Jeffrey Hollon, Applied Optimization, Inc.; Kimberly Kinateder, Applied Optimization; Victoria Carone, Applied Optimization, Inc.; Tamara Payne, Applied Optimization, Inc.; Markus Ernst, Applied Optimization, Inc.; Phan Dao, Applied Optimization Inc.; Charles J. Wetterer, KBR; Stephen Gregory, Stephen A. Gregory, LLC; Anthony Dentamaro, Anthony Dentamaro, LLC; James Frith, AFRL/RVSW; Scott Milster, AFRL/RVSW
Keywords: cislunar, change detection, data fusion, identification, manifold learning
Abstract:
A critical function of successful Space Domain Awareness (SDA) is to provide robust and confident Space Object Identification (SOI) and Change Detection (CD) capabilities to analysts. We have developed a collection of algorithms that provide these capabilities for space objects in the Low-Earth Orbit (LEO) and Geosynchronous Earth Orbit (GEO) regimes when surveyed by ground-based and space-based sensor platforms. The proliferation of space objects outside of GEO demands that existing and new technologies address cislunar SDA. Due to differences between the LEO, GEO, and cislunar space domains, existing technologies cannot be applied directly and must evolve to operate in this new and highly dynamic environment. In this paper, we describe the operation of our current CD methods, the problems they face in generalizing from LEO and GEO to cislunar space, and our initial efforts and assessments to extend the CD methods to cislunar space.
The photometry-based CD methods we have developed can be differentiated by how much historical data they require. The baseline method assumes previous photometry data collected on the satellite are available for processing, while the baseline-less method requires only one collection of photometry data. These two methods are complementary in the sense that if the requirements for the baseline method are not met, changes can be assessed using the baseline-less method.
The baseline method ingests historical data on an object that can be from multiple sensors and embeds the astrometric and photometric information in a high-dimensional manifold that statistically learns the space objects brightness characteristics as illuminated and viewed from any direction and with a specific attitude profile (e.g., Earth-nadir pointing). Thus, our approach uses a fusion of astrometric and photometric measurements. The manifold is referred to as the space objects historical hypersurface, and it may be queried based on the orbital state vectors of new observations. The queried historical observations form a statistical baseline without the need to observe the space object under the same observing conditions. These extracted baseline light curves are statistically modelled to provide an expected light curve against which the new data will be evaluated in order to invoke our CD methods.
The baseline-less method comprises a set of three algorithms designed to detect near real-time changes. They differ in that they utilize local trends in the photometric data from one collection, in order to develop an understanding of the space objects behavior. When the local trends are broken, based on statistically driven thresholds, the baseline-less method will flag that a change has occurred.
To demonstrate and test our two CD approaches on objects in cislunar space, we generated a set of notional ground- and space-based photometric observations of targets in selected cislunar orbits. Using the Satellite Visualization and Signature Tool (SVST), we then simulated their photometric light curves. Changes in the targets pose are imposed in the simulations to mimic real-world attitude control changes that cislunar space objects are expected to perform. The baseline and baseline-less methods were tested on this dataset, and their ability to correctly detect these changes in space object behavior was quantified to demonstrate their level of success. Further, we demonstrated that the use of our baseline method provides the ability to perform CD along with SOI.
Date of Conference: September 27-20, 2022
Track: Non-Resolved Object Characterization