Steven Ingram, Lockheed Martin; Rupinder Paul, Virginia Tech; Andrew Zizzi, Lockheed Martin Advanced Technology Center; Brian Mayer, Virginia Tech; Naren Ramakrishnan, Virginia Tech; Moses Chan, Lockheed Martin
Keywords:
Abstract:
As the number of objects sent into space and into orbit around the Earth is likely to continue increasing, Space Situational Awareness will continue to require up-to-date methods and algorithms that can predict, and eventually mitigate issues caused by satellite anomalies. A cognitive fusion modeling approach, which entails combining informative features from different data modalities, presents a unique approach to orbital anomaly detection tasks. In this paper, we investigate two methods that can potentially make our fusion approach more feature rich and provide higher confidence when estimating if a satellite is “of interest.”
First, we investigate the use of unsupervised outlier detection algorithms using a set of delta features which represents the change in the value of satellites orbital elements. For this task, we make use of “structured” sensor data that comprises of Two-Line element (TLE) sets which represent satellites orbital path. In the second method, we make use of unstructured infrared images from the Space Object Tracking (SpOT) facility of satellites in orbit to detect patterns of anomalous behavior. With help of a very comprehensive set of known satellite anomalies as noted by online sources, we will examine and evaluate several scenarios based on historical events that document the utility of these additional techniques for assisting with cognitive fusion modeling.
Date of Conference: September 11-14, 2018
Track: Poster