Effective Prediction of V2I Link Lifetime and Vehicle's Next Cell for Software Defined Vehicular Networks: A Machine Learning Approach

Abstract

Predicting, in the one hand, the time duration that a vehicle remains associated to a cell i.e. Network Attachment Point (NAP) and, on the other hand, the next cell can help anticipating network control decisions to provide services with stringent requirements despite vehicle mobility. In this paper, we propose a machine learning based approach for Software Defined Vehicular Networks that allows a cell to estimate the attachment duration of each newly associated vehicle at the association request time, as well as, a prediction of the upcoming cell, performed at the SDN controller that controls the cells. Our proposed models have been evaluated on a large dataset, which we have generated based on a real mobility trace from the city of Luxembourg, and the evaluation shows promising results in terms of prediction accuracy.