Long-term application-level wireless link quality prediction

The knowledge of a future throughput value for a user equipment (UE) in Long Term Evolution (LTE) or any other transmission technology is very valuable. It can be used in rate adaptation algorithms so that radio channel congestions may be mitigated thus allowing for better quality of experience of the wireless user. Such control usually would happen at the application layer so that the control loops at different layers may work together and thus create a stable operating point. However, the metrics that are learned and predicted are available at the core of the radio link. In this paper we identify a suitable application-level link quality metric (which we call bits_U/prb) for prediction, and analyze the performance of the predictions at differing Rayleigh velocities. We find that the prediction of the application-level link quality can be 90 to 97 % accurate.