Download presentation
Presentation is loading. Please wait.
Published byJohn Parker Modified over 8 years ago
1
Maximal Independent Set and Connected Dominating Set Xiaofeng Gao Research Group on Mobile Computing and Wireless Networking Univ. of Texas at Dallas
2
Catalog Introduction Preliminary Voronoi Division Euler’s Formula Discussion for Holes Xiaofeng Gao, Univ. of Texas at Dallas26/6/2008
3
INTRODUCTION Section 1 Xiaofeng Gao, Univ. of Texas at Dallas36/6/2008
4
Connected Dominating Set A virtual backbone for wireless ad hoc network Garey and Johnson (1978) proved that finding a Minimum Connected Dominating Set in a general graph is NP-hard. Researchers always use approximation algorithms to select a feasible solution (as close as possible to OPT) Xiaofeng Gao, Univ. of Texas at Dallas46/6/2008
5
Communication Model Unit Disk Graph (UDG) ◦ G=(V, E) ◦ Each v in V denotes a center of a disk with radius 1 ◦ Each (u, v) in E denotes the distance between u and v is less than or equal to 1. Disk Graph (DG) ◦ G=(V, A) ◦ Each v in V denotes a center of a disk with radius r v ◦ Each (u, v) in E means v lies within u’s range. Say, dist(u,v) <= r u Xiaofeng Gao, Univ. of Texas at Dallas56/6/2008
6
Common Techniques Find a Dominating Set D ◦ Partition ◦ Final Maximal Independent Set (MIS) Connect D as a CDS ◦ Steiner Tree ◦ Spanning Tree Greedy Approach ◦ Spider (Guha & Khuller) Xiaofeng Gao, Univ. of Texas at Dallas66/6/2008
7
Approximation Ratio Given graph G mcds(G): the size of optimal CDS solution for G mis(G): the size of selected MIS set for G connect(G): the size of disks selected to connect the MIS into CDS. Xiaofeng Gao, Univ. of Texas at Dallas76/6/2008
8
Approximation Ratio (2) Formula Compare the relationships between mis(G) and mcds(G) is crucial to reduce the approximation ratio Xiaofeng Gao, Univ. of Texas at Dallas86/6/2008
9
Milestones Wan et.al. (2002) ◦ mis(G) <= 4 ∙ mcds(G) + 1 Wu et.al. (2006) ◦ mis(G) <= 3.8 ∙ mcds(G) +1.2 Funke et.al. (2006) ◦ mis(G) <= 3.748 ∙ mcds(G) + 9 ◦ mis(G) <= 3.453 ∙ mcds(G) + 8.291 Yao et.al. (2008) ◦ mis(G) <= 11/3 ∙ mcds(G) + Constant Xiaofeng Gao, Univ. of Texas at Dallas96/6/2008
10
Our Accomplishment Result ◦ mis(G) <= 3.453 ∙ mcds(G) + 4.839 ◦ mis(G) <= 3.339 ∙ mcds(G) + 4.974 if there’s no hole in the special graph ◦ mis(G) <= 3.478 ∙ mcds(G) + 4.874 if there exists holes in the induced graph Strategy ◦ Voronoi Division ◦ Euler’s Formula Xiaofeng Gao, Univ. of Texas at Dallas106/6/2008
11
PRELIMINARIES Section 2 Xiaofeng Gao, Univ. of Texas at Dallas116/6/2008
12
Induced Graph OPT: the optimal CDS in G. ◦ OPT can dominate the whole graph. ◦ For any disks u, v in OPT, dist(u, v) <= 1 S: a selected independent set ◦ For any disk d in S, there exist a disk t in OPT such that dist(d, t) <= 1 ◦ For any two disks d 1, d 2 in S, dist(d 1,d 2 )>1 Make Induced Graph G’ ◦ Change radius of disks in OPT into 1.5 ◦ Change radius of disks in V\OPT into 0.5 Xiaofeng Gao, Univ. of Texas at Dallas126/6/2008
13
Properties for Induced Graph The areas covered by OPT will contain all the other disks. For any selected independent set S, any two disks will not intersect each other. If a disk belongs to both OPT and S, we can consider it as two distinct disks, one belongs to OPT and another belongs to S. Xiaofeng Gao, Univ. of Texas at Dallas136/6/2008
14
Example – Original Graph Xiaofeng Gao, Univ. of Texas at Dallas146/6/2008
15
Example – Induced Graph Xiaofeng Gao, Univ. of Texas at Dallas156/6/2008
16
Rough Bound Xiaofeng Gao, Univ. of Texas at Dallas166/6/2008
17
Calculation Xiaofeng Gao, Univ. of Texas at Dallas176/6/2008
18
VORONOI DIVISION Section 3 Xiaofeng Gao, Univ. of Texas at Dallas186/6/2008
19
Definition Let S be a set of n nodes in Eucliean space. For each node pi in S, the Voronoi Cell V(p i ) of p i is the set of points that are closer to p i than to any other nodes of S. Xiaofeng Gao, Univ. of Texas at Dallas196/6/2008
20
Example |OPT| = 2; |MIS| = 7 Xiaofeng Gao, Univ. of Texas at Dallas206/6/2008
21
Analysis Each disk in MIS will occupy a Voronoi Cell with more area than a circle with radius 0.5 The minimum Voronoi Cell inside the graph should be a heptagon. Xiaofeng Gao, Univ. of Texas at Dallas216/6/2008
22
Minimum Area for Triangle Non-Boundary ◦ area(P 3 )=1.299 Boundary ◦ area(E 3 )=1.178 Xiaofeng Gao, Univ. of Texas at Dallas226/6/2008
23
Minimum Area for Quadrangle Xiaofeng Gao, Univ. of Texas at Dallas23 area(P 4 ) = 1 area(A 4 ) > 1.136 area(E 4 ) = 0.972 6/6/2008
24
Minimum Area for Pentagon Xiaofeng Gao, Univ. of Texas at Dallas24 area(P 5 ) = 0.908 area(A 5 ) > 0.950 area(P 5 ) = 0.908 6/6/2008
25
Minimum Area for Hexagon Xiaofeng Gao, Univ. of Texas at Dallas25 area(P 6 ) = 0.866area(E 6 ) = 0.855 area(A 6 ) = 0.886 6/6/2008
26
Heptagon and Others area(P k )> area(P 6 ) (k>=7) Xiaofeng Gao, Univ. of Texas at Dallas266/6/2008
27
Minimum Area Area(A 7 ) = 0.853 Xiaofeng Gao, Univ. of Texas at Dallas276/6/2008
28
Modified Upper Bound Xiaofeng Gao, Univ. of Texas at Dallas286/6/2008
29
EULER’S FORMULA Section 4 Xiaofeng Gao, Univ. of Texas at Dallas296/6/2008
30
Symbols s i : the minimum area of the non-boundary i-polygon cell; s i ': the minimum area of the boundary i- polygon cell. We have ◦ s 3 >= s 4 >= s 5 >= s 6 <= s 7 <= s 8 <= … ◦ s 3 ' >= s 4 ' >= s 5 ' >= s 6 ' >= s 7 ' <= s 8 ' <=... Set s i = s 6 (i >=7) and s i ' = s 7 ' (i >= 8) Xiaofeng Gao, Univ. of Texas at Dallas306/6/2008
31
3-Regularization Make the degree of every vertex in Voronoi Division be exactly 3. Xiaofeng Gao, Univ. of Texas at Dallas316/6/2008
32
Euler’s Formula : the outer boundary of MCDS f i : no. of non-boundary Voronoi Cells with i edges; f i ': no. of boundary VC with i edges. m: no. of edges n: no. of vertices By Euler’s Formula: Xiaofeng Gao, Univ. of Texas at Dallas326/6/2008
33
Equations G' is a cubic graph ◦ 2m = 3n ◦ (1) be no. of edges in outer face, every edge is in two faces exactly, (2) For boundary cell, it must have one edge belonging to the outer face (3) 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas33
34
Calculation Combining (1), (2), and (3), we have Together with Euler’s Formula, remove n, we have 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas34
35
Calculation – cont. Since all Voronoi cells are contained in the area constructed by MCDS, Combining two formula together, 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas35
36
Calculation – cont. Since we have Finally, we get that 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas36
37
DISCUSSION WITH HOLES Section 5 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas37
38
Modification k: no. of holes in G‘ : no. of edges in all holes. Then we have 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas38
39
Modification – cont. 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas39 We have Since
40
Guess 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas40
41
Example 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas41
42
Reference Xianyue Li, Xiaofeng Gao, Weili Wu, A Better Theoretical Bound to Approximate Connected Dominating Set in Unit Disk Graph, Submitted to The 3rd International Conf erence on Wireless Algorithms, Systems and Applications (W ASA 2008), Oct. 26-28, 2008. S.Funke, A.Kesselman and U.Meyer, A Simple Improved Distributed Algorithm for Minimum CDS in Unit Disk Graphs, ACM Transactions on Sensor Networks, 2(3), 444- 453, (2006). 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas42
43
Questions? 6/6/2008Xiaofeng Gao, Univ. of Texas at Dallas43
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.