ENHANCED VANET ROUTING THROUGH ADAPTIVE HYBRID POSITION-CLUSTER BASED PROTOCOL: A MACHINE LEARNING APPROACH

Abstract

In VANETs, the high mobility of nodes along with changing topology and diverse traffic conditions make it challenging to implement efficient as well as reliable routing. Most traditional routing approaches-whether they are position-based or cluster-based cannot hold their optimal performance in all scenarios. In this context, this paper presents a novel Adaptive Hybrid Position-Cluster Based Protocol using machine learning to enhance the efficiency of VANETs' routing. AHPCBP incorporates a real-time awareness of positions through dynamic clustering mechanisms and leverages machine learning to predict mobility patterns of the vehicle and to optimise the clustering process. Thus, supervised models trained from past movement patterns can be utilised to forecast future changes in topology and update its routing strategies beforehand. The protocol introduces a new clustering algorithm that dynamically adjusts cluster size and formation based on traffic density, vehicle speed, and predicted movement patterns. Extensive evaluations with realistic urban mobility traces demonstrate that AHPCBP significantly outperforms existing protocols, achieving a 27% increase in packet delivery ratio, a 35% reduction in end-to-end delay, and a 42% decrease in routing overhead. It proves to be highly effective in urban environments with varying densities and dynamic mobility patterns. Moreover, the prediction accuracy of vehicle trajectory forecasting is strong with machine learning, which is 89%, leading to stabilization in cluster creation and later maintenance. Such results show the promising integration of machine learning with hybrid position-cluster-based routing to enhance future VANET communication systems, providing adaptability and reliability in widely varied traffic scenarios.