Integrált LEO műholdas és földi hálózatok QoS jellemzőinek predikciója autoenkóder alapú gépi tanulással
Prediction of QoS characteristics of integrated LEO satellite and terrestrial networks using autoencoder-based machine learning
Keywords:
Starlink, Non-Terrestrial Network, Autoencoder, Low Earth Orbit Satellites, Round-Trip Time, Machine Learning, /, nem Földi hálózat, autoenkóder, alacsony Föld körüli pályán keringő műholdak, körforgási idő, gépi tanulásAbstract
We analyzed over 1 million RTT samples from Starlink’s LEO system collected over 30 days, revealing significant fine-grained temporal variability. The RTT distribution shows dual peaks, indicating dynamic link-layer modes. Hourly aggregated features capture behavioral trends, while an autoencoder-based imputation reconstructs missing data effectively. We also evaluate RTT characteristics under VPN usage, highlighting performance shifts. These insights inform good modelling and robust interpretation of RTT behavior in NTN-TN integrated networks.
Kivonat
A Starlink LEO rendszeréből 30 nap alatt gyűjtött több mint 1 millió RTT-mintát elemeztünk, amely során jelentős, finomszemcsés időbeli változékonyságot tártunk fel. Az RTT-eloszlás kettős csúcsokat mutat, ami dinamikus adatkapcsolati üzemmódokat sugall a szolgáltatásra vonatkozóan. Az óránkénti összesített jellemzők rögzítik a viselkedési trendeket, míg az autoenkóder hatékonyan rekonstruálja a hiányzó adatokat. VPN-használat esetén is értékeljük az RTT-jellemzőket, kiemelve a teljesítménybeli eltolódásokat. Ezek a jellemző metrikák megfelelő modellezést és az RTT viselkedésének hatékony értelmezést teszik lehetővé NTN-TN integrált hálózatokban.
References
Laniewski, Dominic, et al. Wetlinks: a large-scale longitudinal starlink dataset with contiguous weather data. 2024 8th Network Traffic Measurement and Analysis Conference (TMA). IEEE, 2024.
Mohan, Nitinder, et al. A multifaceted look at starlink performance. Proceedings of the ACM Web Conference 2024. 2024.
Komodromos, Zacharias M., Wenkai Qin, and Todd E. Humphreys. Signal simulator for Starlink Ku-Band downlink. Proceedings of the 36th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2023). 2023.
Gal, Zoltan, and Djamila Talbi. Insights into low Earth orbit satellite communication dynamics: quality of service analysis of connection behavior, latency, and Doppler shift. 2024 IEEE 4th International Conference on Electronic Communications, Internet of Things and Big Data (ICEIB). IEEE, 2024.
Ramanathan, Alagappan, and Sangeetha Abdu Jyothi. Weathering a Solar Superstorm: Starlink Performance during the May 2024 Storm. Proceedings of the 2nd International Workshop on LEO Networking and Communication. 2024.
Laniewski, Dominic, et al. Starlink on the Road: A First Look at Mobile Starlink Performance in Central Europe. 2024 8th Network Traffic Measurement and Analysis Conference (TMA). IEEE, 2024.
Wong, Aaron, et al. Network Performance Analysis of Starlink: A Proliferated Low Earth Orbit Satellite Internet System. 2024 17th International Conference on Signal Processing and Communication System (ICSPCS). IEEE, 2024.
Giannakopoulos, Panagiotis, et al. perfCorrelate: Performance variability correlation framework. Future Generation Computer Systems 170 (2025): 107827.
Zhang, Menghao, et al. SuperFE: A Scalable and Flexible Feature Extractor for ML-based Traffic Analysis Applications. Proceedings of the Twentieth European Conference on Computer Systems. 2025.
Goodman, Eric L., Chase Zimmerman, and Corey Hudson. Packet2vec: Utilizing word2vec for feature extraction in packet data. arXiv preprint arXiv:2004.14477 (2020).
Kulin, Merima, et al. A survey on machine learning-based performance improvement of wireless networks: PHY, MAC and network layer. Electronics 10.3 (2021): 318.
Sarhan, Mohanad, et al. Feature extraction for machine learning-based intrusion detection in IoT networks. Digital Communications and Networks 10.1 (2024): 205-216.
Olaoye, Godwin, and Ayuns Luz. Hybrid Models for Medical Data Analysis. Available at SSRN 4742530 (2024).
Abiri, Najmeh, et al. Establishing strong imputation performance of a denoising autoencoder in a wide range of missing data problems. Neurocomputing 365 (2019): 137-146.
Jiao, Xu, et al. Improving Generalization for Missing Data Imputation via Dual Corruption Denoising Autoencoders. International Conference on Applied Computational Intelligence, Informatics and Big Data. Springer, Cham, 2025.
