Abstract
In wireless ad hoc networks, network partition, a large-scale topology change, would occur unexpectedly and disrupt the on-going routing paths as well as application connections. By observing the group mobility patterns that are aggregated from the movement of mobile nodes, the prospective network partitioning can be predicted and therefore we can reduce the amount of disruptions. In this paper, we proposed a group mobility model that uses mobility parameters, such as distance, velocity and acceleration for network partition prediction. We then demonstrate how to predict network partition with these parameters, and illustrate how to apply the prediction to routing protocols. We also propose a group-id based clustering algorithm, which exactly determines the mobility groups as well as elects the clusterheads.