In April 2024, the Asteroid Institute, a program of the B612 Foundation, announced that an algorithm had identified 27,500 new asteroid candidates without any new observations of the sky. The discoveries came entirely from mining archival images in the NOIRLab Source Catalog, collected between October 2012 and October 2019, by reprocessing more than 1.7 billion individual point-source detections.
The work used an algorithm called THOR - Tracklet-less Heliocentric Orbit Recovery - developed by researchers at the University of Washington’s DiRAC Institute. THOR can link points of light across images taken on different nights into a single asteroid orbit, removing the traditional requirement that an object be seen several times in one night. It ran on ADAM, an open-source cloud-based astrodynamics platform built on Google Cloud, and the search consumed about 8.5 million virtual CPU hours. The approach is a preview of how computation will help mine the far larger sky surveys expected from the Vera C. Rubin Observatory.