You are on page 1of 3

Homework 1

ECE 6563 Networked Control Systems

Due: Sept.6, 2018 (Sept.13 for Q/Q3 sections)

Q. 1
If `ij is the shortest path distance (number of edges one needs to follow) between vertices vi and vj , the
diameter of a graph is defined as
diam(G) = max `ij .
vi ,vj ∈V

Similarly, if we let `?i (known as the eccentricity of vertex vi ) be the longest distance to any vertex from
vertex i, i.e.,
`?i = max `ij ,
vj ∈V

then the radius of a graph is defined as

radius(G) = min `?i .


vi ∈V

Find the radius and diameter for the four graphs LN , CN , SN , KN .

Q.2
One example of a proximity graph is the visibility graph in which two nodes are adjacent if and only if
they are visible from each other. What this means is that there are no obstacles along the line between
the two nodes.

a
Given the environment below, where the solid areas are obstacles, draw the corresponding visibility graph
and write the adjacency matrix.

b
In general, if we run the consensus equation over the positions of agents, whose inter-agent interactions
are given by a visibility graph, would the graph always stay connected if it started connected? Prove
that this is indeed the case (at least give good reasons for this), or provide a counter-example, i.e., an
example in which the visibility graph would become disconnected.

1
c
Consider now the two environments depicted below, representing agents with heterogeneous and lim-
ited sensing capabilities. In figure, dashed-line areas correspond to the portion of the domain agents can
see (note that areas and corresponding agent are denoted by the same color). Similarly to Q.2.a., draw
the corresponding visibility graph and write the adjacency matrix.

Q. 3
Let G = (V, E) be an undirected graph with adjacency matrix A. The length of any generic path in G
is equal to the number of edges composing the path. Therefore, a path of length 2 between vertexes vi
and vj can be generally represented as:
vi → vk → vj
where vk is neighbor to both vi and vj .

a
Show that the total number of paths of length 2 from i to j, via any other vertex in G, is:
(2)
Nij = [A2 ]ij

where the notation [M ]ij denote the ij th entry of matrix M .

b
A special case of all the possible paths that start from node i, are loops ending exactly on node i itself.
We call these paths cycles, and we denote the number of cycles of length r in the graph as L(r) . Following
the results from Q.3.a, show that:
L(r) = Trace(Ar )
[Note: cycles consisting of the same vertices in the same order but with different starting points should be
counted separately. Similarly, loops composed by the same number of vertices but traversed in opposite
directions should be counted separately as well.]

c
Remembering that the adjacency matrix is symmetric (i.e. A has n real eigenvalues), shows that:
X
L(r) = λri
i

where λi is the ith eigenvalue of the adjacency matrix A.

2
Q. 4
Assume a networked system composed by N nodes. Recall that if xi is scalar, with its derivative given
by the consensus equation X
ẋi = (xj − xi ), i = 1, . . . , N,
j∈Ni

this can be written as


ẋ = −Lx,
where L is the Laplacian of the network graph (assume undirected) and x = (x1 , . . . , xN )T .
If instead we have that:
ẋ = −L2 x,

a
Does this choice of dynamics solve the consensus problem? If so, why? Also, if so, what is the rate of
convergence?

b
What are the corresponding node-level dynamics, i.e., find

ẋi =???.

c
In light of Q.3.a, can you give a graph theoretic interpretation to your answer to Q. 4.b?

Q. 5
An undirected graph G is connected if and only if its Laplacian’s second smallest eigenvalue λ2 is non-
zero. Show that the number of connected components (i.e., connected subgraphs that are disconnected
from each other) is equal to the number of zero eigenvalues of the Laplacian.

You might also like