AI Approach Uncovers Dozens of Hidden Planets in NASA's TESS Data
Astronomers at the University of Warwick have validated over 100 exoplanets, including 31 newly detected planets, using a new artificial intelligence tool applied to data from NASA's Transiting Exoplanet Survey Satellite (TESS), a space mission that monitors the sky for the subtle dimming of starlight caused when planets pass in front of their host stars.
Published in MNRAS, the team applied their newly developed AI pipeline called RAVEN to observations of over 2.2 million stars collected during TESS's first four years of operations. They focused on finding planets that orbit close to their stars, completing an orbit in less than 16 days, providing the most accurate assessment of how common these short-period worlds are.
"Using our newly developed RAVEN pipeline, we were able to validate 118 new planets, and over 2,000 high-quality planet candidates, nearly 1,000 of them entirely new," said first author Dr. Marina Lafarga Magro, Postdoctoral Researcher at the University of Warwick.
RAVEN's Edge
Modern planet-hunting missions routinely identify thousands of possible planets (candidates), but confirming which signals are real, and understanding how common different types of planets are, remains a major challenge with the current methods.
"The challenge lies in identifying if the dimming is indeed caused by a planet in orbit around the star or by something else, like eclipsing binary stars, which is what RAVEN tries to answer," said Warwick's Dr. Andreas Hadjigeorghiou, who led the development of the pipeline.
"Its strength stems from our carefully created dataset of hundreds of thousands of realistically simulated planets and other astrophysical events that can masquerade as planets. We trained machine learning models to identify patterns in the data that can tell us the type of event we have detected, something that AI models excel at."
"In addition, RAVEN is designed to handle the whole process in one go, from detecting the signal, to vetting it with machine learning and statistically validating it. This gives the pipeline an additional edge over contemporary tools that only focus on specific parts of the workflow."
Planetary Prevalences
With this well-characterized set of validated planets, the team was able to move beyond individual discoveries and study the population of close-in exoplanets in detail. In a companion MNRAS study, they measured how frequently close orbiting planets occur around sun-like stars; mapping results across orbital period and planet size with unprecedented detail.
They found that around 9–10% of sun-like stars host a close-in planet, which was consistent with NASA's Kepler mission—a space telescope that previously measured how common planets are around other stars, but RAVEN had uncertainties up to ten times smaller.
The study also provides the first direct measurement of "Neptunian desert" planets, showing they occur around just 0.08% of sun-like stars.
"For the first time, we can put a precise number on just how empty this 'desert' is," said Dr. Kaiming Cui, Postdoctoral Researcher at Warwick and first author of the population study. "These measurements show that TESS can now match, and in some cases surpass, Kepler for studying planetary populations."
A Foundation for Future Discoveries
Together, these studies demonstrate how large astronomical data and new AI developments go hand in hand, generating new discoveries while stress-testing AI on difficult research problems as well as transforming both planet discovery and planetary population science.
The team has also released interactive tools and catalogs, allowing other researchers to explore the results and identify promising targets for future observations with ground-based telescopes and upcoming missions such as ESA's PLATO.
Related Articles
- Planetary systems discovered by TESS
- Exoplanet discoveries using new methods
- Astronomy technology advancements
© 2026 Phys.org
