ReinWiFi: A Reinforcement-Learning-Based Framework for the Application-Layer QoS Optimization of WiFi Networks

Qianren Li, Bojie Lv, Yuncong Hong, and Rui Wang*
*Corresponding author: Prof. Rui Wang (wang.r@sustech.edu.cn)

Example

This is an example of the proposed ReinWiFi framework’s performance. The left graph visualizes the transmission latency, while the right plots the change in RTT and actions. Each UE initially follows the standard EDCA, with the scheduling parameters rate and CW denoted as 0. The proposed ReinWiFi framework is activated to optimize the QoS in the vision of a significant RTT increase.

System

The proposed system consists of multiple STAs and single AP, in which both infrasture mode wifi connection and wifi P2P connections exist.

Testbed

The proposed system consists of multiple STAs and single AP, in which both infrasture mode wifi connection and wifi P2P connections exist.

Results

Average cost where the traffic follows the same as dataset collection phase.

Average cost under unseen environment named as scenario 6 to 11.