This article is more than 1 year old

Algorithm spots 104 asteroids in huge piles of data

Rocks stood out like a THOR thumb for code

Researchers at The Asteroid Institute have developed a way to locate previously unknown asteroids in astronomical data, and all it took was a massive amount of cloud computing power to do it.

Traditionally, asteroid spotters would have to build so-called tracklets of multiple night sky images taken in short succession that show a suspected minor planetoid's movement. If what's observed matches orbital calculations, congratulations: it's an asteroid. 

Asteroid Institute scientists are finding a way around that time sink with a novel algorithm called Tracklet-less Heliocentric Orbit Recovery, or THOR, that can comb through mountains of data, make orbital predictions, transform sky images, and match it to other data points to establish asteroid identity.

To prove the concept, THOR co-creator and Asteroid Institute graduate fellow Joachim Moeyens focused on a thirty-day window of data pulled from the NOIRLab Source Catalog, a 68-billion-point dataset of National Optical and Astronomy Observatory telescope images. 

In that data, Moeyens discovered 104 asteroids that were validated and added to the Minor Planet Center's asteroid registry, which the Asteroid Institute's parent organization the B612 Foundation said means "researchers can now begin systematic explorations of large datasets that were previously not usable for discovering asteroids."

In addition to the NOIRLab dataset, THOR is also able to test data from the Large Synoptic Survey Telescope (LSST) dataset, a massive collection of astronomical observations that generates 20TB of data a night. 

Dr Ed Lu, executive director at the Asteroid Institute, said that THOR's results prove its creators have developed "a new computationally-driven method of discovering asteroids," albeit one that required "enormous computational power," Lu said

THOR runs inside the Asteroid Institute's Asteroid Discovery Analysis and Mapping (ADAM) platform that can build 3D models of asteroid orbits and maps out the inner solar system. Combining the two, said Moeyens, will give researchers the ability to go beyond finding new items in historic data.

"With additional development, ADAM::THOR will be able to perform real-time asteroid discovery on observations as they come in from telescopes around the globe," Moeyens said.

While the approach may be new, using old astronomical data to make new discoveries isn't. In 2020, researchers from the University of Warwick in the UK designed an algorithm able to identify potential exoplanets in NASA's Kepler mission readings. 

These type of algorithms may have another stargazing use, too: there are concerns that Earth's growing network of artificial satellites might make space observation more difficult, and AI might be needed to separate the stars from the orbiting internet links. ®

More about


Send us news

Other stories you might like