TASC: Topology Adaptive Spatial Clustering for Sensor Networks Reino Virrankoski, Dimitrios Lymberopoulos and Andreas Savvides Embedded Networks and Application Lab Electrical Engineering Department Yale University, New Haven Infocom 2005
Ju-Mei Li Outline Introduction TASC Distributed Leader Election Discovering Local Network Structure Weight computation Grouping Similar Densities Density reachability Evaluation Conclusion
Ju-Mei Li Introduction A good topology of large-scale sensor networks should help Sensor nodes coordination Network management Data aggregation and compression Goal Through the development of weights and dynamic density reachablility Topology Adaptive Spatial Clustering Scheme (TASC)
Ju-Mei Li TASC: Distributed Leader Election Input information 2-hop neighborhood Inter-node distance measurements Min. cluster size MinPoints Each node uses input information to compute Weight Number of density reachable node Midmost position on each shortest path, biggest weight
Ju-Mei Li TASC: Distributed Leader Election f g b a c e h d i k j BroadcastToNeighborhood(weight) Select the heaviest density reachable node as nominee BroadcastToNeighborhood(nominee) Select the heaviest density reachable node as nominee BroadcastToNeighborhood(nominee) Density reachable nodes of node i = 4 Density reachable nodes of node j = 7 Density reachable nodes of node k = 3 Select the closest nominee as leader BroadcastToNeighborhood(leaderID, nodeID) Select the closest nominee as leader BroadcastToNeighborhood(leaderID, nodeID)
Ju-Mei Li TASC: Distributed Leader Election f g b a c e h d i k j If this node is leader until election timeout; BroadcastToNeighborhood(clustermenbers) If this node is leader until election timeout; BroadcastToNeighborhood(clustermenbers) If clustersize is received If clustersize < min. cluster size = 4 select the closest neighbor for which clustersize ≥ min. cluster size = 4 and joints its cluster BroadcastToNeighborhood(leaderID, clustersize) If clustersize is received If clustersize < min. cluster size = 4 select the closest neighbor for which clustersize ≥ min. cluster size = 4 and joints its cluster BroadcastToNeighborhood(leaderID, clustersize)
Ju-Mei Li TASC: Weight computation ABCDE A-B A-B-C A-B-C-D A-B-C-D-E B-C B-C-D B-C-D-E C-D C-D-E D-E
Ju-Mei Li TASC: Weight computation Including distance in Weight Computation If node k is found on path from node i to node j in between node a and node b Then the weight increment of node k is given A B C DEG F H
Ju-Mei Li TASC: Density reachability i Sensing range <= transmission range If MinPoints = m = 3 riri Could be large, equal, or small than sensing range
Ju-Mei Li TASC: Density reachability i a b c jk d e Density reachable nodes of node i : node j, node k, node a, node b, and node c
Ju-Mei Li TASC: Density reachability i k j i k j
Ju-Mei Li TASC: Distributed Leader Election
Ju-Mei Li Evaluation PARSEC 100 random scenarios 100 nodes are deployed on 1000*1000 Measurement range 200, 250, 300, 350, 400 Minimum cluster size: 4 Shortest path is done on each node Floyd-Warshall algorithm
Ju-Mei Li Evaluation
Ju-Mei Li Evaluation Measurement range: (a)200, (b)300 (a) (b)
Ju-Mei Li Evaluation Measurement range: (a)200, (b)300, (c)400
Ju-Mei Li Evaluation MinPoints = 2 MinPoints = 4 MinPoints = 6
Ju-Mei Li Conclusion This paper proposed a TASC algorithm Which uses Weight Number of density reachable node To decompose large network into smaller locally clusters
Thank You!!
Ju-Mei Li TASC: Density reachability i k j i j k