A Clustering Method Based on K-Means and Motion Similarity in Vehicle Ad Hoc Network

Vehicle ad hoc network (VANET) have high-speed mobility, reliability and scalability, which can reduce and avoid traffic accidents according to the obtained vehicle and road information. A VANET clustering algorithm (KMSC) combining K-Means and motion similarity is proposed to solve the problem of unstable and even interrupted data transmission between VANETs due to high-speed vehicle movement and variable topology. Using elbow method, the clustering number of K-Means algorithm is no longer randomly selected, which makes the cluster convergence better. The vehicles are then assigned to the cluster based on their kinematic similarity as a measure instead of the distance parameter in the K-Means algorithm. Finally, cluster heads are selected by weight function according to kinematic similarity, node degree and average link duration ratio. The results of joint simulation using SUMO and NS3 platforms show that the KMSC algorithm has smaller end-to-end delay and larger vehicle communication throughput, which ensures the stability of inter-vehicle communication under high mobility.