Gaussian-Binary classification for resident space object maneuver detection

Abstract

Abstract The capability accurately and timely detect a resident space object (RSO) maneuver is a critical task for monitoring space activities. This paper presents a data-driven Gaussian Binary RSO Maneuver Detection (GaBRSOMD) method to detect whether or not there is a maneuver between two tracks of the same RSO on different orbital paths. Using an in-house simulated space catalog environment, the Gaussian Binary Classification (GBC) method is used to detect three types of maneuvers: the maneuver is due to a small impulsive velocity change; a low thrust, and the maneuver is due to either an impulsive or a low thrust which is unknown a priori. Numerical results demonstrate that the proposed GBC model can achieve high accuracy in detecting all three cases of maneuvers. The paper further demonstrates that the GBC approach is robust to noisy data and has advantages over an AutoEncoder method and the classical Principal Component Analysis (PCA) method. Hence, the proposed GBC has great potential for making quick detection decision with both high accuracy and precision.