Q-learning based heterogeneous network selection algorithm

Abstract

A Q-learning based network selection algorithm is proposed which is derived from the autonomy of network selection. It is based on the heterogeneous network composed of WiMax and HSDPA. The algorithm can constantly adjust its strategy for action selection according to network load conditions, traffic attribute and mobility of terminal by the constantly trial and error learning with wireless environment. Finally a suitable network can be selected for every session. Simulation results show that system blocking probability can be lowered and spectrum utility can be increased. © Springer-Verlag Berlin Heidelberg 2012.