Reinforcement Learning-based Contention Window Adjustment for Wireless Body Area Networks

This paper presents a reinforcement learning-based contention window adjustment for wireless body area networks (WBANs), abbreviated RL-CWA. The RL-CWA uses Q-learning to adaptively adjust the contention window according to the acknowledgment reception, thereby reducing the number of collisions. In addition, RL-CWA maintains multiple Q-tables to take into account minimum and maximum contention windows that vary depending on the user priority (UP) of the traffic. To verify the superiority of RL-CWA, an experimental simulation was conducted. The results showed that the RL-CWA obtains better performance than the existing IEEE 802.15.6 MAC.