ, vj ðtÞ, tÞ
S11 S12 >
> N j¼1
S¼ 4 0, >
>
>
>
S21 S22 < X N X N
 lij x~ j ðtÞ   lij v~j ðtÞ, t 2 ½k!, k! þ Þ,
where S11 ¼ ST11 , S12 ¼ ST21 , S22 ¼ ST22 , is equivalent to >
>
>
> j¼1 j¼1
one of the following conditions: >
>
>
> X N
>
>
> _~i ðtÞ ¼ f ðvi ðt  Þ, vi ðtÞ, tÞ  1
v f ðvj ðt  Þ, vj ðtÞ, tÞ,
(1) S11 4 0, S22  S21 S1
11 S12 4 0;
>
> N j¼1
>
>
(2) S22 4 0, S11  S12 S1
22 S21 4 0.
>
:
t 2 ½k! þ , ðk þ 1Þ!Þ, k 2 N,
Lemma 4 (Halanay inequality, Fujisaka and Yamada ð7Þ
1983): Suppose that the non-negative function y(t),
t 2 [, þ1), satisfies where i ¼ 1, 2, . . . , N. T 
Let xðtÞ~ ¼ x~ T1 ðtÞ, . . . , x~ TN ðtÞ , vðtÞ~ ¼ v~T1 ðtÞ, . . . ,
dyðtÞ v~TN ðtÞÞT , f(v(t  ), v(t), t) ¼ ( fT (v 1(t  ), v1(t), t), . .. ,
T
 c1 yðtÞ þ c2 yðt  Þ, t 0, f T(vN (t  ), vN (t), t))) T and ðtÞ ¼ x~ T ðtÞ, v~T ðtÞ .
dt
Then, system (7) can be written as
where constants c1 4 c2 4 0. Then, 8
> _ðtÞ ¼ Fðvðt  Þ, vðtÞ, tÞ þ ðB1  In Þ ðtÞ,
>
>
yðtÞ  j yð0Þj ert , t 0, >
>
< t 2 ½k!, k! þ Þ,
ð8Þ
where j yð0Þj ¼ maxs0 yðsÞ and r is the unique >
> _ðtÞ ¼ Fðvðt  Þ, vðtÞ, tÞ þ ðB2  In Þ ðtÞ,
>
>
solution of >
:
t 2 ½k! þ , ðk þ 1Þ!Þ,
r ¼ c1 þ c2 er :
where
Lemma 5 (Huang, Li, Yu, and Chen 2009): Suppose
that the non-negative function y(t), t 2 [, 1), satisfies Fðvðt  Þ, vðtÞ, tÞ
0 1
0Nn
dyðtÞ B    C
 c1 yðtÞ þ c2 yðt  Þ, t 0, ¼@ 1 A,
dt IN  1NN  In f ðvðt  Þ, vðtÞ, tÞ
N
where c1, c2 are positive constants. Then,
 
ON IN
yðtÞ  j yð0Þj eðc1 þc2 Þt , t 0, B1 ¼ ,
L L
where j yð0Þj ¼ maxs0 yðsÞ.
 326 G. Wen et al.
 
ON IN 1
B2 ¼ T
ðtÞðQ  INn Þ ðtÞ,
ON ON 2
and k 2 N. where
Remark 2: It is not hard to verify that (t) ¼ 02Nn is  
2 ðL þ LT Þ 
an equilibrium point of the switched systems (8). Q¼ :
 
~ and vðtÞ,
Furthermore, by the definitions of (t), xðtÞ ~
one has that the second-order consensus in multi-agent By Lemma 3, Q 4 0 is equivalent to both  4 0 and
system (4) can be achieved if and only if the solution 2 ðL þ LT Þ 4 2 . From condition (1), one obtains
02Nn for system (8) is globally attractive, namely Q 4 0, V(t) 0 and V(t) ¼ 0 if and only if (t) ¼ 02Nn.
k (t)k ! 0, as t ! 1.   Þ ¼
1
PFor
N
notational convenience,
1
PN let vðt
Theorem 1: Suppose that the communication topology N v
j¼1 j ðt  Þ and 
vðtÞ ¼ N v
j¼1 j ðtÞ. For t 2 [k!,
G(A) is strongly connected and balanced, and k! þ ] and arbitrarily given k 2 N, taking the time
Assumption 1 holds. Then, second-order consensus in derivative of V(t) along the trajectories of (8) gives
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

system (4) is achieved if the following conditions hold: _ ¼ T

VðtÞ ðtÞðP  In Þ½Fðvðt  Þ, vðtÞ, tÞ þ ðB1  In Þ ðtÞ
(1) 2 ðL þ L Þ 4 2 ,
T
  
1
(2) 1 ðR1 Þ 4 c012ðQÞ
ðP1 Þ
, ¼ x~ T IN  1NN  In f ðvðt  Þ, vðtÞ, tÞ
N
(3)  4 rþð 3 þ 4 Þ!
rþ 3 þ 4 ,   
1
where þ v~T IN  1NN  In f ðvðt  Þ, vðtÞ, tÞ
N
0  1
2 ðL þ LT Þ  2   1 1  
T
B  2 T C þ ðtÞ PB1 þ BT1 P  In ðtÞ
R1 ¼ @  2 ðL þ L Þ A, 2
 1
 ð 2 þ2 1 Þ  2 ¼ x~ T ðtÞ þ v~T ðtÞ ½ f ðvðt  Þ, vðtÞ, tÞ
 
N ðLþLT Þ   tÞ  x~ T ðtÞ þ v~T ðtÞ
  Þ, vðtÞ,
 1N  f ðvðt
P1 ¼ ,
    
1
 1NN  In f ðvðt  Þ, vðtÞ, tÞ
  N
2 ðL þ LT Þ 
Q¼ ,
  þ x~ T ðtÞ þ v~T ðtÞ ½1N  f ðvðt
  Þ, vðtÞ,
 tÞ
20 2 1
c0 ¼ ( þ ) 2, r is the unique positive solution of  ðL þ LT Þ ON
r ¼  1 þ 2e ,
r 1 ðR1 Þ
1 ¼ 2 ðP1 Þ ,
c0
2 ¼ 1 ðQÞ, 3 ¼ 1 6B ( 2 )C
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ T ðtÞ6 B  ðL þ LT Þ C
c1 þc2 þ ðc1 c2 Þ2 þc3 2 4 @ A
1 ðQÞ , ðþÞc0
4 ¼ 1 ðQÞ , c1 ¼ 22 , c2 ¼ ON
ð 2 þ2 1 Þ þ 2IN
þ  and c3 ¼ ( 1 þ 2(L þ LT))2.
2 3
Proof: Construct the following Lyapunov function 7
candidate  In 7
5 ðtÞ: ð10Þ
1 T
VðtÞ ¼ ðtÞðP  In Þ ðtÞ, ð9Þ
2
~ ¼ ðIN  N1 1NN Þ  In xðtÞ and vðtÞ
Since xðtÞ ~ ¼
where 1
  ðIN  N 1NN Þ  In vðtÞ, one gets
ðL þ LT Þ IN
P¼ :
IN IN x~ T ðtÞ½1N  f ðvðt
  Þ, vðtÞ,
 tÞ ¼ 0,
ð11Þ
It will be shown that V(t) is a valid Lyapunov function v~T ðtÞ½1N  f ðvðt
  Þ, vðtÞ,
 tÞ ¼ 0,
for analysing the error dynamics described by system
(8). According to the Courant–Fischer theorem (Horn and
and Johnson 1985), one has   
1
x~ T ðtÞ 1NN  In f ðvðt  Þ, vðtÞ, tÞ ¼ 0,
 T   N
VðtÞ ¼ x~ ðtÞ ðL þ LT Þ  In xðtÞ
~    ð12Þ
2 1
 v~T ðtÞ 1NN  In f ðvðt  Þ, vðtÞ, tÞ ¼ 0:
þ x~ T ðtÞvðtÞ
~ þ v~T ðtÞvðtÞ
~ N
2
 International Journal of Control 327
  N
Combining (10)–(12), one obtains 2  2 ðL þ LT Þ  X
 kx~ i ðtÞk2
_
VðtÞ
2 i¼1

¼ x~ T ðtÞ þ v~T ðtÞ ½ f ðvðt  Þ, vðtÞ, tÞ X

N
þ 1 kx~ i ðtÞk  kv~i ðtÞk
  Þ, vðtÞ,
 1N  f ðvðt  tÞ i¼1
20 2
1 3

6B  2 ðL þ L Þ
T
ON C 7 ð 2 þ 2 1 Þ þ 2  2 2 ðL þ LT Þ XN

þ T ðtÞ6 B C  In 7 ðtÞ: þ kv~i ðtÞk2

4@ 2 A 5 2
 i¼1
ON  ðL þ LT Þ þ IN
2 ð þ Þ 2 XN
ð13Þ þ kv~i ðt  Þk2
2 i¼1
By Assumption 1, one gets 1
¼ k ðtÞkT ðR1  IN Þk ðtÞk
2
x~ T ðtÞ½ f ðvðt  Þ, vðtÞ, tÞ  1N  f ðvðt
  Þ, vðtÞ,
 tÞ 
þ  ðt  ÞkT ðS1  IN Þk ðt  Þk , ð16Þ
X
N
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

¼  T ½ f ðvi ðt  Þ, vi ðtÞ, tÞ

ðxi ðtÞ  xðtÞÞ where
i¼1 0  1
  Þ, vðtÞ,
 f ðvðt  tÞ 2 ðL þ LT Þ  2   1
B  C
X
N R1 ¼ @ 2 2 ðLþLT Þ A,
kx~ i ðtÞkð 2 kv~i ðt  Þk þ  1
 1 kv~i ðtÞkÞ  ð 2 þ2 1 Þ2
i¼1
"  
X
N X
N 0 0

2
kx~ i ðtÞk þ 2 2
kv~i ðt  Þk 2 S1 ¼ ,
2 2 0 ð þ Þ 2
i¼1 i¼1
#
X
N ~
kxðtÞk ¼ ðkx~ 1 ðtÞk, kx~ 2 ðtÞk, . . . , kx~ N ðtÞkÞT , ~
kvðtÞk ¼
þ 1 kx~ i ðtÞk  kv~i ðtÞk , ð14Þ ðkv~1 ðtÞk, kv~2 ðtÞk, . . . , kv~N ðtÞkÞT , ~
k ðtÞk ¼ ðkxðtÞk T
,
i¼1 ~ T ÞT , and k ðt  Þk ¼ ðkxðt
kvðtÞk ~  ÞkT , kvðt~  ÞkT ÞT .
and On the other hand, one has
1 T
v~T ðtÞ½ f ðvðt  Þ, vðtÞ, tÞ  1N  f ðvðt
  Þ, vðtÞ,
 tÞ VðtÞ ¼ ðtÞðP  In Þ ðtÞ,
2
X
N  T  
¼ x~ ðtÞ ðL þ LT Þ  In xðtÞ
~
¼  T ½ f ðvi ðt  Þ, vi ðtÞ, tÞ
ðvi ðtÞ  vðtÞÞ 2
i¼1 
þ x~ T ðtÞvðtÞ
~ þ v~T ðtÞvðtÞ
~
  Þ, vðtÞ,
 f ðvðt  tÞ 2
1
X
N  T ðtÞðP1  INn Þ ðtÞ, ð17Þ
2
 kv~i ðtÞkð 2 kv~i ðt  Þk þ 1 kv~i ðtÞkÞ
i¼1 where
" #  
X
N X
N
N ðL þ LT Þ 
2 2 2 2
 þ 1 kv~i ðtÞk þ kv~i ðt  Þk : P1 ¼ :
2 i¼1
2 i¼1  
ð15Þ Thus, according to Equation (16) and the following
facts:
Combining (13)–(15) gives
1
_
VðtÞ VðtÞ  2 ðP1 Þ T ðtÞ ðtÞ,
2
2
X
N
ð 2 þ 2 1 Þ
X
N 1
 kx~ i ðtÞk2 þ kv~i ðtÞk2 Vðt  Þ 1 ðQÞ T ðt  Þ ðt  Þ,
2 2 2
i¼1 i¼1
k ðtÞkT R1 k ðtÞk 1 ðR1 Þ T ðtÞ ðtÞ,
ð þ Þ 2 XN XN
þ kv~i ðt  Þk2 þ  1 kx~ i ðtÞk  kv~i ðtÞk k ðt  ÞkT S1 k ðt  Þk  ð þ Þ 2
T
ðt  Þ ðt  Þ,
2 i¼1 i¼1
20 1 3
2 T one obtains
6B 2 ðL þ L Þ O N C 7
þ T ðtÞ6 B C  In 7 ðtÞ _   1 VðtÞ þ
VðtÞ
4@ 2 A 5 2 Vðt  Þ, ð18Þ

ON  ðL þ LT Þ þ IN
2 where 1 ¼ 12 ðR1Þ
ðP1 Þ, 2 ¼ 1cðQÞ
0
and c0 ¼ ( þ ) 2.
 328 G. Wen et al.

For t 2 [k! þ , (k þ 1)!) and arbitrarily given Then, according to (22), one has
k 2 N, taking the time derivative of V(t) along the
trajectories of (8) gives jVðÞj ¼ max VðtÞ  jVð0Þj erðÞ : ð24Þ
t

_ ¼
VðtÞ T
ðtÞðP  In Þ½Fðvðt  Þ, vðtÞ, tÞ þ ðB2  In Þ ðtÞ Combining (23) and (24) yields
  
T 1 VðtÞ  jVðÞj eð 3 þ 4 ÞðtÞ
¼ x~ IN  1NN  In f ðvðt  Þ, vðtÞ, tÞ
N
    jVð0Þj erðÞþð 3 þ 4 ÞðtÞ
,   t 5 !: ð25Þ
T 1
þ v~ IN  1NN  In As V(t) is a continuous function of t, one has
N
T
 f ðvðt  Þ, vðtÞ, tÞ þ ðtÞ½ðPB2 Þ  In ðtÞ, ð19Þ Vð!Þ ¼ lim VðtÞ  jVð0Þj erðÞþð 3 þ 4 Þð!Þ
: ð26Þ
t!!
where
Then,
   
ON IN ON IN jVð!Þj ¼ max VðtÞ
B1 ¼ and B2 ¼ :
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

!t!
L L ON ON
 jVðÞj eð 3 þ 4 Þð!Þ

Similar to the previous analysis, one obtains

 jVð0Þj erðÞþð 3 þ 4 Þð!Þ
¼ jVð0Þj eD ,
2
X
N
ð 2 þ 2 1 Þ X
N
ð27Þ
_ 
VðtÞ kx~ i ðtÞk2 þ kv~i ðtÞk2
2 i¼1
2 i¼1 where D ¼ r(  )  ( 3 þ 4)(!  ) 4 0. For any pos-
ð þ Þ 2 XN
itive integer k, one has
þ kv~i ðt  Þk2
2 i¼1 jVðk!Þj  jVð0Þj ekD : ð28Þ
X
N
T For arbitrary t 4 0, there exists a non-negative integer
þ 1 kx~ i ðtÞk  kv~i ðtÞk þ ðtÞ½ðPB2 Þ  In ðtÞ
i¼1 k, such that k!  t 5 (k þ 1)!. When t 2 [k!, k! þ ),
1 one obtains
 k ðtÞkT ðR2  IN Þk ðtÞk
2   VðtÞ  jVðk!Þj erðtk!Þ
þ  ðt  ÞkT ðS2  IN Þk ðt  Þk , ð20Þ  jVð0Þj ekDrðtk!Þ
where  jVð0Þj ekD
   jVð0Þj eD eð!Þt :
D
ð29Þ
2  1 þ 2 ðL þ LT Þ
R2 ¼ T
,
 1 þ 2 ðL þ L Þ ð 2 þ 2 1 Þ þ 2 When t 2 [k! þ , (k þ 1)!), one has
  VðtÞ  jVðk! þ Þj eð 3 þ 4 Þðtk!Þ
0 0
S2 ¼ :  jVð0Þj ekDr eð 3 þ 4 Þð!Þ
0 ð þ Þ 2
D

It then follows that  jVð0Þj eð 3 þ 4 Þð!ÞrþD

eð!Þt
D
¼ jVð0Þj er eð!Þt : ð30Þ
_  2 ðR2 Þ VðtÞ þ 2 ðS2 Þ Vðt  Þ
VðtÞ
1 ðQÞ 1 ðQÞ Combining (29)–(30) gives
 3 VðtÞ þ 4 Vðt  Þ, ð21Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi VðtÞ  K0 eðD=!Þt , for all t 4 0, ð31Þ
2
c1 þc2 þ ðc1 c2 Þ þc3
where 3 ¼ , c0
4 ¼ 1 ðQÞ, c1 ¼ 22 ,
ð 2 þ2 1 Þ
1 ðQÞ
T 2 where K0 ¼ eDjV(0)j , which indicates that the states of
c2 ¼ 2 þ , c3 ¼ ( 1 þ 2(L þ L )) and
agents exponentially converge to consensus. This
c0 ¼ ( þ ) 2. Based on the above analysis and
completes the proof.
according to Lemma 4, one obtains
Corollary 1: Suppose that the communication topology
VðtÞ  jVð0Þj ert , 0  t 5 , ð22Þ G(A) is a strongly connected and balanced, and
where r is the unique positive solution of r ¼ Assumption 1 holds. Then, second-order consensus in
 1 þ 2er , jVð0Þj ¼ maxs0 VðsÞ. For  5 t 5 !, system (4) is achieved if the following conditions hold:
by using Lemma 5, one obtains (1)  4 ,
(2) 2(L þ LT) 4 max{1, %1, %2},
VðtÞ  jVðÞj eð 3 þ 4 Þt
: ð23Þ (3)  4 rþð 3 þ 4 Þ!
rþ 3 þ 4 ,
 International Journal of Control 329
T
where %1 ¼  2þ 2
1
þ 2 ðþÞð N ðLþL Þþ1Þ
ðÞ , %2 ¼ 1
T
ð 2 þ2 1 Þþ2 ðþÞð ðLþL ÞþÞ
2
þ 2 N
2 ðÞ
, r is the unique positive
r minf 1 , 2 g
solution of r ¼  þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 2 e , 1 ¼ N ðLþLT Þþ, 2 ¼
ðþÞ 2 c1 þc2 þ ðc1 c2 Þ2 þc3 ðþÞ 2 2
 , 3 ¼  , 4 ¼  , 1 ¼  2(L þ 3 2
LT)   2   1, 2 ¼ 22(L þ LT)  ( 2 þ 2 1)  2,
Figure 1. The fixed directed communication topology G(A).
c1 ¼ 22 , c2 ¼ ð 2 þ22 1 Þ þ  and c3 ¼ ( 1 þ
2(L þ LT))2.
With communication
Proof: Construct the same Lyapunov function can-
Without communication
didate V(t) similar to the proof of Theorem 1. By the
Geršgorin disc theorem (Horn and Johnson 1985), the 0.425 0.425 0.425
corollary can be proved by following the proof of
Theorem 1. …
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

Remark 3: In Yu et al. (2010b), the concept of

general algebraic connectivity a(L) is introduced to 0 0.5 1.0 1.5 … time

describe the second-order multi-agent system’s ability Figure 2. Intermittent communication.

to reach consensus. By Definition 6 in Yu et al.
T
(2010b), one has aðLÞ ¼ 2 ðLþL
2
Þ
for a strongly con-
nected and balanced G(A), where L is the Laplacian 2

matrix of the graph. Suppose that  4 . Then, from Trajectory of x1(t)

1
Trajectory of x (t)
Corollary 1, the second-order consensus can be 2
Trajectory of x3(t)
State trajectories of agents

achieved if the general algebraic connectivity a(L) 0

and the communication time duration  are larger than

−1
their corresponding threshold values, respectively.
Remark 4: Suppose that consensus in multi-agent −2

system (4) can be achieved for  ¼  0, where  0 4 0.

−3
Then, it can be concluded from Theorem 1 that consen-
sus in multi-agent system (4) can be achieved for all −4
 2 [0,  0], i.e. the consensus achievement is robust to
 2 [0,  0]. −5
0 2 4 6 8 10 12
t
Remark 5: It can be observed that Theorem 1 is
proved by using a common Lyapunov function Figure 3. Consensus of state trajectories of agents.
approach. Thus, it is interesting and important to
further study how to reduce the conservatism of the 1 ðR1 Þ ¼ 5:3970 4 c012ðQÞ ðP1 Þ
¼ 0:025 and  ¼ 0:425 4
rþð 3 þ 4 Þ!
theoretical results in the future. rþ 3 þ 4 ¼ 0:4232. Therefore, second-order consensus
can be achieved in multi-agent system (4). The position
and velocity states of all agents are, respectively, shown
in Figures 3 and 4, with initial conditions x1 ðtÞ ¼ 1:5,
4. A simulation example x2 ðtÞ ¼ 0:10, x3 ðtÞ ¼ 2:02, v1 ðtÞ ¼ 0:05, v2 ðtÞ ¼ 0:15,
In this section, a simulation example is provided to v3 ðtÞ ¼ 1:10, for all t 2 [, 0]. Furthermore, accord-
 T
verify the theoretical analysis. ing to the definition of (t), i.e. ðtÞ ¼ x~ T ðtÞ, v~T ðtÞ ,
Consider the second-order consensus protocol T
with xðtÞ~ ¼ ðx~ 1 ðtÞ, x~ 2 ðtÞ, x~ 3 ðtÞÞ and vðtÞ
~ ¼ ðv~1 ðtÞ, v~2 ðtÞ,
with time delayed nonlinear velocities in system (4), v~3 ðtÞÞT , it is reasonable to denote k (t)k as the consen-
where the communication topology among the mul- sus errors of multi-agent system (4). The evolution
tiple agents is shown in Figure 1 with weights on trajectory of k (t)k is given in Figure 5. It can be seen
each edge being 0.2. Let the time delayed nonlinear that the consensus problem in system (4) is indeed
function be f(vi(t  ), vi(t), t) ¼ 0.001cos(vi(t)) þ solved and the simulation results verify the theoretical
0.001sin(vi(t  )) 2 R, where vi(t) 2 R,  ¼ 0.01, and analysis very well.
i ¼ 1, 2, 3. In view of Assumption 1, one obtains
1 ¼ 0.001, 2 ¼ 0.001. Let  ¼ 3,  ¼ 5,  ¼ 0.425 and
! ¼ 0.500. Namely, the communication rate among the 5. Conclusions
dynamic agents is 85%, see Figure 2 for illustration. In this article, a kind of intermittent consensus
By Theorem 1, one has 2 ðL þ LT Þ ¼ 0:6 4 2 ¼ 0:12, protocols for second-order multi-agent systems with
 330 G. Wen et al.
0.6
61225013 and 61104145, the Natural Science Foundation of
Trajectory of v (t)
0.4 1 Jiangsu Province of China under Grant No. BK2011581, the
Trajectory of v2(t)
Research Fund for the Doctoral Program of Higher
0.2 Trajectory of v3(t)
Education of China under Grant No. 20110092120024, the
Velocity trajectories of agents

0 Fundamental Research Funds for the Central Universities of

China, and the Information Processing and Automation
−0.2
Technology Prior Discipline of Zhejiang Province-Open
−0.4 Research Foundation under Grants 20120801 and 20120802.
−0.6

−0.8
References
−1

−1.2 Amster, P., and Mariani, M.C. (2008), ‘Some Results on the
−1.4 Forced Pendulum Equation’, Nonlinear Analysis, 68,
0 2 4 6 8 10 12
t 1874–1880.
Beard, R.W., McLain, T.W., Goodrich, M.A., and
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

Figure 4. Consensus of velocity trajectories of agents. Anderson, E.P. (2002), ‘Coordinated Target Assignment
and Intercept for Unmanned Air Vehicles’, IEEE
Transactions on Robotics Automation, 18, 911–922.
3 Boyd, S., Ghaoui, L.E., Ferion, E., and Balakrishnan, V.
(1994), Linear Matrix Inequalities in System and Control
2.5 Theory, Philadelphia, PA: SIAM.
Fax, A., and Murray, R.M. (2004), ‘Information Flow and
Consensus error ||ζ(t)||

2 Coorperative Control of Vehicle Formations’, IEEE

Transactions on Automatic Control, 49, 1465–1476.
Fujisaka, H., and Yamada, T. (1983), ‘Stability Theory of
1.5
Synchronized Motion in Coupled-oscillator Systems’,
Progress of Theoretical Physics, 69, 32–47.
1
Hong, Y., Chen, G., and Bushnell, L. (2008), ‘Distributed
Observers Design for Leader-following Control of Multi-
0.5 agent Networks’, Automatica, 44, 846–850.
Hong, Y., Hu, J., and Gao, L. (2006), ‘Tracking Control for
0 Multi-agent Consensus with an Active Leader and
0 2 4 6 8 10 12
t Varibale Topology’, Automatica, 42, 1177–1182.
Horn, R.A., and Johnson, C.R. (1985), Matrix Analysis,
Figure 5. Consensus error k (t)k. Cambridge, UK: Cambridge University Press.
Huang, T., Li, C., Yu, W., and Chen, G. (2009),
‘Synchronization of Delayed Chaotic Systems with
time delayed nonlinear dynamics and fixed directed Parameter Mismatches by using Intermittent Linear State
communication topology has been introduced and Feedback’, Nonlinearity, 22, 569–584.
studied. It has been theoretically shown that second- Jadbabaie, A., Lin, J., and Morse, A.S. (2003), ‘Coordination
order consensus can be reached if the communication of Groups of Mobile Autonomous Agents using Nearest
time duration and the general algebraic connectivity Neighbour Rules’, IEEE Transactions on Automatic
Control, 48, 988–1001.
are respectively larger than their corresponding
Li, Z., Duan, Z., Xie, L., and Liu, X. (2012), ‘Distributed
thresholds depending both on the general algebraic
Robust Control of Linear Multi-agent Systems with
connectivity of the topology and the agents’ inherent Parameter Uncertainties’, International Journal of
dynamics. Future work will be focused on the consen- Control, 85, 1039–1050.
sus behaviours of more practical models, such as Li, Z., Liu, X., and Fu, M. (2011), ‘Global Consensus
second-order multi-agent systems with nonlinear Control of Lipschitz Nonlinear Multi-agent Systems’, in
dynamics and quantisation effects, higher-order Proceedings of the 18th IFAC World Congress, Milano,
multi-agent systems with nonlinear dynamics and Italy, pp. 10056–10061.
so on. Li, X., Wang, X., and Chen, G. (2004), ‘Pinning a Complex
Dynamical Network to Its Equilibrium’, IEEE
Transactions on Circuits and Systems I: Regular Papers,
51, 2074–10061.
Acknowledgements Olfati-Saber, R., and Murray, R.M. (2004), ‘Consensus
This work was supported by the Hong Kong Research Problems in Networks of Agents with Switching Topology
Grants Council under the GRF Grant CityU 1114/11E, the and Time-delays’, IEEE Transactions on Automatic
National Nature Science Foundation of China under Grants Control, 49, 1520–2087.
 International Journal of Control 331

Ren, W. (2008), ‘Synchronization of Coupled Harmonic Intermittent Information Transmissions’, International

Oscillators with Local Interaction’, Automatica, 44, Journal of Control, 85, 384–396.
3195–1533. Wen, G., Duan, Z., Li, Z., and Chen, G. (2012b),
Ren, W., and Atkins, E. (2007), ‘Distributed Multi-vehicle ‘Flocking of Multi-agent Dynamical Systems with
Coordinated Control via Local Information Exchange’, Intermittent Nonlinear Velocity Measurements’,
International Journal of Robust and Nonlinear Control, 17, International Journal of Robust and Nonlinear Control,
1002–3200. doi: 10.1002/rnc.1784.
Ren, W., and Atkins, E. (2008), ‘On Consensus Algorithms Wen, G., Duan, Z., Su, H., Chen, G., and Yu, W. (2012c),
for Double-integrators Dynamics’, IEEE Transactions on ‘A Connectivity-preserving Flocking Algorithm for Multi-
Automatic Control, 58, 1503–1033. agent Dynamical Systems with Bounded Potential
Ren, W., and Beard, R.W. (2005), ‘Consensus Seeking in Function’, IET Control Theory and Applications, 6,
Multiagent Systems Under Dynamically Changing 813–821.
Interaction Topologies’, IEEE Transactions on Automatic Wen, G., Duan, Z., Yu, W., and Chen, G. (2012d),
Control, 50, 655–661. ‘Consensus in Multi-agent Systems with Communication
Ren, W., Beard, R.W., and Atkins, E.M. (2007), Constraints’, International Journal of Robust and Nonlinear
‘Information Consensus in Multivehicle Cooperative Control, 22, 170–182.
Downloaded by [University of California, San Diego] at 21:12 07 January 2015

Control’, IEEE Control Systems Magazine, 27, 71–82. Xie, G., and Wang, L. (2005), ‘Consensus Control for a Class
Su, H., Chen, G., Wang, X., and Lin, Z. (2011), of Networks of Dynamic Agents: Fixed Topology’,
‘Adaptive Second-order Consensus of Networked Mobile in Proceedings of the 44th IEEE Conference on Decision
Agents with Nonlinear Dynamics’, Automatica, 47, and Control and the 2005 European Control Conference,
368–375. Seville, Spain, pp. 96–101.
Toner, J., and Tu, Y. (1998), ‘Flocks, Herds, and School: A Yu, W., Chen, G., Wang, Z., and Yang, W. (2009),
Quantitative Theory of Flocking’, Physics Review E, 58, ‘Distributed Consensus Filtering in Sensor Networks’,
4828–4858. IEEE Transactions on Systems, Man, and Cybernetics-Part
Topaz, C.M., and Bertozzi, A.L. (2005), ‘Swarming Patterns B, 39, 1568–1577.
in a Two-dimensional Kinematic Model for Biological Yu, W., Chen, G., and Cao, M. (2010a), ‘Some Necessary
Groups’, SIAM Journal of Applied Mathematics, 65, and Sufficient Conditions for Second-order Consensus in
152–174. Multi-agent Dynamical Systems’, Automatica, 46,
Vicsek, T., Cziroók, A., Ben-Jacob, E., Cohen, O., and 1089–1095.
Shochet, I. (1995), ‘Novel Type of Phase Transition in a Yu, W., Chen, G., Cao, M., and Kurths, J. (2010b), ‘Second-
System of Self-driven Particles’, Physics Review Letters, 75, order Consensus for Multi-agent Systems with Directed
1226–1229. Topologies and Nonlinear Dynamics’, IEEE Transactions
Wang, J., Tan, Y., and Mareels, I. (2009), ‘Robustness on Systems, Man, and Cybernetics-Part B: Cybernetics, 40,
Analysis of Leader-follower Consensus’, Journal of System 881–891.
Sciences and Complexity, 22, 186–206. Zhang, H., and Lewis, F.L. (2012), ‘Adaptive Cooperative
Wen, G., Duan, Z., Li, Z., and Chen, G. (2012a), ‘Consensus Tracking Control of Higher-order Nonlinear Systems with
and Its L2-gain Performance of Multi-agent Systems with Unknown Dynamics’, Automatica, 48, 1432–1439.