Hao Peng, Rutgers, The State University of New Jersey, Xiaoli Bai, Rutgers, Rutgers, The State University of New Jersey
Keywords: Space Situational Awareness, Orbit Prediction, Machine Learning
Abstract:
With the assumption that a machine learning (ML) approach can learn the underlying pattern of the orbit prediction errors from historical data, in this paper, the support vector machine (SVM) is used to improve the accuracy of orbit prediction of resident space object (RSO) in the future. We first show that the trained SVM model can capture the relationship between the chosen learning variables and the target orbit prediction error with both good average and individual performances. Then through a series of experiments, we show that the performance can be further improved with more training data, until adequate data is provided. Moreover, the correction capability of the trained SVM model is limited to the future horizon and its generalization capability will be reduced greatly if the orbit is predicted too far in the future. At the last part of the paper, the effect of model and measurement errors are investigated, including an idealistic case without any error. The results show that the residual errors after the ML-modification will increase as the measurement error in the system increases, but the trained SVM still shows good capability to improve the orbit prediction accuracy. Some insights for future studies are also provided in the paper.
Date of Conference: September 19-22, 2017
Track: Astrodynamics