Optical based statistical space objects tracking for catalogue maintenance

Cai, H 2019, Optical based statistical space objects tracking for catalogue maintenance, Doctor of Philosophy (PhD), Science, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Cai.pdf Thesis application/pdf 9.74MB
Title Optical based statistical space objects tracking for catalogue maintenance
Author(s) Cai, H
Year 2019
Abstract The number of space objects has grown substantially in the past decades due to new launches, regular mission activities, and breakup events. This has significantly affected the space environment and the development of the space industry. To ensure safe operation of space assets, Space Situational Awareness (SSA) has attracted considerable attention in recent years. One primary strategy in SSA is to establish and maintain a Space Object Catalogue (SOC) to provide timely updated data for SSA applications, e.g., conjunction analysis, collision avoidance manoeuvring. This thesis investigates three techniques for SOC maintenance, namely the tracklet association method for initial orbit determination, the multi-target tracking method for the refinement of orbital state estimation, and multi-sensor tasking method for the optimisation of sensor resources.

Generally speaking, due to the limited number of optical sensors used to track the large population of space objects, the obtained observational arcs for many targets are very short. Such short arcs, which contain a small number of angular observations, are referred as tracklets. Given such limited data, typical orbit determination methods, e.g., Laplace, Gaussian, Double-R methods, may fail to produce a valid orbital solution. By contrast, tracklet association methods compare and correlate multiple tracklets across time, and following successful association, a reliable initial orbital state can be further determined for SOC maintenance. This thesis proposes an improved initial value problem optimisation method for accurate and efficient tracklet association, and a common ellipse method to distinguish false associations of tracklets from objects in the same constellation. The proposed methods are validated using real optical data collected from the Mount Stromlo Observatory, Canberra, Australia.

Furthermore, another challenging task in SSA is to track multiple objects for the maintenance of a catalog. The Bayesian multi-target tracking filter addresses this issue by associating measurements to initially known or newly detected targets and simultaneously estimating the timevarying number of targets and their orbital states. In order to achieve efficient tracking of the new space objects, a novel birth model using the Boundary Value Problem (BVP) approach is proposed. The proposed BVP birth model is implemented in the Labelled Multi-Bernoulli (LMB) filter, which is an efficient multi-target tracker developed based on the Random Finite Set (RFS) theory, for improved computational efficiency of new space object tracking. Simulation results indicate that the computational efficiency of the proposed method significantly outperforms the state-of-the-art methods.

Finally, as limited sensors are available for SOC maintenance, an appropriate sensor tasking scheme is essential for the optimisation of sensor resources. The optimal sensor tasking command allocates multiple sensors to take the best action and produce useful measurements for more accurate orbital state estimation. In this thesis, an analytical form is derived for the Rényi divergence of LMB RFS in which each target state density is a single Gaussian component. The obtained analytical Rényi divergence is formulated as a reward function for multi-sensor tasking, which improves the computational efficiency, especially for large-scale space object tracking. In addition, this thesis further investigates the benefits of using the analytical Rényi  divergence and various space-based and ground-based sensor networks for accurate tracking of objects in geosynchronous Earth orbit.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Science
Subjects Aerospace Engineering not elsewhere classified
Keyword(s) space object catalog
tracklet association
multi-target tracking
multi-sensor tasking
birth model
Version Filter Type
Access Statistics: 93 Abstract Views, 137 File Downloads  -  Detailed Statistics
Created: Wed, 14 Aug 2019, 09:01:00 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us