Precise orbit determination with satellite laser ranging observations using a batch filter based on particle filtering

다국어 초록 (Multilingual Abstract)

This study presents a batch filter based on particle filtering (PF) for precise orbit determination (POD) with satellite laser ranging (SLR) observations. As a verification of the performance of the presented batch filter based on PF (PF Batch), one-dimensional numerical simulations and accuracy assessments are initially conducted under various initial errors and non-Gaussian noise environments. Under conditions of large initial state error or large non-Gaussian measurement noise, a PF Batch yields more robust and accurate estimation results than a batch filter based on unscented transformation (UT). In addition, a sensitivity analysis of estimation parameters is performed, showing that the developed PF Batch does not require the heavy scaling parameter tuning required for a batch filter based on UT (UT Batch). A complexity analysis is also performed to consider the computational burden.

The significance of this research is to propose a PF Batch for POD and to examine the characteristics of the POD process and results by using the PF Batch.
For applications of POD, the Yonsei laser-ranging precise orbit determination system (YLPODS) is developed for SLR observations of low Earth orbiting satellites. SLR normal point (NP) observations of the CHAMP satellite are used for measurements of POD.
The characteristics of a PF Batch are compared with those of a UT Batch and those of a batch least-squares filter (BLSF) under various environmental conditions. For large initial error conditions, the initial position errors of a reference orbit are assumed to be 10 m, 100 m, and 500 m. For non-Gaussian conditions, five simulated measurement noises are added to SLR observations. A post-fit residual test and an external orbit comparison are performed for orbit quality assessment purposes. The results show that the performance of the PF Batch is better than that of the other two filters in large initial error and non-Gaussian environmental conditions. For the post-fit residuals and the 3D errors by means of external orbit comparison under large initial errors, while the precision of the UT Batch and the PF Batch is maintained at a level of about 10 cm, the precision of the BLSF is not. For the post-fit residuals and the 3D errors under non-Gaussian environmental conditions, the precision of the PF Batch and the BLSF is maintained at the level of under 70 cm regardless of the arc, but the precision of the UT Batch is not.
Moreover, the PF Batch does not require the heavy scaling parameter tuning that is required for a UT Batch. These results show that POD using the PF Batch is advantageous in terms of accuracy and convenience.