ML‐Assisted Adaptive Fade Mitigation for Satellite Connectivity

ABSTRACT The demand for reliable and high‐quality wireless networks driven by resource‐intensive applications and services is growing rapidly and invoked the emergence of beyond 5G (B5G) and 6G paradigm and technologies which include satellite systems as part of the nonterrestrial networks (NTN). The satellite links, however, are severely affected by rain attenuation, particularly at the frequencies of interest for 6G systems, leading to significant signal degradation. Hence, fade mitigation techniques (FMT) are critical for sustaining the link against fade, thus improving link availability and reliability. Adaptive FMT (AFMT) is deployed to enable links to react and compensate for signal degradation based on channel conditions and fading severity. However, conventional AFMT performance is limited by the uncertainties of the channel conditions. To circumvent this obstacle, the capabilities of artificial intelligence (AI) algorithms are explored to predict channel conditions reliably and enhance AFMT performance. This paper proposes the use of deep learning models, particularly the long short‐term memory (LSTM) model in MATLAB R2020b to predict rain attenuation and optimize AFMT strategies. Exploiting its ability to process large datasets and learn complex patterns, the LSTM‐based model integrates time diversity (TD) and site diversity (SD) independently and with joined selection combining (SC) and maximal ration combining (MRC), to preemptively address signal degradation and maintain optimal quality of service (QoS). This new model is trained and validated using real rain attenuation data obtained from Chilbolton and Chilton Observatories in England. Simulation results demonstrated significant improvements, with the LSTM‐based approach achieving 90% accuracy in predicting attenuation for a 0.01% time probability. The proposed hybrid AFMT (TD + SD + MRC) strategy outperformed other configurations considered, delivering at least 21 dB gain. These findings highlight the potential of AI‐driven AFMTs to enhance QoS, enabling better planning and decision‐making for reliable and high‐quality satellite communication signals.