Admissible region splitting for multi-fidelity orbit propagation




Chiu, Alexander

Journal Title

Journal ISSN

Volume Title



Admissible regions allow for active tasking of sensors for orbit determination of newly detected objects. These regions are defined using physics-based constraint hypotheses applied to detected objects. The regions can be represented probabilistically using Gaussian mixtures and propagated forward in time, which is computationally expensive. Multi-fidelity orbit uncertainty propagation can help speed up this process, but performance degrades when applied to large initial distributions. A new method for splitting admissible regions is proposed based on k-means clustering. A sensitivity analysis is performed using the partial derivatives of orbital elements with respect to spherical coordinates. The clustering algorithm leads to enhanced multi-fidelity performance by keeping the points closer together over time. The accuracy and runtime improvements are demonstrated using simulated optical and radar test cases.


LCSH Subject Headings