C. Channing Chow II, Cloudstone Innovations LLC; Jason Baldwin, Complex Futures LLC; Charles J. Wetterer, KBR; Micah Dilley, KBR; Keric Hill, KBR; Paul Billings, KBR; Christopher Craft, KBR; James Frith, Air Force Research Laboratory
Keywords: cislunar, orbit determination, astrodynamics, data fusion, space situational/domain awareness
Abstract:
The high levels of nonlinearity evident in the cislunar domain present stressing challenges when estimating the trajectories for and maintaining custody of cislunar objects. In our previous paper, we demonstrated the use of the Unscented Kalman Filter (UKF) and Gaussian Mixture Models (GMM) in the orbit determination (OD) and tracking of objects in cislunar space using astrometric and photometric observations of non-Keplerian orbits that arise due to a more pronounced influence of the Moons gravity field. It was also shown that the use of high-fidelity 3-body dynamics is essential in capturing accurate representations of the motion of objects in cislunar space. The UKF provided relatively good results for a variety of orbits and data cadences, while the GMM demonstrated promise in improving performance through large data gaps. In this paper, we expand our cislunar OD research in two areas: (1) we apply the Adaptive Entropy-based Gaussian-mixture Information Synthesis (AEGIS) approach to improve the uncertainty realism within the estimation filter, and (2) we compare filter performance using different observation types independently, namely electro-optical (EO) and passive radio frequency (PRF), and then show improvements with their data fusion. In the first OD area, performance using AEGIS is compared directly to use cases from our previous paper in which a single UKF and fixed GMMs were used. The orbits under consideration, here, are examples from three separate elemental periodic orbit families that have been differentially corrected in a high-fidelity dynamical system: H1, L1, and W4W5. The individual orbits sample a range of possibilities within each family, including orbits near theoretical bifurcation points between families (i.e., where a single orbit belongs to both families). In the second OD area, filter performance improvement is achieved primarily by reducing or eliminating large data gaps in the EO data by filling in with PRF data. The efficacy of the strategy is justified since PRF observations are still possible during times when EO observations are excluded due to EO constraints. This result is derived from processing synthetic observations from the four orbits from our previous paper that were the most challenging due to large data gaps. We model and simulate hypothetical, Earth ground-based, EO and PRF sensor networks to generate the expected right ascension (RA), declination (DEC), and visual magnitude (VMAG) measurements, and the time difference of arrival (TDOA) and frequency difference of arrival (FDOA) measurements, respectively. All filters are constructed from components using the modular Infinity Filter Framework (IFF).
Date of Conference: September 27-20, 2022
Track: Cislunar SSA