Machine Learning Argument of Latitude Error Model for LEO Satellite Orbit and Covariance Correction
arXiv:2602.16764v1 Announce Type: new Abstract: Low Earth orbit (LEO) satellites are leveraged to support new position, navigation, and timing (PNT) service alternatives to GNSS. These alternatives require accurate propagation of satellite position and velocity with a realistic quantification of uncertainty. It is commonly assumed that the propagated uncertainty distribution is Gaussian; however, the validity of this assumption can be quickly compromised by the mismodeling of atmospheric drag. We develop a machine learning approach that corrects error growth in the argument of latitude for a diverse set of LEO satellites. The improved orbit propagation accuracy extends the applicability of the Gaussian assumption and modeling of the errors with a corrected mean and covariance. We compare the performance of a time-conditioned neural network and a Gaussian Process on datasets computed with an open source orbit propagator and publicly available Vector Covariance Message (VCM) ephemerides
arXiv:2602.16764v1 Announce Type: new Abstract: Low Earth orbit (LEO) satellites are leveraged to support new position, navigation, and timing (PNT) service alternatives to GNSS. These alternatives require accurate propagation of satellite position and velocity with a realistic quantification of uncertainty. It is commonly assumed that the propagated uncertainty distribution is Gaussian; however, the validity of this assumption can be quickly compromised by the mismodeling of atmospheric drag. We develop a machine learning approach that corrects error growth in the argument of latitude for a diverse set of LEO satellites. The improved orbit propagation accuracy extends the applicability of the Gaussian assumption and modeling of the errors with a corrected mean and covariance. We compare the performance of a time-conditioned neural network and a Gaussian Process on datasets computed with an open source orbit propagator and publicly available Vector Covariance Message (VCM) ephemerides. The learned models predict the argument of latitude error as a Gaussian distribution given parameters from a single VCM epoch and reverse propagation errors. We show that this one-dimensional model captures the effect of mismodeled drag, which can be mapped to the Cartesian state space. The correction method only updates information along the dimensions of dominant error growth, while maintaining the physics-based propagation of VCM covariance in the remaining dimensions. We therefore extend the utility of VCM ephemerides to longer time horizons without modifying the functionality of the existing propagator.
Executive Summary
The article presents a machine learning approach to correct error growth in the argument of latitude for LEO satellites, addressing the challenge of mismodeled atmospheric drag that compromises the Gaussian assumption of uncertainty distribution. The authors develop and compare time-conditioned neural networks and Gaussian Processes to predict argument of latitude error as a Gaussian distribution, using open-source orbit propagator data and publicly available VCM ephemerides. The method extends the utility of VCM ephemerides to longer time horizons by correcting dominant error growth dimensions while preserving physics-based propagation in others.
Key Points
- ▸ Machine learning models are used to correct error growth in LEO satellite orbits due to mismodeled atmospheric drag.
- ▸ Time-conditioned neural networks and Gaussian Processes are compared for their performance in predicting argument of latitude error.
- ▸ The correction method extends the applicability of the Gaussian assumption and improves orbit propagation accuracy.
- ▸ The approach maintains physics-based propagation of VCM covariance in non-dominant error dimensions.
Merits
Innovative Approach
The use of machine learning to correct orbit propagation errors is innovative and addresses a critical issue in LEO satellite operations.
Practical Application
The method extends the utility of VCM ephemerides, making it practical for longer time horizons without modifying existing propagators.
Comprehensive Comparison
The comparison between time-conditioned neural networks and Gaussian Processes provides valuable insights into model performance.
Demerits
Limited Scope
The study focuses on the argument of latitude error, which may not capture all aspects of orbit propagation errors.
Data Dependency
The effectiveness of the models is dependent on the quality and availability of open-source orbit propagator data and VCM ephemerides.
Generalizability
The models' performance may vary with different LEO satellite configurations and atmospheric conditions.
Expert Commentary
The article presents a significant advancement in the field of satellite orbit propagation by leveraging machine learning to address the critical issue of mismodeled atmospheric drag. The innovative use of time-conditioned neural networks and Gaussian Processes to correct error growth in the argument of latitude demonstrates a robust approach to enhancing orbit propagation accuracy. The method's ability to extend the applicability of the Gaussian assumption and maintain physics-based propagation in non-dominant error dimensions is particularly noteworthy. However, the study's focus on a one-dimensional error model may limit its comprehensive applicability. Future research could explore the integration of additional error dimensions and the impact of varying atmospheric conditions on model performance. The practical implications of this research are substantial, particularly for improving PNT services and space traffic management. Policymakers and industry stakeholders should take note of the potential benefits of investing in machine learning technologies for aerospace applications.
Recommendations
- ✓ Further research should explore the integration of additional error dimensions to provide a more comprehensive correction model.
- ✓ The impact of varying atmospheric conditions on model performance should be thoroughly investigated to ensure robustness across different scenarios.
