NASA deploys ExoMiner++ AI to accelerate search for nearby Earth-like exoplanets

NASA is increasingly turning to artificial intelligence to speed up the search for exoplanets, using a deep-learning tool called ExoMiner++ to sift through vast amounts of data from its Transiting Exoplanet Survey Satellite, or TESS.

Launched in 2018, TESS surveys nearly the entire sky to detect planets orbiting bright, nearby stars.

Scientists have so far confirmed more than 6,000 exoplanets, largely using data from NASA’s Kepler Space Telescope and TESS, with the latter accounting for over 700 discoveries. The sheer volume of data has made automated analysis essential.

According to NASA, an earlier version of the tool, ExoMiner, confirmed 370 planets using Kepler data. ExoMiner++, an upgraded version designed to analyse both TESS and Kepler datasets, has identified about 7,000 planet candidates within TESS observations. The system uses deep-learning neural networks to analyse light curves, detecting subtle dips in starlight that may indicate a planet passing in front of its host star.

The AI distinguishes genuine planetary transits from false signals caused by phenomena such as eclipsing binary stars.

In one analysis, researchers trained the algorithm on around 10,800 TESS light curves, enabling it to identify three planet candidates that had previously been overlooked.

NASA noted that TESS’s wide sky coverage contrasts with Kepler’s narrower field of view, allowing the mission to monitor far more stars. All TESS data and AI tools, including ExoMiner++, are publicly available, with the software released as open source on GitHub.

Looking ahead, NASA’s Nancy Grace Roman Space Telescope, expected to launch in the mid-2020s, is projected to detect tens of thousands of additional exoplanet transits.

Scientists say AI-driven tools will be crucial in analysing this data and accelerating future discoveries.