An Intelligent Machine Learning Based Routing Scheme for VANET

Abstract

Today, Vehicular Ad-hoc Networks (VANET) have become an interesting research topic for developing Intelligent Transport Systems. In urban environments, vehicles move continuously and at different speeds, which leads to frequent changes in the network topology. The main issue faced in an urban scenario is the performance of routing protocols when delivering data from one vehicle to another. This paper introduces ECRDP, an Efficient Clustering Routing approach using a new clustering algorithm based on Density Peaks Clustering (DPC) and Particle Swarm Optimization (PSO). First, the PSO algorithm is applied to determine the cluster heads, or a new fitness function for finding the best solutions is formulated using the DPC algorithm. Next, clustering is performed based on the reliability of links parameter between vehicles. Then, a maintenance phase is proposed to update the cluster heads and redistribute the vehicles in the clusters. Finally, the effectiveness of the suggested scheme is evaluated by a simulation operated by MATLAB on a real urban scenario. The results achieved show an overall increase in stability, proven by a reduction in change rate by 74%, and an improvement in performance indicated by an increase in intra-cluster throughput by 34% and inter-cluster by 47%, as well as an overall reduction of average delay by 16%.

Keywords

• Clustering
• density peak clustering
• machine learning
• particle swarm optimization
• routing
• vehicular ad-hoc NETwork