SAMNET: Self-adaptative multi-kernel clustering algorithm for urban VANETs

Abstract

Nowadays, vehicle ad hoc networks (VANET) are becoming one of the trends that motivate many service providers in urban areas. In this work, we take into account the random and continuous evolution of traffic in the VANET environment. We adopt a system to model the mode of evolution based on commutation. The proposed system is defined as a finite collection of linear submodels. Thus, for each subsystem, it is necessary to identify the discrete state and to establish a specific sub-model to model the overall system. However, according to recent studies, the adoption of an efficient VANET clustering algorithm can promote road safety, provide a means of entertainment for passengers, and promote message routing. In this article, a clustering algorithm based on a Self-Adaptative Multi-Kernel clustering for urban VANET (SAMNET) is also provided. SAMNET is based on a set of measurement data, representing the unpredictable density of vehicle nodes, acceleration or deceleration, and the limited radio range of the communication scheme used. The proposed algorithm takes advantage of the concept of identifying these data generated by linear sub-models which communicate through an unpredictable dynamic switching. It is a self-adapting clustering algorithm that consists of modeling each sub-model based on a linear regression function. SAMNET is broken down into three stages: (i) initialization of clusters, (ii) adaptation of clusters, (iii) fusion of clusters. To assess the comparative effectiveness of SAMNET, many experiments are carried out. The results obtained show that the proposed methodology provides almost optimal results and works well with regard to the average lifetime of the clusters and the data delivery rate.