On-line Client Association Scheme Based on Reinforcement Learning for WLAN Networks

Emergent WLAN applications demand more and more network resources. At the same time, the increasing number of users leads to dense WLAN networks. In this context, the standard client-Access Point association has been shown to reach far from optimal performance. Thanks to the SDN paradigm, centralised programmable strategies could be applied by the WLAN controllers to choose the best Access Point for each user when many Access Points are in the vicinity of a user. In this paper, we present an on-line centralized client-Access Point association scheme based on reinforcement learning which aims at minimizing the clients dissatisfaction in terms of obtained rate. The scheme is implemented in a SDN controller which pilots the Access Points. Simulation results show the superiority of our scheme over the IEEE standard signal strength based association in terms of aggregated throughput and users satisfaction.