Accession Number:

AD1158717

Title:

Applying Machine Learning Anomaly Detection Techniques to U.S. Navy Space System Operations

Descriptive Note:

[Technical Report, Memorandum Report]

Corporate Author:

NAVAL RESEARCH LAB WASHINGTON DC

Personal Author(s):

Report Date:

2021-01-31

Pagination or Media Count:

44

Abstract:

This report documents the first year of a Karles Fellowship research project investigating applications of machine learning to enhanced spacecraft operations. The first year of the fellowship was primarily comprised of research scope determination, literature review, data collection, and algorithm selection and development. In recent years the United States U.S. Department of Defense DoD has placed an increased emphasis on the development of autonomous capabilities, and this has been echoed in U.S. Navy research and development strategy. Machine learning technology represents a near-term opportunity to incrementally improve autonomous capabilities through the augmentation of existing technology. In the longer term, it is an investment opportunity into new technology which may drastically improve the capabilities of DoD systems. Practical approaches to the autonomy problem must focus on removing the most significant barriers to autonomy before more sophisticated technology becomes realistic. In the context of space system operations, health monitoring and fault management has been identified by both government and commercial entities as one of the largest inhibitors to space system autonomy. The increasing size and complexity of space systems as well as the rapid adoption of satellite constellations has quickly made it impractical for traditional ground-based human monitoring to be sustainable. This work primarily investigates the use of machine learning for automated anomaly detection in satellite telemetry. Anomaly detection is one of the foundational responsibilities of autonomous health monitoring because the detection of off-nominal state is typically the first step in the operational fault detection and remediation process.

Descriptors:

Subject Categories:

  • Optical Detection and Detectors
  • Manned Spacecraft

Distribution Statement:

[A, Approved For Public Release]