Switching in the Rain: Predictive Wireless x-haul Network Reconfiguration

4G, 5G, and smart city networks often rely on microwave millimeter-wave x-haul links. A major challenge associated with these high frequency links is their susceptibility to weather conditions. In particular, precipitation may cause severe signal attenuation, which significantly degrades the network performance. this paper, we develop a Predictive Network Reconfiguration (PNR) framework that uses historical data predict future condition of each link then prepares ahead time for imminent disturbances. The PNR has two components: (i) an Attenuation Prediction (AP) mechanism; (ii) Multi-Step (MSNR) algorithm. AP mechanism employs encoder-decoder Long Short-Term Memory (LSTM) model sequence attenuation levels link. MSNR algorithm leverages predictions dynamically optimize routing admission control decisions aiming maximize utilization, while preserving max-min fairness among nodes using (e.g., base-stations) preventing transient congestion be caused by switching routes. We train, validate, evaluate dataset containing over 2 million measurements collected from real-world city-scale backhaul network. results show framework: predicts accuracy, RMSE less than 0.4 dB prediction horizon 50 seconds; can improve instantaneous utilization more 200% when compared reactive reconfiguration algorithms cannot leverage information about