The evolution of wireless communications is expected to rely on machine learning (ML)-based capabilities to provide proactive management of network resources and sustain quality-of-service (QoS) and user experience. New use cases in the area of vehicular communications, including so-called vehicle-to-everything (V2X) schemes, will benefit strongly from such advances.

The objective of this challenge is to leverage the Berlin V2X dataset to predict QoS. More specifically, several features from the primary and secondary cells of two commercial LTE mobile network operators in the metropolitan area of Berlin can be used alongside GPS and context information to predict the achieved throughput in the downlink.

The results of this challenge will serve as a cornerstone to deploy general solutions for wireless QoS prediction in a multi-domain setup (i.e., across operators, devices and/or link directions).

The evaluation metric for this competition is Root Mean Squared Error.

The data includes information on:

Physical layer parameters (such as signal strength and signal quality). Cellular radio resource management like cell identity, modulation and coding scheme, assigned resource blocks, and carrier aggregation (i.e., primary/secondary cell data). Wireless Quality-of-Service (QoS) in the form of downlink throughput. GPS-positioning information. Side information (traffic and weather).