A Statistical Learning Framework for QoS Prediction in V2X

Abstract

Managing QoS is one of the most critical and challenging aspects for connected and automated driving to be accepted in reality, as pre-agreed QoS Key-Performance-Indicators (KPIs) such as throughput, latency, and packet delivery ratio may not be guaranteed at all times. Enabling notifications with QoS predictions to vehicle applications presents a way to act upon potential QoS degradation. This, in turn, will make possible to improve overall system reliability while enhancing safety of connected and automated driving. To meet the V2X requirements and deliver QoS prediction with accuracy information, this paper proposes a prediction framework that combines a channel prediction model that maps contextual information into prediction of channel characteristics, with a statistical learning model that delivers QoS prediction with statistical guarantees.