상세정보
학술대회
Machine Learning Approach to Initial Orbit Determination of Unknown LEO Satellites
- 발행일
- 201805
- 출처
- International Conference on Space Operations (SpaceOps) 2018, pp.1-11
- 초록
- Initial orbit determination should be performed to know the unknown satellite orbit with minimum observation data sets from ground station. Satellite tracking system such as optical, radar, and/or radio can acquire and track the satellite based on the initial orbit. Then a normal operational orbit determination is carried out with enough tracking data sets. However, initial acquisition and tracking of the unknown satellite are very difficult with conventional optical telescope, directional radar, and parabolic radio antenna system. When the unknown satellite transmits downlink radio signal, an omnidirectional radio antenna with wide-band receiving capability can acquire the Doppler signal from the satellite. A series of Doppler observations are converted to range rates between the satellite and ground station. In this paper, only one real ground station is defined to measure range rate of the satellite and two virtual ground stations are assumed to predict range rate and slant range. Prediction of range rate and slant range is carried out by machine learning. One month of simulated tracking data set, i.e. range rate and slant range, for the three ground stations are used in the training. Given a satellite pass at the real ground station, predicted satellite tracking data from two virtual ground stations are generated by machine learning. Then, trilateration orbit determination is performed with range rate and slant range data sets from the three ground stations. This study shows that the regression of the machine learning is applicable to the initial orbit determination of the unknown satellites.
- KSP 제안 키워드
- Data sets, Doppler signals, Ground Station, Initial orbit determination, LEO satellite, Machine Learning Approach, Radio Signal, Range data, Satellite tracking data, Satellite tracking system, Virtual ground