Improving low-Earth orbit predictions using two-line element data with bias correction J. C. Bennett * The Satellite Positioning for Atmosphere, Climate and Environment (SPACE) Research Centre, School of Mathematical and Geospatial Sciences, RMIT University, GPO Box 2476, Melbourne, Victoria, 3001, Australia EOS Space Systems Pty. Ltd., Mount Stromlo Observatory, Cotter Road, Weston Creek, Australian Capital Territory, 2611, Australia J. Sang, C. Smith EOS Space Systems Pty. Ltd., Mount Stromlo Observatory, Cotter Road, Weston Creek, Australian Capital Territory, 2611, Australia K. Zhang The Satellite Positioning for Atmosphere, Climate and Environment (SPACE) Research Centre, School of Mathematical and Geospatial Sciences, RMIT University, GPO Box 2476, Melbourne, Victoria, 3001, Australia ABSTRACT In this paper we present results from our orbit prediction study using the publicly available Two-Line Element (TLE) sets. The method presented here is similar to that introduced by Levit and Marshall [1]; however, we also consider the non-spherical low-Earth orbit satellites Grace A and Grace B. A state vector is generated every 10 minutes in the orbit determination (OD) period using SGP4. These generated states are subsequently used as observations in an orbit determination run considering a full set of forces to determine the orbit over the 10-day time span. All information used is from the TLE data sets. Once the orbit has been determined, it is then numerically propagated to obtain a prediction of the object’s position. The TLE-determined orbit is compared to highly accurate satellite laser ranging (SLR) Consolidated Prediction Format (CPF) data to assess the accuracy. We tested the technique by performing 200 independent simulations for Stella, Starlette, Grace A and Grace B and found that it resulted in better orbit pre- dictions 98.5%, 93.4%, 97.5% and 95.5% of the time, respectively, when compared to standard SGP4 propagation. For Starlette and Stella after a 7 day prediction period the average absolute maximum along track bias was reduced by approximately 64% and 74%, respectively. For Grace A and Grace B after a 7 day prediction period the average absolute maximum along track bias was reduced by approximately 68% and 64%, respectively. The TLE-determined orbit contains bias in the along, across track and radial directions with the along track error dominating. If these can be estimated we can obtain an improved orbit prediction. We used our TLE-determined orbit as an initial state and determined an orbit 3 days after the 10 day OD period from only two passes of SLR data from a single station (Mount Stromlo, Australia). We then estimated the bias in the along track direction by fitting a quadratic function to the along track bias data. The error between the TLE-determined orbit and the SLR-determined orbit in the along (minus the quadratic bias), across and radial tracks was then estimated using sinusoidal functions. These estimations were then used to correct the TLE-determined orbit, resulting in drastic improvements in the prediction accuracy of low-Earth objects. For a prediction period of 7 days, the absolute maximum along track error for Grace A reduced from 16.6 km (SGP4) to 4.8 km with the TLE data fitting presented in this paper. With bias estimation this error was reduced to 1.7 km. This demonstrates the ability to obtain much more accurate orbit predictions using only two passes (19 normal point SLR ranging observations) from one station. In the operational sense, the presented method can be used in debris conjunction analyses to improve the accuracy and reliability of the conjunction predictions. This method is currently implemented in EOSSS’ conjunction analysis software. Objects of interest can then be tracked with EOSSS’ tracking facilities and much better orbit predictions can be obtained. Keywords: Orbit Determination, Orbit Prediction, Bias Correction, Low-Earth Satellites * Email: james.cameron.bennett@rmit.edu.au