In this study, a procedure to associate multiple orbit solutions and obtain accurate one to predict long-term trajectory for space surveillance is proposed and analyzed. The procedure consists of three steps: single-arc orbit determination, associatio...
In this study, a procedure to associate multiple orbit solutions and obtain accurate one to predict long-term trajectory for space surveillance is proposed and analyzed. The procedure consists of three steps: single-arc orbit determination, association check, and element fitting. The unscented batch least-squares algorithm was employed to determine the state vector in Cartesian coordinates based on short single-arc observation data. The association between the single-arc orbit solutions is statistically determined by the chi-square test, which provides the probability that two solutions indicate the same object. Finally, the element fitting determines the mean orbital elements that fit the associated single-arc orbit solutions by employing the unscented Kalman filter (UKF) and smoother.
Each step was examined using optical data only, radar data only, and both optical and radar data, and the performances were validated. Through a numerical simulation using pseudo observation data, the single-arc orbit determination accuracy was observed larger than hundreds of meters for optical and radar only cases, while the combined data presented a position error of several tens of meters. A strategy for the second observation was constructed by analyzing the predicted trajectory using the single-arc orbit solution in the South-East-Zenith (SEZ) coordinates. Because the solution provided poor prediction accuracy, a time margin was required for the second observation. Through the chi-square test, the association between the single-arc orbit solutions was determined. The association was not correctly determined when the single-arc orbit solution had errors larger than few kilometers and objects flew in formation. The element fitting reconfirmed the association and provided the mean orbital elements appropriate for long time orbit prediction. Furthermore, this study employed actual optical and radar observation data provided by OWL-Net and LeoLabs, respectively, and both the reliability of the tracking data and the practicality of the proposed procedure were validated.
The proposed procedure is a standalone process that does not require any orbit database, which supports independent operation of the space situational awareness (SSA) system and development of a new database for space objects. It is robust to the a priori errors and methodologically consistent because the nonlinear unscented transform is employed for both single-arc orbit determinations and element fittings. Moreover, the separated step-by-step process provides two additional advantages. First, the association between single-arc solutions is checked and reconfirmed. The numerical simulation verified that the element fitting can distinguish neighboring objects that are not identified in the chi-square test. Second, the dimensionality of the solve-for parameter in each step and the computation cost are reduced by separating the error sources and unknown parameters as short- or long-term. For instance, the effect of drag-related parameters does not affect short propagation, but long propagation only, and long tracking data is required for estimating them.