Partition prediction with Angular clustering algorithm (ACA) in ad hoc networks

In wireless ad hoc networks, when the mobile nodes move with diverse patterns and cause the network to separate into completely disconnected portions, network partitioning occurs. Network partitioning is a wide scale Topology change that can cause sudden and severe disruptions to ongoing network routing and upper layer applications. Its occurrence can be attributed to the aggregate group motion exhibited in the movements of the mobile nodes. By exploiting the group mobility pattern, we can predict the future network partitioning, and thus minimize the amount of disruptions. In this paper, we propose a new characterization of group mobility based on existing group mobility models, which provides parameters that are sufficient for network partition prediction. We then demonstrate how partition prediction can be made using the mobility model parameters, and illustrate the applicability of the prediction information. Furthermore, we use a simple but effective data clustering algorithm that, given the Angles of the mobile nodes in an ad hoc network, it can accurately determine the mobility groups and estimate the characteristic parameters of each group.