Cmjv01i02p0155 by Journal of Computer and Mathematical Sciences

J. Comp. & Math. Sci. Vol. 1(2), 155-162 (2010).

On Minimum Metric Dimension of Circulant Networks 1

BHARATI RAJAN, 1INDRA RAJASINGH, 1CHRIS MONICA. M* and 2PAUL MANUEL

Department of Mathematics, Loyola College, Chennai 600 034 (India) Department of Information Science, Kuwait University, Kuwait 13060 Email: chrismonicam@yahoo.com

ABSTRACT Let M = {v1 , v 2 ,..., vn } be an ordered set of vertices in a graph G. Then

(d (u, v1 ), d (u, v 2 ),..., d (u, vn )) is called the

M-coordinates of a vertex u of G. The set M is called a metric basis if the vertices of G have distinct M-coordinates. A minimum metric basis is a set M with minimum cardinality. The cardinality of a minimum metric basis of G is called minimum metric dimension and is denoted by  (G). This concept has wide applications in motion planning and in the field of robotics. In this paper we determine the minimum metric dimension of certain classes of circulant networks. We prove that for circulant graphs G(n;  {1, 2}),  (G (n;  {1, 2}) = 3, when n = 4 l, 4 l + 2, 4 l + 3, l 1 and 2 < (G (n; ±{1, 2})  4, for 4 l+ 1, l 1. We have similar results for circulant digraphs G(n;  {1, 2, 3}) and certain subclasses of circulant graphs. Key words: metric basis, metric dimension, circulant graphs, robotics, Cayley graph.

1. INTRODUCTION Major issues involved in the design of interconnection networks are quick *This research is supported by The Major Project-No.F.8-5-2004(SR) of University Grants Commission, India.

communication among vertices, high robustness, rich structure in the sense of embeddable properties, fault tolerance and VLSI. The circulant graphs are an important class of topological structures of interconnection networks. They are symmetric with simple structures and easy extendability. Circulant graphs have

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

156

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

been used for decades in the design of computer and telecommunication networks due to their optimal faulttolerance and routing capabilities4. The term circulant comes from the nature of its adjacency matrix; a matrix is circulant if all its rows are periodic rotations of the first one. Circulant matrices have also been employed for designing binary codes. Circulant graphs also constitute the basis for designing certain data alignment networks for complex memory systems18. Most of the earlier research concentrated on using the circulant graphs to build interconnection networks for distributed and parallel systems2,4. For example, undirected circulant networks arise in the context of Mesh Connected Computer suited for parallel processing of data, such as the well-known ILLIAC type computers1. Generally, the ILLIAC network with n2 processors can be represented as a circulant graph G(n2; ±{1, n}). By using circulant graph we can adapt the performance of the network to user needs. For a given number of processors and a given crossbar switch technology, we can choose the performance of the network. If subsequently, the user needs to increase this performance we can increase the degree of the circulant graph without changing the number of processors. The opposite modification is also possible, we can increase the number of processors without changing the degree of the circulant graph. This

flexibility which is not possible with other topologies, allows us to optimize the ratio price/performance according the user. The family of circulant graphs includes the complete graph and the cycle among its members. Circulant graphs are intensively researched in computer science, graph theory and discrete mathematics. 2. An Overview of the Paper Let M = {v1 , v2 ,..., v n } be an ordered set of vertices in a graph G. Then (d (u , v1 ),d (u , v2 ),..., d ( u, vn )) is called the M-coordinates of a vertex u of G. The set M is called a metric basis if the vertices of G have distinct Mcoordinates. A minimum metric basis is a set M with minimum cardinality. If M is a metric basis then it is clear that for each pair of vertices u and v of V \ M, there is a vertex w  M such that d(u, w)  d(v, w). The cardinality of a minimum metric basis of G is called minimum metric dimension and is denoted by (G); the members of a metric basis are called landmarks. A minimum metric dimension (MMD) problem is to find a minimum metric basis. This problem has application in the field of robotics. A robot is a mechanical device which is made to move in space with obstructions around. It has neither the concept of direction nor that of visibility. But it is assumed that it can sense the distances to a set

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010). of landmarks. Evidently, if the robot knows its distances to a sufficiently large set of landmarks its position in space is uniquely determined. The concept of metric basis and minimum metric basis has appeared in the literature under a different name as early as 1975. Slater in16 and later in17 had called these sets as locating sets and reference sets respectively. Slater called the cardinality of a reference set as the location number of G. He described the usefulness of these ideas when working with sonar and loran stations. Independently, Harary and Melter8 discovered these concepts as well but used the terms metric basis and metric dimension, rather than reference set and location number. Chartrand et al.6 have called a metric basis and a minimum metric basis as a resolving set and minimum resolving set. We adopt the terminology of Harary and Melter.

157

Garey and Johnson7 noted that this problem is NP-complete for general graphs by a reduction from 3-dimensional matching. Recently Manuel et al. 15 have proved that the MMD problem is NP-complete for bipartite graphs by a reduction from 3-SAT narrowing down the gap between the polynomial classes and NP-complete classes for the MMD problem. In this paper, we derive a minimum metric basis for certain classes of circulant networks. 3. Topological Properties of Circulant Networks The circulant network is a natural generalization of double loop network, which was first considered by Wong and Coppersmith18. A circulant undirected graph, denoted by G(n;± {1, 2… j}), 1<j  n/2, n  3 is defined as an undirected graph consisting of the vertex set V = {0, 1 … n – 1} and the edge set E = {(i, j): j  i  s (mod n),

If G has p vertices then it is clear that 1(G)  p – 1. Also for the complete graph Kp, the cycle Cp and the complete bipartite graph Km,n, the minimum metric dimensions are (Kp) = p – 1, (Cp) = 2 and (Km, n) = m + n – 2 8. This problem has been studied for grids11, trees, multi-dimensional grids10, Petersen graphs3, Torus Networks 13 , Benes and Butterfly networks15 and Honeycomb networks14.

s{1, 2 … j}}. The graph in Figure 1 is the circulant graph G(8; ±{1, 2}). It is also clear that G(n;±1) is an undirected cycle C n and G(n; ±{1, 2 … n/2}) is a complete graph K n . We observe e that C n = G(n, ±1) is a subgraph of G(n; ±{1, 2 … j}) for every j, 1 < j  n/2.

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

158

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010). 0

is hamiltonian, and is a Cayley graph.

In this paper, we make use of an interesting property of circulant graphs, that it is diametrically uniform12.

3 4

Figure 1. Circulant undirected graph G(8; {1,2}) A circulant digraph was proposed by Elspas and Turner 5. A circulant digraph, denoted by G(n; {1, 2 … j}), 1<j  n–1, n2 is defined as a digraph consisting of the vertex set V = {0, 1 … n – 1} and the edge set E = {(i, j):

j  i  s (mod n) , s  {1, 2 … j}}. See

For each vertex u of a graph G, the maximum distance d(u, v) to any other vertex v of G is called its eccentricity and is denoted by ecc(u). In a graph G, the maximum value of eccentricity of vertices of G is called the diameter of G and is denoted by . Let G be a graph with diameter . A vertex v of G is said to be diametrically opposite to a vertex u of G, if dG(u, v)=. A graph G is said to be a diametrically uniform graph if every vertex of G has at least one diametrically opposite vertex. The set of diametrically opposite vertices of a vertex x in G is denoted by D(x).

Figure 2. It is clear that G(n; 1) is a directed cycle C n and G(n; {1,2…n–1 }) is a

4. Minimum Metric Dimension of Circulant Networks

complete digraph K n . Theoretical properties of circulant graphs have been studied extensively and are surveyed in2. A circulant graph

Theorem 110: Let G be a graph with minimum metric dimension 2 and let {u,v}V be a metric basis in G. Then the following are true: a) There is a unique shortest path between u and v. b) The degree of each u and v is at most 3.

Figure 2. Circulant digraph G(8; {1, 2})

Since G(n;±{1, 2 … j}), 1<j  n/2 is 2j-regular, we have the following Lemma as an application of Theorem 1.

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010). Lemma 1: (G(n; ±{1, 2 … j})) > 2, 1 < j  n/2. Proposition 1: The undirected circulant graph G(n;±{1, 2}) is diametrically uniform. Moreover, for any vertex x in G(n; ±{1, 2}), the set D(x) of diametrically opposite vertices of x satisfies 1.

D( x)  1 if n = 4l + 2

D( x)  2 if n = 4l + 3

D( x)  3 if n = 4l

D( x)  4 if n = 4l + 1.

159

Fix a  V . Then the mirror through the vertex a divides the vertex set of G(n; ±{1, 2}) into two sets S1 and S2. Let S1= { a1 , a2 ,...,a t }, S2 = { b1 , b2 ,...,bt } for t= ( n  1) / 2  . Then ai and bi are images of each other with respect to the mirror through a. See Figure 3.

The proof of Proposition 1 follows by considering the breadth first search tree rooted at the vertex x of the graph. Definition 1: For any vertex x of G(n; {1, 2}), the geometric diameter of C n through x is called a mirror. For any x  V , if D(x) is odd,, then the mirror through x passes through one of the members of D(x). This vertex is denoted by x*.

(n  1) / 2  .

Proposition 2: Let t = Then

{a *, at , bt } {a , a , b , b }  D( a)   t 1 t t 1 t {a *} {a t , bt }

if if if if

n  4l n  4l  1 n  4l  2 n  4l  3

for any vertex a of G(n; ±{1, 2}).

Figure 3. G(12; {1, 2}) Lemma 2: Let G(n; ±{1, 2}) be an undirected circulant graph where n=4l, 4l+ 2 or 4l + 3, l  1. Let M = {a, a1}. Then any two vertices in S1 ( S2 ) have distinct M-coordinates. Proof: Let u, v  S1 be such that d(a, u)=d(a, v). Then u and v are consecutive vertices on C n . Let u  a i ,

v  ai 1 . Without loss of generality,, let i be odd. The shortest paths

aa1a 3 a5 ...ai and aa 2 a 4 ...a i 1 are of

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

160

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

equal length. Then the lengths of shortest paths aa 2 a 4 ...a i 1 and a1 a 2 a 4 ...ai 1 are the same and d(a1, ai) = d(a, ai) – 1 = d(a,ai+1) – 1 = d(a1, ai+1)–1  d(a1, ai+1). Thus any two vertices in S1 have distinct M-coordinates. The proof is similar for any two vertices in S2. Theorem 2: Let G(n; ±{1, 2}) be an undirected circulant graph. Then, for l  1 , 1. (G (n; ±{1, 2}) = 3, when n = 4l, 4l + 2, 4l +3. 2. 2 < (G (n; ±{1, 2})  4, when n = 4l + 1. Proof: Let n = 4l. Let a be a vertex of G(4l;±{1,2}). By the structure of G(4l; ±{1, 2}), vertices a2 i 1 , a2 i ,

b2 i 1 , b 2i , 1  i  t 2  are at distance i from a. Note that i < . The vertices at distance  are a*, at , bt where a* is a vertex on the mirror through a. Let M={a, a1}. Then by Lemma 2, any two vertices in S1(S2) have distinct M-coordinates. Now among all pairs (x, y), x  S1, y  S2 only the pairs ( a2 i , b2 i 1 ),

Figure 4. G(9; {1, 2}) and M = {a , p, q, r} Similarly, for i > 2, d( b2 i 1 , a2 )=d( b2 i 1 , b3 )+d( b3 , a2 ) = i – 2 + d( b3 , a2 ). Hence d( a2 i , a2 )  d( b2 i 1 , a2 ), for

1  i  t 2  1. By Theorem 1, (G(n;± {1, 2}) > 2. Thus M = {a, a1 , a2 } is a minimum metric basis. Proofs for G(n; ±{1,2}) when n=4l+2, 4l+3 or 4l+1 are similar to that of G(n; ±{1, 2}). It is important to note that the mirror through the vertex a does not pass through any of the diametrically opposite vertices of a when n= 4l +1 or 4l+ 3. See Figure 4 and Figure 5.

1  i  t 2   1 , where a2 (t 2 1) = a*, are equidistant from both a and b. Now, include a2 in M. It ready follows that d( a4 , a2 )d( b3 , a2 ). Now, for or i>2, d( a2 i , a2 ) = d( a2 i , a4 ) + d( a4 , a2 ) = i – 2 + d( a4 , a2 ).

Figure 5. G(11; {1, 2}) and M = {a , p, q}

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010). The following theorems can be proved proceeding along the same lines as in Theorem 2. Theorem 3: Let G(n; ±{1, 2, 3}), n  8 be a circulant undirected graph. Then 1. 2 <  (G(n, ±{1, 2, 3}))  5, when n = 6l, l  2 . 2. 2 < (G(n, ±{1, 2, 3}))  4, when n = 6l + 2, 6l+ 4, l  1 . 3. 2 < (G(n, ±{1, 2, 3}))  5, when n = 6l + 1, l  2 . 4. 2 < (G(n, ±{1, 2, 3}))  4, when n = 6l + 3, 6l + 5, l  1 .

Figure 6. G(9; {1, 2, 3}) and M = {p, q, r} A metric basis of a digraph G(V, E) is a subset M  V such that for each pair of vertices u, v V \ M, there exists a vertex w  M such that d(w, u) d(w, v). We have the following conjecture. See Figure 6. Conjecture 1: Let G (n; {1, 2… j} be a directed circulant graph. Then

(G (n; {1, 2 … j}) = j, 1  j  n  1 .

161

4. CONCLUSION We have obtained the minimum metric dimension of undirected circulant graphs G(n; ±{1, 2}), G(n; ±{1, 2, 3}). The problem for G(n; {1, 2 … j}), j > 2 is under investigation. R E FE R E N C ES 1. G.H. Barnes, R.M. Brown, M. Kato, D. J. Kuck, D. L. Slotnick, R. A. Stokes, “The ILLIAC IV computers”, IEEE Transactions on computers, 17, 746-757(1968). 2. J.C. Bermond, F. Comellas, and D.F. Hsu, “Distributed loop computer networks: A survey”, ournal of Parallel and Distributed Computing, 24, 2-10 (1995). 3. Bharati Rajan, Indra Rajasingh, J. A. Cynthia and Paul Manuel, “On Minimum Metric Dimension”, The Indonesia-Japan Conference on Combinatorial Geometry and Graph Theory, September 13-16, 2003, Bandung, Indonesia. 4. F.T. Boesch and J. Wang, “Reliable Circulant Networks with Minimum Transmission Delay”, IEEE Transactions on Circuit and Systems, 32, 1286-1291 (1985). 5. B. Elspas and J. Turner, “Graphs with circulant adjacency matrices”, Journal of Combinatorial Theory, 9, 297-307 (1970). 6. Gary Chartrand, Linda Eroh, Mark A. Johnson, Ortrud R. Oellermann, “Resolvability in graphs and the metric dimension of a graph”, Discrete Applied Mathematics,

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

162

10.

11.

12.

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010). 105, 99-113 (2000). M. R. Garey and D. S. Johnson, Computers and Intractability: A Guide to the Theory of NP Completeness, W. H. Freeman and Company (1979). F. Harary and R. A. Melter, “The metric dimension of a graph”, Ars Combinatorica, 2, 191-195 (1976). S. Khuller, E. Rivlin and A. Rosenfeld, “Graphbots: Mobility in discrete spaces”, IEEE Transactions on Systems, Man and Cybernetics, 28(1), 29-38 (1998). S. Khuller, B. Ragavachari, and A. Rosenfeld, “Landmarks in Graphs”, Discrete Applied Mathematics, 70, 217-229 (1996). R.A. Melter and I. Tomcscu, “Metric bases in digital geometry”, Computer Vision, Graphics, and Image processing, 25, 113-121 (1984). Paul Manuel, Bharati Rajan, Indra Rajasingh and Amutha Alaguvel, “Tree Spanners, Cayley Graphs and Diametrically Uniform Graphs”, LNCS, 2880, 334-345 (2003). Paul Manuel, Bharati Rajan, Indra Rajasingh, Chris Monica. M, “Landmarks in Torus Networks”, Journal

14.

15.

16. 17.

18.

19.

of Discrete Mathematical Sciences & Cryptography, 9(2), 263-271 (2006). Paul Manuel, Bharati Rajan, Indra Rajasingh, Chris Monica. M, “On Minimum Metric Dimension of Honeycomb Networks”, Journal of Discrete Algorithm, 6(1), 20-27 (2008). Paul D. Manuel, Mostafa I. AbdEl-Barr, Indra Rajasingh and Bharati Rajan, “An Efficient Representation of Benes Networks and its Applications”, Journal of Discrete Algorithm, 6(1), 11-19 (2008). P.J. Slater, “Leaves of trees”, Congr. Number., 14, 549-559 (1975). P.J. Slater, “Dominating and reference sets in a graph”, J. Math. Phys. Sci., 22, 445-455(1988). G.K. Wong and D.A. Coppersmith, “A combinatorial problem related to multimodule memory organization”, Journal of Association for Computing Machinery, 21, 392401 (1974). J. Xu, Topological Structures and Analysis of Interconnection Networks, Kluwer Academic Publishers (2001).

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)