A Two-level Cooperative Clustering Scheme for Vehicular Communications

A two-level cooperative network scheme for vehicular communications is proposed in this paper. The network architecture consists on a set of deployed RoadSide Units (RSU) based on the expected density of nodes in the network whereas these nodes are organized as a vehicular ad-hoc network. The position of the RSUs is obtained using the k-means algorithm along with the gap statistic to obtain the optimal number of base stations. The nodes are clustered using spectral clustering based on the geographical position and dynamics of each node, subject to their predictable and highly correlated behavior with the environment. The head-cluster is chosen using concepts of coalitional games in order to extend the stability of the cluster. Additionally, using the beacons sent by the head-cluster to the RSU, a prediction in the dynamic behavior of the clustered nodes is achieved. The RSUs are interconnected using LTE links to provide a cooperative scheme, granting an optimal selection of the head-cluster, and prolonging its lifetime. Finally, the proposed two-level network scheme along with the clustering prediction method are analyzed and compared with the commonly used clustering techniques in a real scenario. The simulation results show the positive impact of the cooperative scheme developed predicting the movement of the clusters.