AI-driven, QoS prediction for V2X communications in beyond 5G systems

On the eve of 5G-enabled Connected and Automated Mobility, challenging Vehicle-to-Everything services have emerged towards safer and automated driving. The requirements that stem from those services pose very strict challenges to the network primarily with regard to the end-to-end delay and service reliability. At the same time, the in-network Artificial Intelligence that is emerging, reveals a plethora of novel capabilities of the network to act in a proactive manner towards satisfying the aforementioned challenging requirements. This work presents PreQoS, a computationally-efficient, predictive Quality of Service mechanism that focuses on Vehicle-to-Everything services. PreQoS is able to timely predict specific Quality of Service metrics, such as uplink and downlink data rate and end-to-end delay, in order to offer the required time window to the network to allocate more efficiently its resources. Geographical space discretization and clustering techniques are applied in advance to the prediction process for computational and communication requirements minimization. On top of that, the proactive management of those resources enables the respective Vehicle-to-Everything services and applications to perform any potential Quality of Service-related required adaptations in advance. The evaluation of the proposed mechanism based on a realistic, simulated, Connected and Automated Mobility environment proves the viability and validity of such an approach.