1. An adaptive Multihop Clustering Scheme for Highly Mobile Ad Hoc Networks Tomoyuki Ohta, Shinji Inoue, and Yoshiaki Kakuda ISADS 2003

2. Outline Introduction Adaptive Multihop Clustering Scheme Simulation Results Conclusion

3. Introduction A clustering scheme is aimed at managing a number of mobile devices by utilizing hierarchical structure of the networks Cluster member Clusterhead Cluster gateway

4. Introduction (cont.) An effective hierarchical structure in ad hoc network where mobile devices may move at high mobility The role of each mobile device for the hierarchical structure is adaptive to dynamic change of the topology The overhead for management of the hierarchical structure is small

5. Introduction (cont.) This paper proposes An adaptive multihop clustering scheme for highly mobile ad hoc networks Merger and division of clusters Does not depend on mobility Satisfy the above requirements

6. Adaptive Multihop Clustering Scheme Action 1: A node exchanges the information with the neighboring nodes periodically by using hello packet A cluster member can get information which the neighboring cluster members Action 2: A cluster member periodically broadcasts information to other cluster members in the same cluster by using periodic notification packet A cluster member can share information with other cluster members in the same cluster

7. Adaptive Multihop Clustering Scheme (cont.) States NSN (Normal state Node) Cluster member CN (Control Node) Cluster head BN (Border Node) Gateway BCN (Border and Control Node) Both a cluster head and a gateway ON (Orphan Node) Node Vi does not have any cluster ID

8. Adaptive Multihop Clustering Scheme (cont.) Assignment and update of cluster ID Case 1: State(i ) = ON and when i received a hello packet from an ON node, i recognizes that all neighboring nodes have been also orphan nodes Lets Cluster(i ) = i i

9. Adaptive Multihop Clustering Scheme (cont.) Assignment and update of cluster ID (cont.) Case 2: State(i ) = ON and when i received a hello packet from a neighboring node with its cluster ID=k Cluster (i ) = k

10. Adaptive Multihop Clustering Scheme (cont.) Assignment and update of cluster ID (cont.) Case 3: When i received a hello packet from the neighboring node t and the packet indicates Cluster(t ) = m, i recognizes that all the neighboring nodes holding cluster ID k have disappeared Cluster(i ) = m

11. Adaptive Multihop Clustering Scheme (cont.) State transition in nodes This condition represents that the cluster ID is updated and the updated cluster ID equals to the node ID held by node vi This condition represents that the cluster ID is updated and the updated cluster ID does not equal to the node ID held by node vi This condition represents that a neighboring node appeared, which holds a cluster ID being different with cluster ID held by node vi This condition represents that the cluster ID’s held by all neighboring nodes became the same as the cluster ID held by node vi This condition represents that all neighboring nodes of node v disappeared

12. Adaptive Multihop Clustering Scheme (cont.) Division and Merger of Clusters In order to maintain the hierarchical structure efficiently, load for each cluster should be equivalent U denote the upper bound If Ci > U -> Division L denote the lower bound If Ci Merger

13. Adaptive Multihop Clustering Scheme (cont.) Merger C G + C C ≤ U Clusterhead g send a cluster merge request packet to Clusterhead c Clusterhead c reply packet including ’answer yes’ C G NCCCN N GGG

14. Adaptive Multihop Clustering Scheme (cont.) Division To divide cluster Ci into two neighboring clusters Ci and Cj Clusterhead i sends a cluster divide packet to cluster member j Clusterhead i and Clusterhead j send cluster ID update packets ijnm Cluster Id update packet

15. Adaptive Multihop Clustering Scheme (cont.)

16. Simulation Results 200 mobile nodes 750m × 750m Transmission range of radius 100m Each node periodically(200 ms) sends hello packets The duration of each simulation is 80 seconds

17. Simulation Results (cont.) upper bound lower bound

18. Simulation Results (cont.)

19. Conclusion If the cluster range is set appropriately, the proposed clustering scheme can constitute almost the same size clusters Regardless of whether mobility and degree of nodes change The overhead for the management of the hierarchical structure is uniformly distributed to clusters