A Novel Online Machine Learning Based RSU Prediction Scheme for Intelligent Vehicular Networks

Abstract

Wireless networks development to support the highly dynamic vehicular environment pose several significant challenges for vehicular network services and applications, in efforts to guarantee seamless communication. Intelligent Vehicular Networks goal is to provide high-quality services that can learn and forecast clients' needs and intentions. Machine Learning (ML) is one type of artificial intelligence that can be effective in utilizing the vehicular network's data to predict users movements and allocate resources ahead of time. In this paper, we propose a novel online ML-based Roadside Unit (RSU) prediction scheme for mobility management in Vehicular Networks, to provide seamless mobile connectivity to vehicles and enhance the performance of the prediction model. An Online Probabilistic Neural Network (O-PNN) prediction model is designed and adjusted for VANETs mobile IP protocol. Extensive simulation experiments were performed on the Network Simulator NS-2, and the performance of the prediction model is studied with different traffic and mobility scenarios. Our results showed a high accuracy rate in comparison to several other machine learning models.