A Receding Horizon Based Sensor Tasking for Tracking Space Objects in Cislunar Regime
Open Access
- Author:
- Dasari, Vishnu Tej
- Graduate Program:
- Aerospace Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- August 21, 2024
- Committee Members:
- Puneet Singla, Thesis Advisor/Co-Advisor
Roshan Thomas Eapen, Committee Member
Amy Pritchett, Program Head/Chair
Robert G. Melton, Committee Member - Keywords:
- space situational awareness
conjugate unscented transform
kalman filter
mutual information - Abstract:
- Cislunar space is a growing area of interest as several countries plan missions to explore the moon and its surrounding regions for scientific, economical and national security reasons. With this increased interest, there is an apt need to accurately track spacecraft in the cislunar regime. This thesis develops an optimal sensor tasking approach to track space objects in the cislunar regime with the help of space-based sensors. The developed approach utilizes the non-product quadrature method known as the Conjugate Unscented Transform (CUT) to accurately and efficiently propagate the state uncertainty through the chaotic Circular Restricted Three Body Problem (CR3BP). The CUT approach further leads to the design of a higher order Kalman filter to fuse the model predictions with sensor observations. An information theoretic metric, Mutual Information (MI), also known as the Kullback-Leibler divergence metric, is used to assess the quality of a sensor observation. The MI provides the ratio between the prior and posterior probability density functions, which gives a quantitative assessment of the reduction in uncertainty due to sensing decisions. An exhaustive search algorithm is used to maximize mutual information over all objects, sensors, and time steps for a given problem. Given the combinatorial growth associated with the exhaustive search, sub-optimal sequential algorithms are designed while exploiting the submodularity property of the MI metric. Different variants of the sequential algorithms (e.g. receding horizon window or sequential-in-time) are discussed depending on the size of the problem. A receding horizon approach solves the sensor tasking problem for a given subset of time steps before moving on to the next subset. This is done until the entire time span has been solved, meaning the sensors are tasked for each time step of the time span. Numerical results are shown for the receding horizon window applied to two examples of an Earth-based space object tracking problem and a cislunar object tracking problem. The results show that the implementation of the receding horizon window approach was a success as the tasking provides the desired reduction in uncertainty of all objects over the time span. Specifically, the state errors are statistically analyzed and shown to not significantly impact the object tracking over time.