Downlink SNR Estimation of Wi-Fi Clients using Machine Learning

Wi-Fi, based on the IEE802.11 standard, is the most popular commercial wireless technology today. IEEE 802.11 specification includes mechanisms for channel feedback from clients, but these are rarely implemented due to implementation complexity and signaling overhead. Therefore, Wi-Fi Access Points have to rely on indirect mechanisms to estimate the downlink channel quality. In this work, we describe a method to estimate the downlink Signal-to-Noise Ratio (SNR) of a Wi-Fi client, based on the measurement of uplink transmission metrics. We apply a ML-based approach to develop a predictive model for the same and validate it using actual measurements in a campus network.