Home / Research Areas / Data Science

Data Science

USRA NAMS Data Science groups are using machine learning to enable computers and other automated systems to perform tasks that have historically required human cognition and human decision-making abilities.

Image courtesy of NASA Aeronautics

Overview

Active Projects

All

Overview

USRA Aviation Data Science group works with NASA and the Federal Aviation Administration (FAA) to enable new advances in traditional aviation with human pilots, to detect mission-critical anomalies in autonomous search and rescue drones, and to make flight more efficient and accessible.

USRA Environmental Data Science groups work with NASA, U.S. Geological Survey, and research partners to conduct research on air quality and health impact models, atmospheric sciences and aerosol sciences in Earth and Space environments, and to develop new approaches for studying various hazards and natural disasters.
Active Projects

USRA NAMS Aviation Data Science group conducts research in anomaly detection in aviation with interpretable deep learning model.

Graphical illustration of the developed interpretable semi-supervised model to identify anomalies during the approach to landing of commercial aircrafts.

Image courtesy of USRA NAMS

The identification of precursors to safety incidents in aviation data is a crucial task, yet extremely challenging. The main approach in practice, leverages domain expertise to define expected tolerances in system behavior and flags exceedances from such safety margins. However, this approach is incapable of identifying unknown risks and vulnerabilities. Improving the ability to identify emerging vulnerabilities in current operations helps to increase awareness of new threats. Proactively addressing safety requires developing, testing, and validating new approaches that can process and model large amounts of historically recorded heterogeneous data. Such data describe the operations of millions of flights over multiple years and covers various diverse regions in the National Airspace System. Data science and machine learning approaches have the potential to automatically identify anomalous events in these observed data. However, the events identified still need to be reviewed and assessed by subject matter experts familiar with the procedures in order to better understand how operations are carried out, and their safety implications. Possible vulnerabilities can then be addressed by mitigating the contributing factors with proper countermeasures. These may include improved pilot/controller training or developing automation safety processes, which, when in place, help to avoid states that result in an increased likelihood of an incident or accident that may result in damage to the aircraft, injury, or loss of life.