EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

BSTAR Parameter Correction and Improved Orbit Prediction Accuracy for Low-Earth Orbit Space Objects Based on LR-MS-TCN

Bingxu Yao, Xin Chen, Bowen Yang

IEEE Access · 2026 · 14 · 88083-88094

DOI: 10.1109/access.2026.3698264 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Bingxu Yao und 2 weiteren Autor:innen. Die Arbeit nennt 19 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid growth of low-Earth-orbit (LEO) satellites has intensified orbital congestion and collision risks, wherein accurate orbit prediction is critical for space situational awareness and mission safety. Atmospheric drag, represented by the BSTAR parameter in two-line element (TLE) sets, dominates the orbital uncertainty of LEO targets below 800 km. However, traditional statistical methods and shallow neural networks struggle with long-sequence nonlinear time-varying perturbations and error accumulation in long-horizon extrapolation. To address these limitations, this study proposes a local residual multi-scale temporal convolutional network (LR-MS-TCN) model for BSTAR parameter correction to enhance LEO orbit prediction accuracy. The model integrates a multi-scale feature fusion module to extract perturbation features at different temporal scales, adopts a dilated causal TCN backbone to capture long-range temporal dependencies, and introduces a local residual compensation branch with a gated fusion mechanism to strengthen the representation of short-term sensitive variations near the current epoch. Experiments are conducted on 16 years of TLE data of the FENGYUN-3A satellite, with MAE and RMSE as evaluation metrics. Results show that the proposed model achieves 1.6%, 17.1%, 23.7%, and 27.6% 3D RMSE improvement at 24 h, 48 h, 72 h, and 96 h prediction horizons, respectively, outperforming MSF-TCN, PlainTCN, SingleScaleTCN, LSTM, and GRU. It also maintains stable performance under quiet, storm, and strong geomagnetic storm conditions and exhibits favorable cross-target generalization on the FENGYUN-3B satellite. The LR-MS-TCN model effectively improves BSTAR correction accuracy and long-horizon orbit prediction stability, providing a reliable data-driven approach for high-precision LEO orbital propagation.

Zitieren

APA

Yao, B., Chen, X., & Yang, B. (2026). BSTAR Parameter Correction and Improved Orbit Prediction Accuracy for Low-Earth Orbit Space Objects Based on LR-MS-TCN. IEEE Access, 14, 88083-88094. https://doi.org/10.1109/access.2026.3698264

Vancouver

Yao B, Chen X, Yang B. BSTAR Parameter Correction and Improved Orbit Prediction Accuracy for Low-Earth Orbit Space Objects Based on LR-MS-TCN. IEEE Access. 2026;14:88083-88094. doi:10.1109/access.2026.3698264

BibTeX
@article{Yao2026,
  author = {Yao, Bingxu and Chen, Xin and Yang, Bowen},
  title = {BSTAR Parameter Correction and Improved Orbit Prediction Accuracy for Low-Earth Orbit Space Objects Based on LR-MS-TCN},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {88083--88094},
  doi = {10.1109/access.2026.3698264}
}
Kostenlos starten →