A Multi-Armed Bandit Model for Non-Stationary Wireless Network Selection

The amount of wireless networks and technologies such as 5G have been rapidly increasing in the last years. With this growth in number it has become more relevant to be able to select the best network with the goal of maximizing the quality perceived by the user. The Multi-Armed Bandit (MAB) model is a viable approach to describe the problem of the best wireless network selection by a multi-Radio Access Technology (multi-RAT) device. This work proposes a new model that uses real network parameter values that change over time, that is, a non-stationary scenario. While there exist multiple MAB algorithms intended to work in stationary scenarios, this work proposes a new set of algorithms intended to work in non-stationary and more realistic environments. These new algorithms are able to measure the non-stationarity of the environment and adapt accordingly. This is especially relevant when the goal is to select among different network technologies, such as 5G, LTE or WiFi.