Vehicular Trace File

Vehicle-to-Everything (V2X) potential to support Intelligent Transportation System (ITS) is challenged by its inherent high mobility, changing topology and consequently link instability. The quest to minimize the effect of changing topology has led centroid-based clustering algorithms to exploit Cluster Head (CH) longevity approaches to improve stability while compromising on throughput performance. Most K-means based schemes particularly reselect cluster seeds at every reclustering phase. These approaches come at the cost of link throughput and stability performance, respectively. While sustaining throughput performance our work seeks to improve stability performance of persistent reclustering schemes. Exploiting Simulation of Urban Mobility (SUMO) to generate realistic vehicular traffic traces, we developed a k-means based SNR clustering scheme (KmSNR) upon which we built two memory-based reclustering schemes that leverages on the knowledge of current cluster status information to estimate the position of successive centroids and consequently determine successive Cluster Heads (CHs).We then compare the performance of these two memory-based re-clustering schemes with their corresponding memoryless (discrete) SNR-based and K-means Floyd-Warshall (KmFW) schemes. We found that the memory-based schemes outperforms their corresponding memoryless reclustering schemes in terms of jitter (>5x) and stability (>1.4x) with little or no compromise on throughput performance.