LEO Uydu Takımyıldızları için Alanlar Arası Transfer Öğrenmesi: Benzerlik Tabanlı Çoklu Kaynak Yaklaşımı

Öz

Starlink, OneWeb ve Iridium gibi Alçak Dünya Yörüngesi (LEO) uydu takımyıldızlarının hızlı genişlemesi, küresel bağlantı için büyük fırsatlar yaratırken yörünge tahmini, trafik yönetimi ve kaynak tahsisi gibi alanlarda önemli zorluklar ortaya çıkarmıştır. Geleneksel yörünge yayılım modelleri (ör. SGP-4) ve fizik tabanlı yaklaşımlar, atmosferik sürtünme, uzay havası ve yörünge heterojenliği nedeniyle genellikle gerekli doğruluğu sağlayamamaktadır. Makine öğrenimi (ML) teknikleri tahmin doğruluğunu artırmada güçlü bir potansiyele sahip olsa da, büyük ve kaliteli veri kümelerine bağımlılıkları yeni veya gelişmekte olan takımyıldızlar için sınırlayıcıdır. Bu çalışma, heterojen takımyıldızlar arasındaki yörünge benzerliklerinden yararlanarak minimum hedef veriyle doğru yörünge periyodu tahmini sağlayan benzerlik tabanlı çok kaynaklı transfer öğrenimi (MSTL) çerçevesini sunmaktadır. Performansı %461’e kadar düşürdüğü gösterilen geleneksel fizik tabanlı özellik mühendisliğinden farklı olarak, yöntemimiz Two-Line Element (TLE) verilerinden doğrudan çıkarılan minimalist bir özellik seti (yükseklik, eğim, eksantriklik) kullanır. Benzerlik odaklı kaynak seçimi ve filtrelenmiş çoklu bilgi entegrasyonu sayesinde, önerilen çerçeve yalnızca 25 etiketli hedef örnekle tahmin hatasını %88,2 azaltır (RMSE = 0,045 dk, R² = 0,9972). Sonuçlar, alan farkındalıklı benzerlik filtrelemenin karmaşık özellik mühendisliğinden daha iyi performans gösterdiğini ve fizik tabanlı alanlarda transfer öğrenimine dair geleneksel varsayımları sorguladığını ortaya koymaktadır. Çalışma, yeni LEO operatörleri için ölçeklenebilir, hesaplama açısından verimli ve pratik bir çözüm sunarak kapsamlı veri toplama gereksinimini ortadan kaldırır.

Anahtar Kelimeler

• Transfer Öğrenimi
• LEO Uyduları
• Alan Uyarlama
• Yörünge Tahmini
• Uydu Konstellasyonları

Cross-Domain Transfer Learning for LEO Satellite Constellations: A Similarity-Based Multi-Source Approach

Öz

The rapid expansion of Low Earth Orbit (LEO) satellite constellations such as Starlink, OneWeb, and Iridium has created new opportunities for global connectivity while introducing major challenges in orbit prediction, traffic management, and resource allocation. Traditional orbit propagation models (e.g., SGP-4) and physics-informed approaches often fail to meet accuracy requirements due to atmospheric drag, space weather, and orbital heterogeneity. Although machine learning (ML) techniques show strong potential for improving prediction accuracy, their dependence on large, high-quality datasets limits their applicability to new constellations. This paper presents a similarity-based multi-source transfer learning (MSTL) framework that leverages orbital similarities across heterogeneous constellations to enable accurate orbital period prediction with minimal target data. Unlike conventional physics-informed feature engineering, which can degrade performance by up to 461%, our method employs a minimalist feature set (altitude, inclination, and eccentricity) directly extracted from Two-Line Element (TLE) data. Through similarity-driven source selection and filtered multi-source knowledge integration, the proposed framework reduces prediction error by 88.2% (RMSE = 0.045 min, R² = 0.9972) using only 25 labeled samples from the target constellation. The findings show that domain-aware similarity filtering outperforms complex feature engineering, challenging conventional assumptions about transfer learning in physics-based domains. This work offers a scalable, efficient, and practical solution for emerging LEO operators, enabling rapid model development without extensive data collection.

Anahtar Kelimeler

• Transfer Learning
• LEO Satellites
• Domain Adaptation
• Orbital Prediction
• Satellite Constellations

Kaynakça

1. [1] Z. M. Kassas, S. Kozhaya, H. Kanj, J. Saroufim, S. W. Hayek, M. Neinavaie, N. Khairallah, and J. Khalife, “Navigation with multi-constellation LEO satellite signals of opportunity: Starlink, OneWeb, Orbcomm, and Iridium,” IEEE/ION Position, Location and Navigation Symposium (PLANS), pp. 338–343, (2023).
2. [2] Y. Su, Y. Liu, Y. Zhou, J. Yuan, H. Cao, and J. Shi, “Broadband LEO satellite communications: Architectures and key technologies,” IEEE Wireless Communications, vol. 26, no. 2, pp. 55–61, (2019).
3. [3] C. Wang, D. Bian, S. Shi, J. Xu, and G. Zhang, “A novel cognitive satellite network with GEO and LEO broadband systems in the downlink case,” IEEE Access, vol. 6, pp. 25987–26000, (2018).
4. [4] O. B. Osoro, E. J. Oughton, A. R. Wilson, and A. Rao, “Sustainability assessment of low Earth orbit (LEO) satellite broadband megaconstellations,” arXiv preprint, arXiv:2309.02338, (2023).
5. [5] Z. M. Kassas, S. Hayek, and J.-H. Ahmad, “LEO satellite orbit prediction via closed-loop machine learning with application to opportunistic navigation,” IEEE Aerospace and Electronic Systems Magazine, vol. 40, no. 1, pp. 34–49, (2025).
6. [6] K. Wang, J. Liu, H. Su, A. El-Mowafy, and X. Yang, “Real-time LEO satellite orbits based on batch least-squares orbit determination with short-term orbit prediction,” Remote Sensing, vol. 15, no. 1, (2023).
7. [7] X. Mao, D. Arnold, M. Kalarus, S. Padovan, and A. Jäggi, “GNSS-based precise orbit determination for maneuvering LEO satellites,” GPS Solutions, vol. 27, no. 3, p. 147, (2023).
8. [8] C. Westphal, L. Han, and R. Li, “LEO satellite networking relaunched: Survey and current research challenges,” ITU Journal on Future and Evolving Technologies, vol. 4, pp. 711–744, (2023).

9. [9] G. Acciarini, A. G. Baydin, and D. Izzo, “Closing the gap between SGP4 and high-precision propagation via differentiable programming,” Acta Astronautica, vol. 226, pp. 694–701, (2025).
10. [10] Y.-M. Guan, Z.-H. Li, and X.-L. Gao, “Research on intelligent orbit life prediction model of low Earth orbit vehicle,” Aerospace Science and Technology, vol. 159, p. 109977, (2025).
11. [11] Z. M. Kassas and J. Saroufim, “LEO PNT frameworks for non-cooperative satellites with poorly known ephemerides: Open-loop SGP4, tracking, and differential,” IEEE Aerospace and Electronic Systems Magazine, pp. 1–15, (2025).
12. [12] Y.-M. Guan and Z.-H. Li, “Research on intelligent computing model of aerodynamic fusion 6-DOF trajectory and attitude of low-orbit vehicle,” Celestial Mechanics and Dynamical Astronomy, vol. 136, no. 6, p. 59, (2024).
13. [13] G. Marie and D. G. Ghazi, “TLE prediction using machine learning for satellite maneuver detection,” Regional Student Conferences, p. 98101, (2025).
14. [14] Y. Ying, Z. Ziyu, Z. Qingrui, M. Rong, and C. Pei, “Maneuver detection of cataloged space object based on machine learning,” Chinese Control Conference (CCC), pp. 6038–6043, (2024).
15. [15] L. Jia, C. Ezhilarasu, and I. Jennions, “A review of the application of transfer learning in fault diagnosis and its potential in aerospace condition based maintenance,” Preprints, (2025).
16. [16] M. Hedayati, A. Barzegar, and A. Rahimi, “Fault diagnosis and prognosis of satellites and unmanned aerial vehicles: A review,” Applied Sciences, vol. 14, no. 20, (2024).
17. [17] Z. Zhang, “From orbit to infrastructure: Transfer learning across scales,” Springer Nature Singapore, pp. 151–191, (2025).
18. [18] admin, “United Kingdom satellite communications research report,” Serumset (ResearchAndMarkets.com), (2024).
19. [19] T. Mortlock and Z. Kassas, “Assessing machine learning for LEO satellite orbit determination in simultaneous tracking and navigation,” IEEE Aerospace Conference, pp. 1–8, (2021).
20. [20] H. Liu, S. Yu, and X. Wang, “Mega-constellation satellite maneuver forecast via network with attention mechanism,” Advances in Space Research, vol. 75, no. 6, pp. 4942–4962, (2025).
21. [21] S. Wang, Z. Jiang, X. Li, and J. Ma, “MLDR: An O(-v-/4) and near-optimal routing scheme for LEO mega-constellations,” IEEE Internet of Things Journal, vol. 11, no. 10, pp. 18663–18675, (2024).
22. [22] G. K. Kumar and P. Sankar, “Artificial intelligence powered satellite communications and Sentinel satellite constellations: An overview and future perspectives,” SN Computer Science, vol. 6, no. 6, p. 735, (2025).
23. [23] J. Zhang, Y. Cai, C. Xue, Z. Xue, and H. Cai, “LEO mega constellations: Review of development, impact, surveillance, and governance,” Space: Science & Technology, pp. 1–17, (2022).
24. [24] G. Di Mauro, M. Lawn, and R. Bevilacqua, “Survey on guidance navigation and control requirements for spacecraft formation-flying missions,” Journal of Guidance, Control, and Dynamics, vol. 41, no. 3, pp. 581–602, (2018).
25. [25] A. Sajit, A. Gupta, R. Kukar, and S. Ramakrishnan, “Deployment and maintenance of satellite constellations for strategic applications,” International Conference on Advances in Aerospace and Navigation Systems, pp. 297–306, (2024).
26. [26] A. Esmaeilkhah and R. J. Landry, “Spatial sensitivity of navigation using signal-of-opportunity (SOOP) from Starlink, Iridium-Next, GlobalStar, OneWeb, and Orbcomm constellations,” Engineering Proceedings, vol. 88, no. 1, (2025).
27. [27] H. Peng and X. Bai, “Comparative evaluation of three machine learning algorithms on improving orbit prediction accuracy,” Astrodynamics, vol. 3, no. 4, pp. 325–343, (2019).
28. [28] H. Peng and X. Bai, “Machine learning approach to improve satellite orbit prediction accuracy using publicly available data,” The Journal of the Astronautical Sciences, vol. 67, no. 2, pp. 762–793, (2020).
29. [29] M. Zhai, Z. Huyan, Y. Hu, Y. Jiang, and H. Li, “Improvement of orbit prediction accuracy using extreme gradient boosting and principal component analysis,” Open Astronomy, vol. 31, no. 1, pp. 229–243, (2022).
30. [30] H. Peng and X. Bai, “Gaussian processes for improving orbit prediction accuracy,” Acta Astronautica, vol. 161, pp. 44–56, (2019).
31. [31] J. Pihlajasalo, H. Leppäkoski, S. Ali-Löytty, and R. Piché, “Improvement of GPS and BeiDou extended orbit predictions with CNNs,” European Navigation Conference (ENC), pp. 54–59, (2018).
32. [32] B. Li, Y. Zhang, J. Huang, and J. Sang, “Improved orbit predictions using two-line elements through error pattern mining and transferring,” Acta Astronautica, vol. 188, pp. 405–415, (2021).
33. [33] G. Curzi, D. Modenini, and P. Tortora, “Two-line-element propagation improvement and uncertainty estimation using recurrent neural networks,” CEAS Space Journal, vol. 14, no. 1, pp. 197–204, (2022).
34. [34] H. Ren, X. Chen, B. Guan, Y. Wang, T. Liu, and K. Peng, “Research on satellite orbit prediction based on neural network algorithm,” High Performance Computing and Cluster Technologies Conference (HPCCT), pp. 267–273, (2019).
35. [35] F. Caldas and C. Soares, “Precise and efficient orbit prediction in LEO with machine learning using exogenous variables,” IEEE Congress on Evolutionary Computation (CEC), pp. 1–8, (2024).
36. [36] Y. Zhang, Y. Wang, H. Cao, Y. Hu, Z. Lin, K. An, and D. Li, “Self-similar traffic prediction for LEO satellite networks based on LSTM,” IET Communications, vol. 19, no. 1, (2024).
37. [37] B. Stock, M. McVarthy, and S. Servadio, “A data-driven approach to estimate LEO orbit capacity models,” arXiv preprint, arXiv:2507.19365, (2025).
38. [38] H. Xu, J. Liao, Y. Luo, and Y. Meng, “A VMD-SVM method for LEO satellite orbit prediction with space weather parameters,” Remote Sensing, vol. 17, no. 5, (2025).
39. [39] W. Zhang, K. Zhang, X. Li, and J. Huang, “Improving LEO short-term orbit prediction using LSTM neural network,” Advances in Space Research, vol. 76, no. 1, pp. 481–496, (2025).
40. [40] S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Transactions on Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345–1359, (2009).
41. [41] F. Zhuang, Z. Qi, K. Duan, D. Xi, Y. Zhu, H. Zhu, H. Xiong, and Q. He, “A comprehensive survey on transfer learning,” Proceedings of the IEEE, vol. 109, no. 1, pp. 43–76, (2021).
42. [42] H. Peng and X. Bai, “Improving orbit prediction accuracy through supervised machine learning,” Advances in Space Research, vol. 61, no. 10, pp. 2628–2646, (2018).
43. [43] H. Chen, H. Luo, B. Huang, B. Jiang, and O. Kaynak, “Transfer learning-motivated intelligent fault diagnosis designs: A survey, insights, and perspectives,” IEEE Transactions on Neural Networks and Learning Systems, vol. 35, no. 3, pp. 2969–2983, (2023).
44. [44] D. Tuia, C. Persello, and L. Bruzzone, “Domain adaptation for the classification of remote sensing data: An overview of recent advances,” IEEE Geoscience and Remote Sensing Magazine, vol. 4, no. 2, pp. 41–57, (2016).
45. [45] A. M. Ali, A. Gupta, and H. A. Hashim, “Deep reinforcement learning for sim-to-real policy transfer of VTOL-UAVs offshore docking operations,” Applied Soft Computing, vol. 162, p. 111843, (2024).
46. [46] X. B. Peng, M. Andrychowicz, W. Zaremba, and P. Abbeel, “Sim-to-real transfer of robotic control with dynamics randomization,” IEEE International Conference on Robotics and Automation (ICRA), pp. 3803–3810, (2018).
47. [47] K. Weiss, T. M. Khoshgoftaar, and D. Wang, “A survey of transfer learning,” Journal of Big Data, vol. 3, no. 1, p. 9, (2016).
48. [48] L. Li, L. Zhu, and W. Li, “HISATFL: A hierarchical federated learning framework for satellite networks with cross-domain privacy adaptation,” Electronics, vol. 14, no. 16, p. 3237, (2025).
49. [49] L. Duan, D. Xu, and I. Tsang, “Learning with augmented features for heterogeneous domain adaptation,” arXiv preprint, arXiv:1206.4660, (2012).
50. [50] B. Sun, J. Feng, and K. Saenko, “Return of frustratingly easy domain adaptation,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 30, (2016).
51. [51] H. Zhao, S. Zhang, G. Wu, J. M. Moura, J. P. Costeira, and G. J. Gordon, “Adversarial multiple source domain adaptation,” Advances in Neural Information Processing Systems, vol. 31, (2018).
52. [52] D. Li, W. Xie, Y. Li, and L. Fang, “FedFusion: Manifold-driven federated learning for multi-satellite and multi-modality fusion,” IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–13, (2023).
53. [53] R. Sasso, M. Sabatelli, and M. A. Wiering, “Multi-source transfer learning for deep model-based reinforcement learning,” arXiv preprint, (2022).
54. [54] M. Long, Y. Cao, J. Wang, and M. Jordan, “Learning transferable features with deep adaptation networks,” International Conference on Machine Learning (ICML), pp. 97–105, (2015).
55. [55] Y. Ganin and V. Lempitsky, “Unsupervised domain adaptation by backpropagation,” International Conference on Machine Learning (ICML), pp. 1180–1189, (2015).
56. [56] W. Zhang, L. Deng, L. Zhang, and D. Wu, “A survey on negative transfer,” IEEE/CAA Journal of Automatica Sinica, vol. 10, no. 2, pp. 305–329, (2022).
57. [57] M. J. Sorocky, S. Zhou, and A. P. Schoellig, “Experience selection using dynamics similarity for efficient multi-source transfer learning between robots,” IEEE International Conference on Robotics and Automation (ICRA), pp. 2739–2745, (2020).
58. [58] www.celestrak.org, “Celestrak: Current NORAD Two-Line Element Sets”, (2020).
59. [59] www.celestrak.org/NORAD/ELEMENTS/table.php?GROUP=starlink&FORMAT=json-pretty, “Starlink Satellite Constellation Data”, (2024).
60. [60] www.celestrak.org/NORAD/ELEMENTS/table.php?GROUP=oneweb&FORMAT=json-pretty, “OneWeb Satellite Constellation Data”, (2024).
61. [61] www.celestrak.org/NORAD/ELEMENTS/table.php?GROUP=iridium&FORMAT=json-pretty, “Iridium Satellite Constellation Data”, (2024).
62. [62] F. R. Hoots and R. L. Roehrich, “Models for propagation of NORAD element sets,” Aerospace Defense Command, United States Air Force, (1980).
63. [63] D. A. Vallado, P. Crawford, R. Hujsak, and T. S. Kelso, “Revisiting Spacetrack Report #3: Rev 2,” AIAA Astrodynamics Specialist Conference and Exhibit, (2006).
64. [64] H. D. Curtis, “Orbital mechanics for engineering students,” Butterworth-Heinemann, 4th ed., (2019).