Machine learning enabled Wi-Fi saturation sensing for fair coexistence in unlicensed spectrum

In the past few years, machine learning (ML) techniques have been extensively applied to provide efficient solutions to complex wireless network problems. As such, Convolutional Neural Network (CNN) and Q-learning based ML techniques are most popular to achieve harmonized coexistence of Wi-Fi with other co-located technologies such as LTE. In the existing coexistence schemes, a co-located technology selects its transmission time based on the level of Wi-Fi traffic generated in its collision domain which is determined by either sniffing the Wi-Fi packets or using a central coordinator that can communicate with the co-located networks to exchange their status and requirements through a collaboration protocol. However, such approaches for sensing traffic status increase cost, complexity, traffic overhead, and reaction time of the coexistence schemes. As a solution to this problem, this work applies a ML-based approach that is capable to determine the saturation status of a Wi-Fi network based on real-time and over-the-air collection of medium occupation statistics about the Wi-Fi frames without the need for decoding. In particular, inter-frame spacing statistics of Wi-Fi frames are used to develop a CNN model that can determine Wi-Fi network saturation. The results demonstrate that the proposed ML-based approach can accurately classify whether a Wi-Fi network is saturated or not