LI et al.: SECOND-ORDER GLOBAL CONSENSUS IN MULTIAGENT NETWORKS 567
all agents are randomly interconnected pairwise according to
the edge probability matrix using the weighted, directional
position, and velocity information based on the dynamical
topology of the directed random switching network G(t).
Therefore, a more general version of multiagent networks
of linearly coupled second-order dynamical systems can be
formulated as follows:
⎧
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎩
˙x
i
(t)=v
i
(t)
˙v
i
(t)= f (x
i
(t), v
i
(t), t) + α
N
j=1
W
ij
(t)B(x
j
(t)−x
i
(t))
+β
N
j=1
W
ij
(t)B(v
j
(t) − v
i
(t)), i =1, 2, ...,N
(1)
where x
i
(t) = (x
1
i
(t), ...,x
n
i
(t))
T
∈ R
n
and v
i
(t) =
(v
1
i
(t), ...,v
n
i
(t))
T
are the position and velocity variable
vectors of agent i, respectively. f (x
i
(t), v
i
(t), t) =
( f
1
(x
i
(t), v
i
(t), t), ..., f
n
(x
i
(t), v
i
(t), t)
)
T
: R
n
× R
n
×
R
+
→ R
n
represents a continuous but not necessarily dif-
ferentiable vector-valued function, which models the inher-
ent nonlinear dynamics of the uncoupled agent i. α>
0andβ>0 represent the position and velocity cou-
pling strengths between any two agents in G(t). B ∈
R
n×n
denotes the inner coupling configuration between the
agents.
The notations used throughout this paper are quite standard.
Let R
+
be the set of positive real numbers. The symmetric
part of a matrix C ∈ R
m×m
is indicted with sym(C) =
1/2(C + C
T
). N refers to the set of all nonnegative integers.
Letting x ∈ R
+
, indicate x the largest nonnegative integer
which is smaller than x.
Definition 1: The second-order globally nonlinear con-
sensus in the multiagent dynamical network (1) with
random switching directed topology is considered to be
achieved if, for any initial conditions x
i
(t
0
), v
i
(t
0
) ∈
R
N
, lim
t→∞
x
i
(t) − x
j
(t)=0 and lim
t→∞
v
i
(t)−
v
j
(t)=0, i = j, ∀i, j = 1, 2, ...,N.
In what follows, for i = 2, 3, ...,N,letX
i1
(t) = x
i
(t) −
x
1
(t) and V
i1
(t) = v
i
(t) − v
1
(t) be the position and
velocity differences between agent i and agent 1, respectively,
in the multiagent dynamical network G(t). For convenience,
we define
¯
X(t) = (X
T
21
(t),...,X
T
N1
(t))
T
,
¯
V (t) = (V
T
21
(t),...,
V
T
N1
(t))
T
and F(
¯
X(t),
¯
V (t), t) = ( f
T
(x
2
(t), v
2
(t), t) −
f
T
(x
1
(t), v
1
(t), t),..., f
T
(x
N
(t), v
N
(t), t) − f
T
(x
1
(t),
v
1
(t), t))
T
, the following error dynamical system in compact
vector form can be derived from (1)
˙
¯
X(t)
˙
¯
V (t)
=
O
(N−1)n
I
(N−1)n
αS
W
(t) ⊗ B β S
W
(t) ⊗ B
¯
X(t)
¯
V (t)
+
0
(N−1)n
F(
¯
X(t),
¯
V (t), t)
(2)
where (3), as shown at the bottom of the page, holds.
Correspondingly, for a multiagent network associated with
the time-averaged directed topology corresponding to the
Laplacian matrix
¯
L, we can obtain the following error
dynamical system:
˙
¯
X
E
(t)
˙
¯
V
E
(t)
=
O
(N−1)n
I
(N−1)n
α
¯
S
W
⊗ B β
¯
S
W
⊗ B
¯
X
E
(t)
¯
V
E
(t)
+
0
(N−1)n
F(
¯
X
E
(t),
¯
V
E
(t), t)
(4)
where
¯
S
W
=
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎣
−
¯
W
12
−
j=2
¯
W
2 j
¯
W
23
−
¯
W
13
···
¯
W
2N
−
¯
W
1N
¯
W
32
−
¯
W
12
−
¯
W
13
−
j=3
¯
W
3 j
···
¯
W
3N
−
¯
W
1N
.
.
.
.
.
.
.
.
.
.
.
.
¯
W
N2
−
¯
W
12
¯
W
N3
−
¯
W
13
··· −
¯
W
1N
−
j=N
¯
W
Nj
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎦
(5)
with
¯
W = (
¯
W
ij
)
N×N
,
¯
W
ij
= W
eij
P
ij
, i, j = 1, 2, ...,N.
In addition
¯
X
E
(t) =
X
T
E21
(t), ...,X
T
EN1
(t)
T
and
¯
V
E
(t) =
V
T
E21
(t), ...,V
T
EN1
(t)
T
where
X
Ei1
(t) = x
Ei
(t) − x
E1
(t)
and
V
Ei1
(t) = v
Ei
(t) − v
E1
(t) i = 2, 3, ...,N.
F(
¯
X
E
(t),
¯
V
E
(t), t) = ( f
T
(x
E2
(t), v
E2
(t), t)
− f
T
(x
E1
(t), v
E1
(t), t),..., f
T
(x
EN
(t), v
EN
(t), t)
− f
T
(x
E1
(t), v
E1
(t), t))
T
.
In the following, we present some assumptions
and lemmas which will be used to derive our main
results.
Assumption 1 [19]: For the nonlinear function f in (1),
there exist two constant matrices M
1
= (w
ij
)
n×n
and M
2
=
S
W
(σ (t)) =
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎣
−W
12
(t) −
j=2
W
2 j
(t) W
23
(t) − W
13
(t) ··· W
2N
(t) − W
1N
(t)
W
32
(t) − W
12
(t) −W
13
(t) −
j=3
W
3 j
(t) ··· W
3N
(t) − W
1N
(t)
.
.
.
.
.
.
.
.
.
.
.
.
W
N2
(t) − W
12
(t) W
N3
(t) − W
13
(t) ··· −W
1N
(t) −
j=N
W
Nj
(t)
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎦
(3)