Attention-Based Blind Adaptive Receive Beamforming for Interference Limited NGSO Satellite Systems

This paper presents an interference mitigation framework that can be applied on the user side for Non-Geostationary Satellite Orbit (NGSO) systems that share adjacent, overlapped frequencies to prevent unintentional co-frequency interference (CFI) scenarios. We introduce a novel Attention-based Beamformer (AttBF) model and explore its blind adaptive beamforming capabilities at the user terminals (UTs) side for spatial NGSO-to-NGSO downlink interference nulling, utilizing estimation-free data (e.g., received time-domain signals, frequency-domain representations, and sample covariance matrices (SCMs)) as direct inputs. We present a comprehensive performance evaluation of the proposed AttBF model against traditional deep learning (DL) models across various interference scenarios, encompassing both low spatial correlation (at UT’s side-lobe) and high spatial correlation (at UT’s main-lobe). To facilitate this research and future investigations into the interference management of NGSO systems, we implement innovative and extensive realistic satellite orbiting simulation and data generation methodologies, introducing new open datasets for the community. The results demonstrate that the proposed AttBF-based beamformer, particularly when employing SCMs input, achieves superior performance in mitigating interference compared to time-and frequency-domain inputs. Our findings highlight the enhanced nulling capabilities of the AttBF-based approach compared to DL-based models, such as convolutional neural networks (CNNs), and traditional methods, including zero forcing beamformer (ZFBF) and sample matrix inversion (SMI), underscoring the potential of advanced DL techniques for improving the reliability and efficiency of NGSO systems.