Title: A Q-learning approach for adjusting CWS and TxOP in LAA for Wi-Fi and LAA coexisting networks

Authors: Tzu-Teng Pan; I-Sung Lai; Shang-Juh Kao; Fu-Min Chang

Addresses: Department of Computer Science and Engineering, National Chung Hsing University, Taichung, Taiwan ' Department of Business Administration, Chaoyang University of Technology, Wufeng District, Taichung, Taiwan ' Department of Computer Science and Engineering, National Chung Hsing University, Taichung, Taiwan ' Department of Finance, Chaoyang University of Technology, Wufeng District, Taichung, Taiwan

Abstract: This paper proposes a new approach to adjust values of both Contention Window Size (CWS) and Transmission Opportunity (TxOP) simultaneously by Q-learning algorithm. We optimise the adjustment of two parameters combination to maximise network throughput and achieve differentiated service. To use the Q-learning algorithm to adjust both parameters dynamically, we define the agent, environment, state and action. We also develop a reward function to help the agent find better combinations. The simulation results reveal that the system throughput of the proposed approach is 12%, 13% and 7.4% better than Fair Downlink Traffic Management (FDTM), Multi-Agent Reinforcement Learning (MARL) and Maglogiannis' Q-Learning Scheme (QLS) respectively. Compared to fixed values of CWS and TxOP, the throughput of the Wi-Fi network increases by 20.7%. When the network environment changes from uniform scenarios to uneven scenarios, the adjusting time of our approach is also 97.5% and 67% less than FDTM and MARL.

Keywords: Q-learning; listen-before-talk; coexistence networks.

DOI: 10.1504/IJWMC.2023.133061

International Journal of Wireless and Mobile Computing, 2023 Vol.25 No.2, pp.147 - 159

Accepted: 19 Jul 2022
Published online: 29 Aug 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article