NECESSARY AND SUFFICIENT CONDITIONS FOR SOLVING CONSENSUS PROBLEMS 1709
Lemma 2 (Olfati-Saber and Murray [4])
Given a digraph, it is balanced if and only if w
l
=1 is
the left eigenvector of the Laplacian matrix associated
with the zero eigenvalue, i.e. 1
T
L =0.
For an undirected graph, its Laplacian matrix L
is positive semi-definite. For a connected undirected
graph, there is only one zero eigenvalue of L, all the
other ones are positive and real.
Given a graph G, denote
+
(L) as the set of nonzero
eigenvalues of the Laplacian matrix L of G.
2.2. Consensus problems of second-order multi-agent
system
We consider a group of n identical agents indexed by
1,...,n. All those agents share a common state space R.
A continuous-time model of the n agents with second-
order dynamics is described as follows:
˙x
i
= v
i
,
˙v
i
= u
i
, i =1,...,n
(1)
where the state x
i
∈R is the position of agent i , v
i
∈R
its velocity and u
i
∈R its control input.
The control objective is to regulate each agent’s posi-
tion and velocity such that all the agents asymptoti-
cally travel with zero relative positions and common
velocities.
More concretely, let :R
n
→R be a function of state
x(t) or velocity v(t), where x(t) =[x
1
(t), x
2
(t),...,
x
n
(t)]
T
and v(t ) =[v
1
(t),v
2
(t),...,v
n
(t)]
T
with x(0)
and v(0) denoting the initial state and the initial velocity
of the system. The -consensus problem is a distributed
way to calculate (x(0)) and (v(0)) by applying inputs
u
i
that only depend on the states and velocities of itself
and its neighbors. We say a feedback
u
i
(t) =k
i
1
(x
j
1
(t), x
j
2
(t),...,x
j
L
i
(t))
+k
i
2
(v
j
1
(t),v
j
2
(t),...,v
j
L
i
(t)) (2)
is a control protocol with topology G if the cluster
{j
1
,..., j
L
i
}={i}∪N
i
, i =1,...,n.
We say protocol (2) globally asymptotically solves
the -consensus problem if and only if x
i
(t)→(x(0))
+t(v(0)) and v
i
(t)→(v(0)),ast→∞, i =1,...,n.
A special case with (x(0))=
1
n
n
i=1
x
i
(0) and (v (0))=
1
n
n
i=1
v
i
(0) is called average-consensus problem.
3. SAMPLED-DATA CONTROL PROTOCOL AND
INDUCED NETWORK DYNAMICS
In [14], the following continuous-time linear consensus
protocol for system (1) was introduced:
u
i
(t) =
j∈N
i
a
ij
(x
j
(t)−x
i
(t))
+
j∈N
i
a
ij
(v
j
(t)−v
i
(t)), i =1,...,n. (3)
It is a distributed protocol that each agent only needs
information from its local neighbors.
In this paper, we assume that each agent can only
receive the information of positions and velocities from
its neighbors at sampling times. A sampled-data control
protocol is induced from (3) by using periodic sampling
technology and zero-order hold circuit. Let h>0bethe
sampling period, the obtained protocol is given as:
u
i
(t) =
j∈N
i
a
ij
(x
j
(kh)−x
i
(kh))
+
j∈N
i
a
ij
(v
j
(kh)−v
i
(kh)),
if t ∈[kh, kh+h), k =0, 1, 2,...;
i =1,...,n. (4)
In this case, discretizing Equation (1) with sampling
period h, we obtain:
x
i
(kh+h) = x
i
(kh)+hv
i
(kh)+
1
2
h
2
u
i
(kh),
v
i
(kh+h) = v
i
(kh)+hu
i
(kh),
k = 0, 1, 2,...; i =1,...,n.
(5)
where x
i
(kh), v
i
(kh) represent the position and velocity
of the i th agent at the kth sampling time.
Denote (kh) =[
T
1
(kh),
T
2
(kh),...,
T
n
(kh)]
T
, where
i
(kh) =[x
i
(kh), v
i
(kh)]
T
. Then the network dynamics
is summarized as follows:
(kh+h) =(kh), k =0, 1, 2,.... (6)
where =I
n
⊗A+L ⊗(BK) and
A =
1 h
01
, B =
1
2
h
2
h
, K =[−1 −]
Copyright q 2009 John Wiley & Sons, Ltd. Int. J. Robust Nonlinear Control 2010; 20:1706–1722
DOI: 10.1002/rnc