Reinforcement Learning-Empowered Decentralized Channel Selection for Dense Wi-Fi 7 Networks
Rui Xin Xu, Gaolei Li, Zhaohui Yang, Xiao Yang, Jianhua Li · 2023
Wi-Fi 7 (IEEE 802.11be standard) is gaining world-wide popularity due to its easy extensibility and configuration. The dynamic channel bonding (DCB) is an essential technology for Wi-Fi 7 networks to improve network throughput and optimize users' experience. However, the surging density of DCB-based wireless devices has raised up a multidimensional interference problem when scheduling limited radio resources. Furthermore, the existing interference analysis for low-frequency channels is no longer applicable to the commonly used 6GHZ channels in Wi-Fi 7, which can lead to severe co-channel interference in dense networks and significantly reduce the network's effective throughput. To address these problems, in this paper, we first conduct a novel DCB-enabled wireless channel model for dense Wi-Fi 7 networks. And then, we quantify the negative effects of neighboring access points by deriving the interference factors between different channels. Based on this, we propose to formulate the optimized task as a multi-player version of the multi-armed bandit problem, with the aim of maximizing the system's accumulative reward within a finite working time. Finally, we propose a reinforcement learning-empowered decentralized channel selection (RL-DCS) algorithm to address the above problem in the dense Wi-Fi 7 network. Compared to the traditional stochastic selection algorithm and the classical centralized selection algorithm, simulation results show that proposed methods can achieve significant superiority in convergence rate and accumulative total throughput.