Thomas Kelecy, The Stratagem Group; Shawn Abernethy, The Stratagem Group; Frank Jones, The Stratagem Group; Emily Gerber, The Stratagem Group; Hannah Wurzel, The Stratagem Group
Keywords: Rendezvous & Proximity Operations (RPO); Closely Spaced Objects (CSO); Artificial Intelligence (AI); Machine Learning (ML); Course of Action (CoA)
Abstract:
Closely Spaced Object (CSO) relative motion behavior can be categorized into the following classes: (a) normal operational behavior between CSOs, (b) cooperative Rendezvous and Proximity Operations (RPO) behavior, (c) inadvertent RPO behavior due to a satellite anomaly, or (d) provocative RPO behavior (a threat). The goal of this work is to derive and demonstrate the application of a set of observable Indications and Warnings (I&W) metrics using classification techniques that infer the intent of RPO behavior from observed relative motion. This behavior driven intent is intended to be applied in near real-time from ground and/or space-based Electro-Optical (EO) imagery sources. Relative observations extracted from the imagery are simulated to develop Artificial Intelligence (AI) spatiotemporal models that classify the behavior with the primary goal being to demonstrate the ability to quickly discriminate between natural, cooperative and threatening RPO activities. The innovation of this work is the application of AI to spot the relative behavior that might indicate a threatening RPO activity using real-time image data without the need for detailed tracking analytics used in traditional orbit estimation and data association processes. Results show that accurate quantification of correct identification of the four categories of RPO activity behavior for either ground or space-based observations. The best results are obtained when a combined set of ground and space-based observations are used. The data partitioning strategy for training and testing is also of crucial importance. The primary benefit to space safety is near real-time detection and prediction of threats based on observed relative behavior of Resident Space Objects (RSOs) to support rapid execution of Courses of Action (CoA).
Date of Conference: September 27-20, 2022
Track: Conjunction/RPO