Machine learning based centralized dynamic clustering algorithm for energy efficient routing in vehicular ad hoc networks

Abstract

AbstractVehicular ad hoc network (VANET) is a dynamic and self‐configure wireless network that connects vehicles to provide in‐vehicle infotainment services and comfort to both driver and passengers while on the move. Due to the highly dynamic environment, patchy connectivity, roadside barriers, and a scarcity of road side units (RSUs); existing routing protocols of VANET consume high energy during route construction. As a consequence, the VANET lifespan might be shortened. However, saving energy is a critical factor where VANET is projected to play a key role in the building of sustainable green infrastructure. Clustering is an imperative process where vehicle nodes join to make stable clusters based on common mobility features. In VANET, clustering‐based routing algorithms significantly improve network effectiveness and reduce infrastructure costs. This article proposes fuzzy c‐means (FM) machine learning‐based dynamic clustering algorithm (DCA) to identify trustworthy vehicles for energy‐efficient multi‐hop routing by considering different mobility parameters such as position, direction, speed, and remaining energy of each vehicle node. This article also proposes distance and power‐based variation of FM called fuzzy c‐means using distance and residual power (FMDP) to find stable cluster heads (CHs) for performing better data dissemination to the destination. The proposed DCA outperforms the existing k‐means based clustering algorithms as well as topology‐based DSR and AODV routing protocols in terms of packet delivery ratio, application throughput, end‐to‐end delay, network energy consumption, node energy consumption, and RSU energy consumption. Additionally, clustering methods FM and FMDP of suggested DCA improve end‐to‐end delay by 20%, application throughput by 5%, and reduce network energy consumption by 68% compared with the k‐means based clustering model.