Minimum Contention Window Control Using Q-learning in IEEE 802.11 DCF

In order to reduce the collision probability in IEEE 802.11 Distributed Coordination Function (DCF) protocol, if a collision occurs, then the contention window size become twice and a random number is selected for a backoff. The mechanism, however, may suffer from performance degradation depending on the communication environment, especially in dense or sparse networks. Those problems of the conventional DCF have been studied by many researchers. In this paper, we propose a Q-learning algorithm that enhances the performance of IEEE 802.11 DCF by controlling the minimum contention window size of each terminal, rather than using a fixed window for a network. Simulation results shows that our mechanism outperforms the legacy protocol in various network conditions.