Michael Lachut, EOS Space Systems and Space Environment Research Centre; James Bennett, EOS Space Systems and Space Environment Research Centre; David Kooymans, EOS Space Systems
Keywords: Orbit Estimation, Orbit Prediction, Regressioon Modelling
Abstract:
One of the primary aims of the Space Environment Research Centre (SERC) is to build and maintain its own catalogue of objects in near-Earth orbits. This will provide accurate object ephemeris data for object acquisition in both passive and active tracking, covariance matrices for sensor schedule tasking and observation correlation, state and uncertainty propagation for conjunction assessments, and to facilitate the remote manoeuvre of debris using high powered lasers delivered from ground stations, which is the primary objective of SERC. The catalogue comprises objects which pose a threat of colliding with Optus satellites identified in conjunction assessments, potential candidates for the laser manoeuvre, and other objects of interest for ongoing research including Envisat, Topex, small cubesats, and HAMR debris.
In this paper, we will present methods to automate the assessment of orbit predictions and estimations to maintain and build a new catalogue of space assets and debris. The process begins by running an orbit determination seeded by parameters stored in a catalogue and/or publically available. The characteristics of an object of interest determined during the orbit estimation, such as the ballistic coefficient or radiation coefficient, are assessed as to whether they are within physical limits and are also compared against historic estimated values to test validity. In the event of a new object where no historic data is available, non-negative ballistic/radiation coefficients are stored and not used until a sufficient number is available to achieve statistical confidence in the stored values.
The observation residuals generated from the orbit determination (OD) process are assessed as to whether they are within a given threshold based on the nominal noise of each sensor within our network. Comparisons of observations with highly accurate external ephemeris data from the ILRS are presented to show the level of sensor noise. If the RMS of an observation arc is above a threshold, which is set based on the observation weighting factor, the pass is automatically weighted weaker thus increasing the threshold for that pass. At this point, the OD process runs iteratively until all observation arcs stay within their respective thresholds. This ensures that outliers are tagged and kept quarantined in the catalogue.
The next step is to assess the consistency of the generated ephemeris data from the current OD process with that from a previous OD process, i.e., the resultant difference in states from both ephemeris data is projected onto a satellite coordinate system to access their difference in the radial, along-track and cross-track directions (RSW). The so-called Hills equations are used to predict the functional form of the projected difference in the RSW coordinate system and to set the basis functions for non-linear regression fitting. The coefficients from the regression models are then compared to predefined thresholds from historic regression models stored in the catalogue. If any of the current coefficients from the regression fitting of the current process exceed those predefined thresholds, the ephemeris data is quarantined and assessed manually for prediction accuracy. Examples are provided where ephemeris data from a current OD process is flagged as significantly different to that which was generated in a prior OD process.
Date of Conference: September 11-14, 2018
Track: Poster