|
Beam-hopping optimization for regenerative satellites using a hybrid LSTM-transformer model
|
Authors: Issiakou Worou Wouignanso, Olfa Ben Yahia, Stephane Martel, Guillaume Mantelet, Gunes Karabulut Kurt, Brunilde Sanso
Status: Final
Date of publication: 30 September 2026
Published in: ITU Journal on Future and Evolving Technologies, Volume 7 (2026), Issue 3, Pages 340-356
Article DOI : https://doi.org/10.52953/GSZQ4141
|
Abstract:
This paper presents a hybrid Long Short-Term Memory (LSTM)-transformer model for optimizing Beam Hopping (BH) in multibeam satellite communication systems. BH is a dynamic resource allocation technique that improves spectral efficiency by selectively illuminating ground regions based on traffic demand and channel conditions. The proposed system takes as inputs traffic demand measurements, Channel State Information (CSI), and historical allocation data, and produces as output a BH schedule together with associated Key Performance Indicators (KPIs). The architecture combines three stacked bidirectional LSTM layers, which capture temporal traffic dynamics at multiple scales, with a transformer encoder, which identifies long-range inter-beam spatial correlations through multi-head self-attention. Training is driven by a four-term multi-objective loss that jointly optimizes Spectral Efficiency (SE), Jain's fairness, Quality of Service (QoS) compliance and power utilization. Monte Carlo simulations with stochastic channel and traffic realizations demonstrate that the proposed method improves SE compared with the Deep learning-Based Optimization (DBO) benchmark. The hybrid model also outperforms a pure transformer baseline trained under identical conditions, while reducing training time and maintaining a favourable trade-off between performance and computational cost. In addition, the proposed approach achieves balanced performance in terms of fairness, QoS compliance, power utilization and Signal-to-Interference-plus-Noise Ratio (SINR). A sensitivity analysis of the loss-weighting coefficients further characterizes the trade-off between SE and fairness. |
Keywords: Beam hopping (BH), Long Short-Term Memory (LSTM), multibeam, resource allocation, satellites, spectral efficiency, transformer
Rights: © International Telecommunication Union, available under the CC BY-NC-ND 3.0 IGO license.
|
|
| ITEM DETAIL | ARTICLE | PRICE | |
|---|
ENGLISH 
Full article (PDF) |
|
| 0
| Free of charge | DOWNLOAD |
|
|
| |