Performance optimization of QoS-supported dense WLANs using machine-learning-enabled enhanced distributed channel access (MEDCA) mechanism

Abstract

Quality of service (QoS) implementation in a wireless local area network (WLAN) enables the prediction of network performance and utilization of effective bandwidth for multimedia applications. In QoS-supported WLAN, enhanced distributed channel access (EDCA) adjusts back-off parameters to implement priority-based channel access at the medium access control (MAC) layer. Although conventional QoS-supported EDCA in WLANs can provide a certain degree of QoS guarantee, the performance of best effort data (low-priority) traffic is sacrificed owing to the blind use of a binary exponential back-off (BEB) mechanism for collision avoidance among WLAN stations (STAs). In EDCA, the BEB mechanism exponentially increases the contention window (CW[AC]) for any specific priority access category (AC) when collision occurs and resets it to its initial size after successful data transmission. This increase and reset of CW[AC] is performed regardless of the network density inference, i.e., a scarce WLAN does not require an unnecessary exponential increase in CW[AC]. Similarly, a dense WLAN causes more collisions if CW[AC] is reset to its initial minimum size. Machine-learning algorithms can scrutinize an STA’s experience for WLAN inference. Therefore, in this study, we propose a machine-learning-enabled EDCA (MEDCA) mechanism for QoS-supported MAC layer channel access in dense WLANs. This mechanism utilizes a Q-learning algorithm, which is one of the prevailing models of machine learning, to infer the network density and adjust its back-off CW[AC] accordingly. Simulation results show that MEDCA performs better as compared to the conventional EDCA mechanism in QoS-supported dense WLANs.