![[2602.21116] Attention-Based SINR Estimation in User-Centric Non-Terrestrial Networks](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.21116] Attention-Based SINR Estimation in User-Centric Non-Terrestrial Networks
Summary
This paper presents a novel low-complexity framework for estimating the signal-to-interference-plus-noise ratio (SINR) in user-centric non-terrestrial networks using multi-head self-attention mechanisms.
Why It Matters
As satellite-based non-terrestrial networks become more prevalent, efficient SINR estimation is crucial for optimizing performance. This research addresses the computational challenges of existing methods, offering a more efficient alternative that can enhance user experience and network capacity.
Key Takeaways
- Introduces a dual multi-head self-attention (DMHSA) model for SINR estimation.
- Reduces computational complexity significantly compared to traditional methods.
- Maintains high estimation accuracy with root mean squared error below 1 dB.
- Facilitates the evaluation of multiple user groups for optimal scheduling.
- Supports the integration of DMHSA estimators into existing scheduling procedures.
Electrical Engineering and Systems Science > Signal Processing arXiv:2602.21116 (eess) [Submitted on 24 Feb 2026] Title:Attention-Based SINR Estimation in User-Centric Non-Terrestrial Networks Authors:Bruno De Filippo, Alessandro Guidotti, Alessandro Vanelli-Coralli View a PDF of the paper titled Attention-Based SINR Estimation in User-Centric Non-Terrestrial Networks, by Bruno De Filippo and 2 other authors View PDF HTML (experimental) Abstract:The signal-to-interference-plus-noise ratio (SINR) is central to performance optimization in user-centric beamforming for satellite-based non-terrestrial networks (NTNs). Its assessment either requires the transmission of dedicated pilots or relies on computing the beamforming matrix through minimum mean squared error (MMSE)-based formulations beforehand, a process that introduces significant computational overhead. In this paper, we propose a low-complexity SINR estimation framework that leverages multi-head self-attention (MHSA) to extract inter-user interference features directly from either channel state information or user location reports. The proposed dual MHSA (DMHSA) models evaluate the SINR of a scheduled user group without requiring explicit MMSE calculations. The architecture achieves a computational complexity reduction by a factor of three in the CSI-based setting and by two orders of magnitude in the location-based configuration, the latter benefiting from the lower dimensionality of user reports. We show that both D...
