Enhanced urban clustering in VANETs using online machine learning

Clustering in Vehicular Ad-Hoc Networks (VANETs) is essential to mitigate different challenges and meet the required quality of communications. However, most of the available clustering protocols were designed for highways, and thus become unstable in realistic urban environments with many intersections. In this paper, a Clustering Adaptation Near Intersection (CANI) approach is proposed to ensure clustering stability at intersections. This approach exploits Online Sequential Extreme Learning Machine (OS-ELM) to predict the behavior of the vehicles near an intersection and adapt the clusters accordingly. The main advantage of the developed OS-ELM prediction model is its ability to continuously learn and update in real time. After being validated, the proposed adaptation approach is included in a highway clustering scheme. The resultant clustering protocol is compared to other schemes in a realistic urban environment, and shows significant stability and efficiency performance improvement.