Predictive Channel Estimation and Optimal Switching Control via LSTM in Q/V‐Band Satellite Networks With Smart Gateway Diversity

ABSTRACT The growing demand for broadband satellite connectivity is accelerating the transition toward extremely high frequency (EHF) spectrum, particularly the Q/V‐band (40–50 GHz), which provides substantially larger bandwidth resources compared with conventional Ka‐band allocations and supports the deployment of next‐generation high‐throughput satellite (HTS) systems. Despite these advantages, Q/V‐band links are vulnerable to atmospheric impairments, with rain attenuation representing a primary source of signal degradation and link outages. Ensuring service continuity in this context requires accurate channel state prediction to enable proactive fade mitigation techniques (FMT), such as smart gateway diversity (SGD), which leverages geographically distributed gateways and dynamic resource management to counteract weather‐induced fading. This work investigates artificial intelligence–based channel prediction strategies to enhance SGD performance, using experimental data collected during the Alphasat TDP#5 Q/V‐band campaign. A medium‐term forecasting model based on a long short‐term memory (LSTM) neural network is developed to estimate the future probability of link outage. From a scientific research perspective, the adoption of the link outage probability itself as a driving metric for the SGD decision process represents an innovative aspect of this work, enabling a predictive and reliability‐oriented gateway management strategy. The model is trained and validated on real measurements acquired from the Aldo Paraboni experimental payload. Comparative results demonstrate that the proposed LSTM approach demonstrates improved performance over persistence‐based (Naïve) and linear prediction models, substantially reducing forecasting errors and enabling timely, seamless gateway handovers. These findings confirm the effectiveness of machine learning (ML) techniques in improving the reliability and resilience of next‐generation EHF satellite networks.