Automatic Hazard Detection and Avoidance


From debris to congestion in low-earth-orbit (LEO) to maneuvers and anti-satellite weapons, there are a lot of threats posed to space operations.

Increased use of on-board sensing is improving hazard awareness, but with an increase of false detections. Confidence in the information is key to enabling avoidance actions.

We’re interested in applying machine learning to threat detection in space to mitigate false detection and improve space operations by making autonomous detection more trustworthy.


NASA, the Department of Defense (DoD), the US intelligence community, and international partners have historically provided space situational awareness as a combined effort. To date, approaches for satellite and debris conjunctions have been effective, but feature a high rate of detections (on the order of 2000 a day) that must be reviewed by humans. The problem is further complicated by an overwhelming number of objects, more than 100 objects that are regularly changing orbits, and high uncertainty in the locations of resident space objects. While 23,000 objects larger than 10 cm in diameter are regularly tracked by the U.S. government, it is estimated that there are more than 500,000 objects between 1 and 10 cm as well as more than 100 million objects smaller than 1 cm are predicted to exist. Furthermore, space surveillance can provide an accurate update on the positions and velocities of resident space objects at intervals from once an orbit to once a month, with a significant amount of uncertainty between observations. 


Ready to learn more?

Sign up now to be invited to the webinar series where we’ll discuss the context behind each problem statement and answer questions from startup and university teams.