International Journal of Automation and Computing 11(3), June 2014, 340-346

DOI: 10.1007/s11633-014-0798-1

Consensus Problem of Heterogeneous Multi-agent

Systems with Time Delay under Fixed
and Switching Topologies
Ping-Ping Dai Cheng-Lin Liu Fei Liu
Key Laboratory of Advanced Process Control for Light Industry (Ministry of Education), Jiangnan University, Wuxi 214122, China

Abstract: Consensus problem is investigated for heterogeneous multi-agent systems composed of first-order agents and second-order
agents in this paper. Leader-following consensus protocol is adopted to solve consensus problem of heterogeneous multi-agent systems
with time-varying communication and input delays. By constructing Lyapunov-Krasovkii functional, sufficient consensus conditions in
linear matrix inequality (LMI) form are obtained for the system under fixed interconnection topology. Moreover, consensus conditions
are also obtained for the heterogeneous systems under switching topologies with time delays. Simulation examples are given to illustrate
effectiveness of the results.

Keywords: Consensus problem, heterogeneous multi-agent systems, switching topologies, communication delay, input delay.

1 Introduction systems, and two classes of consensus protocols with and

without velocity measurements were proposed by combin-
In the past decade, consensus problem, which means that ing the homogeneous domination method with the addi-
each agent reaches an agreement on a common value asymp- tion of power integrator method. Liu and Liu[24] investi-
totically by cooperating with their neighbors, has attracted gated consensus problem for discrete-time heterogeneous
much attention in several areas for its broad engineering multi-agent systems composed of first-order and second-
applications, such as automated highway systems, air traf- order agents, and obtained the sufficient consensus crite-
fic control, congestion control in communication networks, ria for the agents with bounded communication delays un-
etc. der fixed topology and switching topologies, respectively.
In the multi-agent systems, each agent needs certain in- Moreover, Yin et al.[25] investigated the consensus problem
put delay for itself to accept and process information, and for a set of discrete-time heterogeneous multi-agent systems
communication delays between neighboring agents are non- and designed a consensus algorithm based on the event-
negligible when the agents exchange information. So far, triggered communication scheme, where communication is
two types of consensus protocols with communication de- only needed between the agent and its local neighbors. In
lays have been extensively adopted. One is that both the addition, a sufficient condition is obtained to guarantee the
state of the agent and its neighbors are affected by identi- consensus of heterogeneous multi-agent systems in terms
cal delays, while another is that delays only affect the state of linear matrix inequality (LMI) based on the Lyapunov
received from neighbors of the agents. functional method and the Kronecker product technique.
Up to now, there has been much research on consen- Tian and Zhang[26] studied the high-order consensus prob-
sus problem of first-order[1−6] , second-order[7−19] , and high- lem for heterogeneous multi-agent systems with unknown
order[20, 21] multi-agent systems with identical agents dy- communication delays, and got a necessary and sufficient
namics. In some practical situations, however, the agent of consensus condition. Munz et al.[27] considered the multi-
multi-agent systems has different dynamics for various con- agent systems with agents dynamics given by strictly stable
straints. Hence, consensus problem of heterogeneous multi- linear systems with nonuniform communication delays, and
agent systems has attracted many researchers interests. set-valued consensus conditions have been acquired based
For heterogeneous multi-agent systems composed of first- on the properties of the frequency-domain feedback matrix.
order and second-order integrator agents, Zheng et al.[22] In this paper, a general stationary consensus protocol is
discussed the consensus problem under the linear consen- adopted for heterogeneous multi-agent systems with input
sus protocol and the saturated consensus protocol, and the and communication delay to follow a static leader. Firstly,
sufficient conditions were established when the communi- consensus is analyzed for the system with time-varying
cation topologies were undirected connected graphs and input delay and communication delay under fixed topol-
leader-following networks. Zheng and Wang[23] studied the ogy. By constructing Lyapunov-Krasovskii functional, suf-
finite-time consensus problem of heterogeneous multi-agent ficient consensus condition is obtained and expressed in LMI
form. Secondly, a delay-dependent consensus condition in
Brief Paper the form of LMI is also obtained for the system with time-
Manuscript received September 9, 2013; revised December 19, 2013
This work was supported by National Natural Science Foundation varying input delay and communication delay converging
of China (Nos. 61104092, 61134007 and 61203147), the Priority Aca- to the leader s states under switching topologies. By using
demic Program Development of Jiangsu Higher Education Institu- LMI toolbox in Matlab software, the bound of time delay
tions.
 P. P. Dai et al. / Consensus Problem of Heterogeneous Multi-agent Systems with · · · 341

can be determined based on LMI. of system (1) is

ẋi (t) = vi (t),
2 Problem description

n
v̇i (t) = −k2 vi (t − τ (t)) + k1 aij (xj (t − η(t) − τ (t))−
In this paper, it is assumed that the heterogeneous multi-
j=1
agent system is composed of m second-order agents and n-m
first-order agents. The dynamics of heterogeneous multi- xi (t − τ (t))) + bi (x0 − xi (t)), i = 1, 2, · · · , m,
agent systems are given as 
n

⎧ ẋi (t) = aij (xj (t − η(t) − τ (t)) − xi (t − τ (t)))+

⎪
⎨ ẋi (t) = vi (t), j=1

v̇i (t) = ui (t), i = 1, 2, · · · , m (1) bi (x0 − xi (t)), i = m + 1, · · · , n. (4)

⎪
⎩ ẋ (t) = u (t), i = m + 1, · · · , n
i i
Define x̄i (t) = xi (t) − x0 , and we can get
where xi ∈ R, vi ∈ R are the position, velocity and ui ∈ R x̄˙ s (t) = vs (t),
is the control input respectively of agent i. v̇s (t) = −k2 Im vs (t − τ (t)) + k1 [Ass x̄s (t − η(t) − τ (t))+
The interconnection topology of multi-agent systems is
usually described as directed graph (digraph). Let G = Asf x̄f (t − η(t) − τ (t)) − Ds x̄s (t − τ (t))] − Bs x̄s (t),
(V, E, A) as the weighted digraph, V = {1, · · · , n} is a set of ˙x̄f (t) = Af s x̄s (t − η(t) − τ (t)) + Af f x̄f (t − η(t) − τ (t))−
vertices, E ⊆ V × V is a set of edges, and A = [aij ] ∈ Rn×n Df x̄f (t − τ (t)) − Bf x̄f (t) (5)
is a weighted adjacency matrix such that aij > 0 and aii = 0
for all i = 1, · · · , n. The node indexes n belong to a finite where x̄s = [x̄1 , x̄2 , · · · , x̄m ], vs = [v1 , v2 , · · · , vm ], and
index set Λ = {1, · · · , n}. The set of neighbors of node i is x̄f = [x̄m+1 , · · · , x̄n ]. Ass ∈ Rm×m is the adja-
denoted by Ni = {j ∈ V (i, j) ∈ E}. The Laplacian matrix cency matrix of second-order agents between each other,
of the weighted digraph G is defined as L = D − A, where Af f ∈ R(n−m)×(n−m) is the adjacency matrix of first-
 order agents between each other, Asf ∈ Rm×(n−m) and
D = diag{ n j=1 aij , i ∈ Λ} is the degree matrix of G.
Based on the protocols in [8, 15], we proposed a consen- Af s ∈ R(n−m)×m is adjacency matrix between first-order
sus protocol with a static leader as follows and second-order agents, and Bs = diag{b1 , · · · , bm }, Bf =
diag{bm+1 , · · · , bn }.

n
Then, system (5) can be rewritten as
ui = −k2 vi (t) + k1 aij (xj (t) − xi (t))+
j=1 ẏ(t) = H1 y(t) + H2 y(t − τ (t)) + H3 y(t − η(t) − τ (t)) (6)
bi (x0 − xi (t)), i = 1, 2, · · · , m (2) where

n ⎡ ⎤
ui = aij (xj (t) − xi (t)) + bi (x0 − xi (t)), 0 Im 0
⎢ ⎥
j=1 y T (t) = [x̄s (t), vs (t), x̄f (t)]T , H1 = ⎣ −Bs 0 0 ⎦,
i = m + 1, · · · , n 0 0 −Bf
⎡ ⎤
where k1 > 0, k2 > 0, aij > 0, j ∈ Ni is the adjacent 0 0 0
⎢ ⎥
element of neighbors matrix A. x0 ∈ R is a constant that H2 = ⎣ −k1 Ds −k2 Im 0 ⎦
denotes the position of the static leader, and the linked 0 0 −Df
weight bi from agent i to the leader is defined as
and ⎡ ⎤
 0 0 0
bi > 0, if agent is connected to the leader, ⎢ ⎥
H3 = ⎣ −k1 Ass 0 k1 Asf ⎦ .
bi = 0, otherwise.
Af s 0 Af f
In this paper, the interconnection topology composed of Therefore, the consensus convergence problem of multi-
the n agents and the static leader is denoted by Ḡ. agent system (4) is equivalent to the asymptotic stability of
Under the time-varying input delay τ (t) > 0 and the system (6).
communication delay η(t) > 0 between each agent, the pro-
tocol (2) can be written as 3 Consensus analysis

n
First of all, the assumptions about the input delay and
ui = −k2 vi (t − τ (t)) + k1 aij (xj (t − η(t) − τ (t))− communication delay are made as follows.
j=1
Assumption 1. 0  τ (t)  σ, 0  τ̇ (t)  μ < 1, σ > 0,
xi (t − τ (t))) + bi (x0 − xi (t)), i = 1, 2, · · · , m (3) t  0.

n Assumption 2. 0  η(t)  h, 0  η̇(t)  d < 1, h > 0.
ui = aij (xj (t − η(t) − τ (t)) − xi (t − τ (t)))+ In the proof of the main results, moreover, the following
j=1 lemma is needed[28] .
bi (x0 − xi (t)), i = m + 1, · · · , n. Lemma 1. Assume that 0  η(t)  h, h > 0 is piece-
wise continuous function. For arbitrary differentiable vec-
With the consensus algorithm (3), the closed-loop form tor function x(t) : [−h, ∞) → Rn and arbitrary positive
342 International Journal of Automation and Computing 11(3), June 2014

definite matrix M ∈ Rn×n , the following inequality is es- Calculating the derivative of V (t), we get
tablished:
V̇ (t) = ẏ T (t)P y(t) + y T (t)P ẏ(t) + (1 − μ)y T (t − τ (t))
   
t
T
t (Q2 − Q1 )y(t − τ (t)) + (d + μ − 1)y T (t − η(t)−
ẋ (s)ds M ẋ(s)ds 
t−η(t) t−η(t) τ (t))Q2 y(t − η(t) − τ (t)) + σ ẏ T (t)R1 ẏ(t)−
 t
σ −1 (y T (t) − y T (t − τ (t)))R1 (y(t) − y(t − τ (t)))+
h ẋT (s)M ẋ(s)ds, t  0.
t−h hẏ T (t)R1 ẏ(t) − h−1 (y T (t − τ (t)) − y T (t − η(t)−
τ (t)))R1 (y(t − τ (t)) − y(t − η(t) − τ (t))).
3.1 Fixed topology
Based on Lemma 1, we have
In this section, consensus is analyzed for system (4) with
time-varying input delay and time-varying communication V̇ (t)  ϕT (t)M ϕ(t)
delay under fixed topology.
Theorem 1. τ (t) and η(t) satisfy Assumptions 1 and 2, where ϕT (t) = [y T (t), y T (t−τ (t)), y T (t−η(t)−τ (t))] and M
respectively, and the interconnection topology Ḡ of multi- is defined as (7). Therefore, if (7) is established, the system
agent system (4) has the leader as a globally reachable (6) is asymptotically stable, i.e., the agents in system (4)
node. Then, multi-agent system (4) achieves an asymptotic achieve an asymptotic consensus.
consensus such that limt→∞ xi (t) = x0 , limt→∞ vi (t) = 0, In practice, the input delay between the controller and
i ∈ Λ, if the following LMI actuator usually changes relatively slowly, but the deriva-
tive of communication delay may be greater than one or un-
⎡ ⎤ known for various communication constraints. When τ (t)
M11 M12 M13 satisfies Assumption 1 and the derivative of η(t) is not less
⎢ ⎥
⎣ M22 M23 ⎦ < 0 (7) than one or unknown, we choose the following Lyapunov-
∗ M33 Krasovskii functional
 0  t
V (t) = y T (t)P y(t) + ẏ T (s)R1 ẏ(s)dsdθ+
holds, where −σ t+θ
 0  t
ẏ T (s)R2 ẏ(s)dsdθ.
M11 = H1T P + P H1 + Q1 + σH1T R1 H1 + hH1T R2 H1 − −h−σ t+θ

σ −1 R1 Thus, we can obtain the following consensus condition.

M12 = P H2 + σH1T R1 H2 + hH1T R2 H2 + σ −1 R1 
Theorem 2. τ (t) satisfies Assumption 1, and d  1,
M13 = P H3 + σH1T R1 H3 + hH1T R2 H3 or the derivative of η(t) is unknown, and the interconnec-
M22 = (1 − μ)(Q2 − Q1 ) + σH2T R1 H2 + hH2T R2 H2 − tion topology Ḡ of multi-agent system (4) has the leader
as a globally reachable node. The agents in system (4)
σ −1 R1 − h−1 R2
achieve an asymptotic consensus, i.e., limt→∞ xi (t) = x0 ,
M23 = σH2T R1 H3 + hH1T R2 H3 + h−1 R2 limt→∞ vi (t) = 0, i ∈ Λ, if the following LMI
M33 = (d + μ − 1)Q2 + σH3T R1 H3 + hH3T R2 H3 − ⎡ ⎤
Mf 11 Mf 12 Mf 13
h−1 R2 ⎢ ⎥
⎣ Mf 22 Mf 23 ⎦ < 0
∗ Mf 33
and P, Qj , Rj , j = 1, 2, 3 are positive definite matrices with
appropriate dimensions. holds, where
Proof. Choose a Lyapunov-Krasovskii functional for the Mf 11 = H1T P + P H1 + σH1T R1 H1 + (h + σ)H1T R2 H1 +
system (6) as follows:
σ −1 R1 − (h + σ)−1 R2

V (t) = V1 (t) + V2 (t) + V3 (t) Mf 12 = P H2 + σH1T R1 H2 + (h + σ)H1T R2 H2 + σ −1 R1

V1 (t) = y T (t)P y(t) Mf 13 = P H3 + σH1T R1 H3 + (h + σ) + H1T R2 H3 +

 t (h + σ)−1 R2
V2 (t) = y T (s)Q1 y(s)ds+
t−τ (t) Mf 22 = σH2T R1 H2 + (h + σ)H2T R2 H2 − σ −1 R1
 t−τ (t)
y T (s)Q2 y(s)ds Mf 23 = σH2T R1 H3 + (h + σ)H1T R2 H3


t−η(t)−τ (t)
 Mf 33 = σH3T R1 H3 + (h + σ)H3T R2 H3 − (h + σ)−1 R2
0 t
V3 (t) = ẏ T (s)R1 ẏ(s)dsdθ+ and P, Qj , Rj j = 1, 2, 3 are positive definite matrices with
−σ t+θ
 −σ  t appropriate dimensions.
ẏ T (s)R2 ẏ(s)dsdθ. The proof of Theorem 2 is similar to that of Theorem 1,
−h−σ t+θ so we omit it to contain the length of this paper.
 P. P. Dai et al. / Consensus Problem of Heterogeneous Multi-agent Systems with · · · 343

3.2 Switching interconnection topology hold for w = 1, 2, · · · , N , where

The structure of the interconnection topology will change Ms11 = H1wT P + P H1w + Q1 + Q2 + σH1T R1 H1w +
as time changes due to the addition of new links or the ob-
stacles in links. For example, the communication range of (h + σ)H1wT R2 H1w − σ −1 R1 − (h + σ)−1 R2
agents is limited, or the agents change communication links Ms12 = P H2w + σH1wT R1 H2 +
and coupling weights. In this section, we investigate the
consensus problem of heterogeneous multi-agent systems (h + σ)H1wT R2 H2w + σ −1 R1
with switching topologies. Here, we assume that the num- Ms13 = P H3w + σH1wT R1 H3w + (h + σ) + H1wT R2 H3w +
ber of switching topologies is finite. The set of topologies
(h + σ)−1 R2
is defined as Ω = {Ḡw , w = 1, 2, · · · , N }, and each digraph
Ḡw has the leader as the globally reachable node. The sys- Ms22 = (μ − 1)Q1 + σH2wT R1 H2w + (h + σ)H2wT R2 H2w −
tem (4) with time-varying delays and switching topologies σ −1 R1
can be represented as
Ms23 = σH2wT R1 H3w + (h + σ)H1wT R2 H3w
ẋs (t) = vs (t)
w
Ms33 = (d + μ − 1)Q2 + σH3T R1 H3w +
v̇s (t) = −k2 Im vs (t − τ (t)) + k1 [Aw
ss xs (t − η(t) − τ (t))+
(h + σ)H3wT R2 H3w − (h + σ)−1 R2
Aw w
sf xf (t − η(t) − τ (t)) − Ds xs (t − τ (t))]+

Bsw (x0 1m − xs (t)) and P, Qj , Rj , j = 1, 2, 3 are positive definite matrices with

ẋf (t) = Aw w
f s xs (t − η(t) − τ (t)) + Af f xf (t − η(t) − τ (t))− appropriate dimensions.
Proof. Choosing a Lyapunov-Krasovskii functional for
Dfw xf (t − τ (t)) − Bfw (xf (t) − x0 1n ). (8)
system (10)
Taking the variable transformation x̄i (t) = xi (t) − x0 , we
V (t) = V1 (t) + V2 (t) + V3 (t),
can get
V1 (t) = y T (t)P y(t),
x̄˙ s (t) = vs (t)  t
w
v̇s (t) = −k2 Im vs (t − τ (t)) + k1 [Aw
ss x̄s (t − η(t) − τ (t))− V2 (t) = y T (s)Q1 y(s)ds+
t−τ (t)
Aw w w
sf x̄f (t − η(t) − τ (t)) − Ds x̄s (t − τ (t))] − Bs x̄s (t)  t
x̄˙ f (t) = Aw w
f s x̄s (t − η(t) − τ (t)) + Af f x̄f (t − η(t) − τ (t))−
y T (s)Q2 y(s)ds,
t− eta(t)−τ (t)
Dfw x̄f (t − τ (t)) − Bfw x̄f (t). (9)  0  t
V3 (t) = ẏ T (s)R1 ẏ(s)dsdθ+
System (9) can be rewritten as −σ t+θ
 0  t
ẏ(t) = H1w y(t) + H2w y(t − τ (t)) + H3w y(t − η(t) − τ (t)) ẏ T (s)R2 ẏ(s)dsdθ.
(10) −h−σ t+θ

where Calculating the derivative of V (t), we get

⎡ ⎤
0 w
Im 0 V̇ (t) = ẏ T (t)P y(t) + y T P ẏ(t) + y T (t)(Q1 + Q2 )y(t)+
T T ⎢ ⎥
y (t) = [x̄s (t), vs (t), x̄f (t)] , H1w=⎣ −Bsw 0 0 ⎦, (μ − 1)y T (t − τ (t))Q1 y(t − τ (t)) + (d + μ − 1)y T
w
0 0 −Bf
(t − η(t) − τ (t))Q2 y(t − η(t) − τ (t)) + σ ẏ T (t)
⎡ ⎤
0 0 0 R1 ẏ(t) − σ −1 (y T (t) − y T (t − τ (t)))R1 (y(t)−
⎢ ⎥
H2w = ⎣ −k1 Dsw w
−k2 Im 0 ⎦, y(t − τ (t))) + (h + σ)ẏ T (t)R2 ẏ(t)−
w
0 0 −Df
(h + σ)−1 (y T (t − τ (t)) − y T (t − η(t) − τ (t)))
and ⎡ ⎤ R2 (y(t − τ (t)) − y(t − η(t) − τ (t))).
0 0 0
⎢ ⎥
H3w = ⎣ −k1 Awss 0 k1 A w
sf ⎦ . From Lemma 1, we have
Awfs 0 w
Af f
V̇ (t)  ϕT (t)M ϕ(t) (12)
Theorem 3. τ (t) and η(t) satisfy Assumptions 1 and
2, and the interconnection topology Gw of multi-agent sys- where ϕT (t) = [y T (t), y T (t−τ (t)), y T (t−η(t)−τ (t))] and M
tems (8) has the leader as a globally reachable node. Then, is defined as (11). Therefore, if (11) is established, the sys-
multi-agent systems (8) reach an asymptotic consensus, i.e., tem (10) is asymptotically stable, i.e., the agents in system
limt→∞ xi (t) = x0 , limt→∞ vi (t) = 0, i ∈ Λ, if the following (8) reach an asymptotic consensus.
LMIs Remark 1. If τ (t) satisfies Assumption 1, and d  1,
⎡ ⎤
Ms11 Ms12 Ms13 or the derivative of η(t) is unknown, we can use the same
⎢ ⎥ Lyapunov-Krasovskii functional as that in Theorem 2. to
⎣ Ms22 Ms23 ⎦ < 0 (11)
get the consensus condition for the system with switching
∗ Ms33 topologies, but we omit it here to avoid repetition.
344 International Journal of Automation and Computing 11(3), June 2014

4 Simulation results
In this paper, we assume that the multi-agent system is
composed of two second-order agents and three first-order
agents and a static leader, and the leader is the globally
reachable node of the topology.
Example 1. The interconnection topology of system (4)
is given in Fig. 1. The node 0 is the static leader, and the
nodes 1, 2 and the nodes 3, 4, 5 denote second-order agents
and first-order agents respectively. We assume that the
weights of edges are: a12 = 1, a23 = 0.2, a34 = 0.2, a45 =
0.3, a52 = 0.3, b5 = 1.5. and the control gains are chosen as
k1 = 0.1 and k2 = 1. The position of static leader is chosen
as x0 = 1, and the initial position and velocity of each node
are generated randomly.
Fig. 3 Agents with d > 1 under fixed topology

Example 2. Consider a network of five agents and a

static leader given by (8). The interconnection topology of
the system is switched between the topology 1 in Fig. 4 and
the topology 2 in Fig. 5. In the simulation, the intercon-
nection topology is switched from topology 1 to topology 2
every 5 s. In topology 1, we choose the weights as: a12 =
0.2, a32 = 0.3, a42 = 0.05, a45 = 0.05, a52 = 0.3, b2 = 1.2.
In topology 2, we choose the weights as: a15 = 0.1, a21 =
Fig. 1 Interconnection topology with five agents and a leader 0.1, a32 = 0.3, a43 = 0.1, a53 = 0.3, b1 = 0.8. Topology 1
and topology 2 has the same control gains that are given
by k1 = 0.2 and k2 = 1.
1) The input delay and communication delay are time-
varying. We choose η(t) = 6 |sin(0.1t)| (s), and the deriva-
tive of η(t) is less than 1 that satisfies Assumption 1.
By using the LMI toolbox in Matlab for the LMI (7) in
Theorem 1, we can get τ < 0.478 s. In this section, we
assume that τ (t) = 0.3 |sin(t)| (s), and the simulation re-
sults are shown in Fig. 2. Thus, each agent converges to the
leader s state, i.e., limt→∞ xi (t) = x0 , i = 1, 2, · · · , 5.
Fig. 4 Topology 1

Fig. 5 Topology 2

1) Consider the input delays and communication de-

lays are time-varying. Choose η(t) = 6 |sin(0.02πt)| (s)
and the derivative d < 1 satisfying Assumption 1. By
using the LMI toolbox in Matlab for the LMI (11) in
Theorem 2, we get τ < 0.812 s. In this section, we as-
sume τ (t) = 0.3 |sin(0.5πt)| (s), and we obtain that each
Fig. 2 Agents with d < 1 under fixed topology agent in system (8) converges to the leader s state, i.e.,
limt→∞ xi (t) = x0 , i = 1, 2, · · · , 5 and limt→∞ vi (t) = 0,
i = 1, 2. The simulation results are shown in Fig. 6.
2) We choose d > 1 and η(t) = 6 |sin(t)| (s) that satisfies 2) Choose η(t) = 6 |sin(2t)| (s) that satisfies d > 1.
Theorem 2. From the condition (7), we can get τ < 0.417 s. From condition (11), we can get τ < 0.79 s. By choos-
Choosing τ (t) = 0.3 |sin(t)| (s), and we obtain that each ing τ (t) = 0.3 |sin(0.5πt)| (s), the closed-loop system (8)
agent converges to the leader s state, i.e., limt→∞ xi (t) = achieves an asymptotic consensus, i.e., limt→∞ xi (t) = 1,
1, i = 1, 2, · · · , 5 and limt→∞ vi (t) = 0, i = 1, 2. The limt→∞ vi (t) = 0, i = 1, 2, · · · , 5. The simulation results
simulation results are shown in Fig. 3. are shown in Fig. 7.
 P. P. Dai et al. / Consensus Problem of Heterogeneous Multi-agent Systems with · · · 345

[3] F. Xiao, L. Wang. Asynchronous consensus in continuous-

time multiagent systems with switching topology and time-
varying delays. IEEE Transactions on Automatic Control,
vol. 53, no. 8, pp. 1804–1816, 2006.
[4] F. C. Jiang, L. Wang. Finite-time information consensus
for multi-agent systems with fixed and switching topologies.
Physica D, vol. 238, no. 16, pp. 1550–1560, 2009.
[5] Y. K. Zhu, X. P. Guan, X. Y. Luo. Finit-time consensus
for multi-agent systems via nonlinear control protocols. In-
ternational Journal of Automation and Computing, vol. 10,
no. 5, pp. 455–462, 2013.
[6] P. Lin, K. Y. Qin, H. M. Zhao, M. Sun. A new approach
to average consensus problems with multiple time-delays
and jointly-connected topologies. Journal of the Franklin
Institute, vol. 349, no. 1, pp. 293–304, 2012.
[7] J. P. Hu, Y. G. Hong. Leader-following coordination of
Fig. 6 Agents with d < 1 under switching topologies multi-agent systems with coupling time delays. Physica A,
vol. 374, no. 2, pp. 853–863, 2007.
[8] J. H. Qin, W. X. Zheng, H. J. Gao. Consensus of mul-
tiple second-order vehicles with a time-varying reference
signal under directed topology. Automatica, vol. 479, no. 9,
pp. 1983–1991, 2011.
[9] J. Y. Yu, L. Wang. Group consensus in multi-agent sys-
tems with switching topologies and communication delays.
Systems & Control Letters, vol. 59, no. 6, pp. 340–348, 2010.
[10] H. Y. Yang, F. C. Wang, S. Y. Zhang. Consensus of second-
order multi-agent systems with nonsymmetric interconnec-
tion and heterogeneous delays. International Journal of Au-
tomation and Computing, vol. 8, no. 4, pp. 421–428, 2011.
[11] P. Lin, Y. M. Jia. Consensus of a class of second-
order multi-agent systems with time-delay and jointly-
connected topologies. IEEE Transactions on Automatic
Control, vol. 55, no. 3, pp. 778–784, 2010.
[12] Y. P. Gao, L. Wang. Asynchronous consensus of continuous-
Fig. 7 Agents with d > 1 under switching topologies time multi-agent systems with intermittent measurement.
International Journal of Robust and Nonlinear Control,
vol. 83, no. 3, pp. 552–562, 2010.
5 Conclusion [13] C. L. Liu, F. Liu. Dynamical consensus seeking of second-
order multi-agent systems based on delayed state compen-
In this paper, a stationary leader-following consensus sation. Systems & Control Letters, vol. 61, no. 12, pp. 1235–
protocol is adopted to solve the consensus problem for 1241, 2012.
heterogeneous multi-agent systems composed of first-order [14] C. L. Liu, F. Liu. Consensus problem of second-order multi-
and second-order agents with time-varying input delay and agent systems with time-varying communication delay and
communication delay under fixed topology and switching switching topology. Journal of Systems Engineering and
topologies. By constructing Lyapunov-Krasovskii func- Electronics, vol. 22, no. 4, pp. 672–678, 2011.
tional, the sufficient consensus conditions are provided for [15] P. Lin, Y. M. Jia. Consensus of second-order discrete-time
the heterogeneous multi-agent systems with time-varying multi-agent systems with nonuniform time-delays and dy-
namically changing topologies. Automatica, vol. 45, no. 9,
input delay and communication delay under fixed and pp. 2154–2158, 2006.
switching topologies respectively. Furthermore, consensus
conditions are obtained for unknown or different values of [16] S. H. Lin, H. B. Du, X. Z. Lin. Finite-time consensus al-
gorithm for multi-agent systems with double-integrator dy-
the derivative of communication delay. Although the con- namics. Automatica, vol. 47, no. 8, pp. 1706–1712, 2011.
sensus conditions are all formulated in the form of LMIs, the
[17] Y. P. Gao, J. W. Ma, M. Zuo, T. Q. Jiang, J. P. Du. Consen-
main results provide consensus criteria to obtain the time
sus of discrete-time second-order agents with time-varying
delay upper-bounds by using the LMI tool box in Matlab. topology and time-varying delays. Journal of the Franklin
Institute, vol. 349, no. 8, pp. 2598–2608, 2012.
References [18] W. G. Zhang, J. Z. Liu, D. L. Zeng, Y. Hu. Consensus
of second-order multi-agent systems with nonuniform time
[1] R. Olfati-Saber, R. M. Murray. Consensus problems in net- delays. Chinese Physics B, vol. 22, no. 5, Article 050511,
works of agents with switching topology and time-delays. 2013.
IEEE Transactions on Automatic Control, vol. 49, no. 9,
pp. 1520–1533, 2004. [19] M. J. Park, O. M. Kwon, J. H. Park, S. M. Lee, E. J.
Cha. Leader-following consensus criteria for multi-agent
[2] H. B. Du, S. H. Li, S. H. Ding. Bounded consensus algo- systems with time-varying delays and switching intercon-
rithms for muliti-agent systems in directed networks. Asian nection topologie. Chinese Phys B, vol. 21, no. 11, Article
Journal of Control, vol. 15, no. 1, pp. 282–291, 2013. 110508, 2012.
346 International Journal of Automation and Computing 11(3), June 2014

[20] T. Yang, Y. H. Jin, W. Wang, Y. J. Shi. Consensus of high- Ping-Ping Dai received the B. Sc. de-
order continuous-time multi-agent systems with time-delays gree in automation from Jiangnan Univer-
and switching topologies. Chinese Physics B, vol. 32, no. 1, sity, China in 2011. Since September 2011,
Article 020511, 2011. she has been studying in Jiangnan Univer-
sity for the M. Sc. degree.
[21] Z. Q. Wu, Y. Wang. Dynamic consensus of high-order multi- Her research interest includes the coor-
agent systems and its application in the motion control of dination control of multi-agent systems.
multiple mobile robots. International Journal of Automa- E-mail: [email protected]
tion and Computing, vol. 9, no. 1, pp. 54–62, 2012.
[22] Y. Zheng, Y. Zhu, L. Wang. Consensus of heterogeneous
multi-agent systems. IET Control Theory and Applications,
vol. 15, no. 16, pp. 1881–1888, 2011. Cheng-Lin Liu received the B. Sc. de-
gree in electrical engineering and automa-
[23] Y. S. Zheng, L. Wang. Finite-time consensus of hetero- tion from Nanjing University of Science and
geneous multi-agent systems with and without velocity Technology, China in 2003. He received his
measurements. Systems & Control Letters, vol. 61, no. 8, Ph. D. degree in control theory and control
pp. 871–878, 2012. engineering at Southeast University, China
in 2008. He is currently a faculty member
[24] C. L. Liu, F. Liu. Stationary consensus of heterogeneous
at Institute of Automation, Jiangnan Uni-
multi-agent systems with bounded communication delays.
versity, China.
Automatica, vol. 47, no. 9, pp. 2130–2133, 2011.
His research interests include Internet
[25] X. X. Yin, D. Yue, S. L. Hu. Distributed event-triggered congestion control and coordination control of multi-agent sys-
control of discrete-time heterogeneous multi-agent systems. tems.
Journal of the Franklin Institute, vol. 350, no. 3, pp. 651– E-mail: [email protected] (Corresponding author)
669, 2013.
[26] Y. P. Tian, Y. Zhang. High-order consensus of heteroge- Fei Liu received the Ph. D. degree in
neous multi-agent systems with unknown communication control science and control engineering
delays. Automatica, vol. 48, no. 6, pp. 1205–1212, 2012. from Zhejiang University, China. Now, he
is a professor of the Institute of Automa-
[27] U. Munz, A. Papachristodoulou, F. Allgower. Delay tion, Jiangnan University, China.
robustness in consensus problem. Automatica, vol. 46, His research interests include advanced
no. 8, pp. 1252–1265, 2010. control theory and applications, batch pro-
cess control engineering, statistical moni-
toring and diagnosis in industrial process,
[28] K. Q. Gu. An integral inequality in the stability problem of and intelligent technique with emphasis on
time-delay systems. In Proceedings of the 39th IEEE Con-
fuzzy and neural systems.
ference on Decision and Control, IEEE, Sydney, Australia, E-mail: fl[email protected]
pp. 2805–2810, 2000.