Smart Pattern V2I Handover Based on Machine Learning Vehicle Classification

Abstract

The mmwave frequencies will be widely used in future vehicular communications. At these frequencies, the radio channel becomes much more vulnerable to slight changes in the environment like motions of the device, reflections or blockage. In high mobility vehicular communications the rapidly changing vehicle environments and the large overheads due to frequent beam training are the critical disadvantages in developing these systems at mmwave frequencies. Hence, smart beam management procedures are desired to establish and maintain the radio channels. In this thesis, we propose that using the positions and respective velocities of the vehicles in the dynamic selection of the beam pair, and then adapting to the changing environments using machine learning algorithms, can improve both network performance and communication stability in high mobility vehicular communications.