Stability Oriented Clustering in Vehicular Ad Hoc Networks Through Machine Learning

Abstract

Vehicular Ad Hoc Networks (VANETs) are vital for enabling intelligent transportation systems through real-time vehicle-to-vehicle communication. However, maintaining stable network topologies remains a major challenge due to high mobility and frequent topology changes. This study introduces a machine learning–based clustering method designed to establish robust VANET topologies. The proposed approach employs supervised learning models to dynamically identify potential cluster heads based on factors such as mobility patterns, relative velocity, node density, and connection duration. By optimizing cluster formation, the method minimizes frequent re-clustering, strengthens connectivity, and reduces communication overhead. Simulation results demonstrate that the machine learning–based clustering approach significantly improves cluster stability, packet delivery ratio, and communication reliability compared to conventional clustering techniques. Overall, this research highlights the effectiveness of intelligent, data-driven strategies in managing VANET dynamics, thereby enabling reliable vehicular communication and enhanced road safety.