UE-Based Estimation of Available Uplink Data Rates in Cellular Networks

In this paper we propose a passive cellular data rate estimation method, which works with parameters accessible from most Commercial Off-The-Shelf (COTS) modems today. We show that by adjusting the optimization objective we can tweak the estimator to achieve higher accuracy at lower data rates. This is particularly useful in cases where a high data rate cellular link is backed up by a low data rate failover link (e.g. a satellite connection). The way the problem is approached makes this solution easily implementable in a product, for example as an input to a Quality-of-Service (QoS) router. The estimation is performed using both, supervised machine learning algorithms and a simple linear regression. The latter can be considered as a lightweight implementation on systems with low processing power or tight energy restrictions.