13.07.2015 Views

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

242 G. Cal<strong>in</strong>escu and P.-J. Wan5 Algorithm MST-Augmentation for BiconnectivityIn this section, we present a simple algorithm which produces a biconnectedspann<strong>in</strong>g graph H <strong>by</strong> augment<strong>in</strong>g an MST. The algorithm first f<strong>in</strong>ds an EuclideanMST T and <strong>in</strong>itializes H to T . At any non-leaf node v of T ,alocalEuclidean MST T v over all the neighbors of v <strong>in</strong> T is constructed and addedto H. ThustheH is a union of the big MST T and many small MSTs. H is2-connected, as it follows from the follow<strong>in</strong>g argument. Only <strong>in</strong>ternal nodes ofT can be articulation po<strong>in</strong>ts; let u be such a node. Remov<strong>in</strong>g u from T creates anumber of connected components of T , each hav<strong>in</strong>g one vertex neighbor with u<strong>in</strong> T . But the neighbors of u <strong>in</strong> T rema<strong>in</strong> connected <strong>by</strong> T u , the local MST whichdoes not <strong>in</strong>clude u.We refer to this algorithm as MST-Augmentation. Besides be<strong>in</strong>g simpleand very fast (as every vertex has constant degree <strong>in</strong> T , total runn<strong>in</strong>g time isdom<strong>in</strong>ated <strong>by</strong> construct<strong>in</strong>g T and is O(n log n)), this algorithm is best suited toefficient distributed implementation. Another advantage of this algorithm is the<strong>in</strong>dependence of the path-loss exponent.To bound the approximation ratio of MST-Augmentation, we <strong>in</strong>troduce ageometric constant α def<strong>in</strong>ed below. Let o be the orig<strong>in</strong> of the Euclidean plane.A set U of at least two po<strong>in</strong>ts is called as a star-set if its Euclidean MST for{o}∪U is a star centered at o. The star is denoted <strong>by</strong> S U . Note that each starsetconta<strong>in</strong>s at least two but at most six po<strong>in</strong>ts. For any star-set U, let T U bethe m<strong>in</strong>imum spann<strong>in</strong>g tree of U. Then α is def<strong>in</strong>ed as the supreme of the ratioc (T U ) /c (S U ) over all star-sets.Lemma 2. For any κ ≥ 2, 2 κ−1 ≤ α ≤ 1.6 · 2 κ−1 . If κ =2, then α =2.Proof. The lower bound 2 κ−1 is achieved <strong>by</strong> U consist<strong>in</strong>g of two po<strong>in</strong>ts u 1 and u 2such that o is the midpo<strong>in</strong>t of the l<strong>in</strong>e segment u 1 u 2 . Next, we prove the upperbound 1.6 · 2 κ−1 . Consider any star-set U. IfU has exactly six po<strong>in</strong>ts, then thesepo<strong>in</strong>ts form a regular hexagon centered at o, and hence c (T U )= 5 6 c (S U )

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!