영문내용
(English Abstract) |
The significant phase of 802.11ax improvement is to enhance the spatial and resource sharing capabilities in the dense network environment. Spatial Reuse (SR) is one of the spatial mechanisms in 802.11ax that has a function to maximize parallel transmission in WLAN. This mechanism is important because the number of Wi-Fi station being used continues to increase, but space between them is decreasing. One of the SR mechanisms is Overlapping Basic Service Set/ Packet Detect (OBSS/PD) threshold. The function of OBSS/PD threshold is to determine its sensitivity threshold. However, OBSS/PD threshold has a drawback. The drawbacks are if the threshold is too low, it might decrease simultaneous transmission. Furthermore, if it is too high, they have the potential to have a hidden node problem. Currently, the OBSS/PD threshold is a static number that requires administrator to configure it manually. Thus, our contribution is to create a dynamic design to choose the appropriate OBSS/PD threshold. We use one of Reinforcement Learning method called Multi-Armed Bandits (MAB) that can learn from the environment and evaluate the reward. The results of dynamic OBSS/PD using MAB especially the Upper Confidence Bound (UCB) algorithm achieve good throughput and fairness rather than using a heuristic algorithm. |
