Using machine learning to analyze extensive Transiting Exoplanet Survey Satellite (TESS) datasets, researchers have created one of the most accurate catalogs of nearby exoplanets to date.

Astronomers at the University of Warwick have confirmed over 100 exoplanets—including 31 newly discovered worlds—by applying an advanced AI system to TESS observations. The mission monitors stars for subtle dips in brightness, which occur when planets transit, or pass in front of, their host stars.

Figure 1. “Hot Jupiter” Exoplanet Discovered by Hubble Space Telescope Orbiting Extremely Close to Its Star

The findings, published in Monthly Notices of the Royal Astronomical Society (MNRAS), are based on a newly developed AI pipeline called RAVEN AI pipeline. Researchers used it to analyze data from more than 2.2 million stars collected during the first four years of the Transiting Exoplanet Survey Satellite (TESS). Their study focused on planets with extremely short orbital periods—less than 16 days—to better understand how common these close-in worlds are.

According to Marina Lafarga Magro, a postdoctoral researcher at the University of Warwick, the RAVEN system enabled the validation of 118 new planets along with more than 2,000 high-quality candidates, nearly half of which are entirely new discoveries. This dataset represents one of the most detailed samples of close-orbiting planets and will guide future observational studies.

Among the confirmed discoveries are several notable categories: ultra-short-period planets that orbit their stars in under 24 hours, rare “Neptunian desert” planets found in regions where few are expected, and compact multi-planet systems featuring tightly packed planetary pairs around the same star.

RAVEN’s Advantage in Exoplanet Discovery

Modern surveys can flag thousands of potential planets, but confirming which signals are genuine remains a major challenge. Many false positives arise from phenomena like eclipsing binary stars, which can mimic planetary transits.

As explained by Andreas Hadjigeorghiou of the University of Warwick, the difficulty lies in determining whether a dip in starlight is caused by an orbiting planet or another astrophysical event. The strength of RAVEN AI pipeline comes from its training on a vast dataset of realistically simulated planets and lookalike signals. By learning to recognize subtle patterns in the data, the AI can accurately distinguish true planetary events from imposters—an area where machine learning excels.

RAVEN AI pipeline stands out for its ability to manage the entire exoplanet detection process in a single workflow—from identifying signals to vetting them with machine learning and statistically validating the results. This integrated approach gives it a clear advantage over tools that handle only specific stages of analysis.

According to David Armstrong, Associate Professor at the University of Warwick and senior co-author of the study, RAVEN enables consistent and objective analysis of massive datasets. Its rigorous testing and validation mean the results go beyond a simple list of candidates, providing a reliable sample for studying how different types of planets are distributed around Sun-like stars.

Mapping Planetary Prevalence Across Star Systems

Using this large, well-defined dataset, researchers were able to move beyond individual discoveries and analyze broader trends in planetary systems. In a companion study published in Monthly Notices of the Royal Astronomical Society (MNRAS), they examined how frequently close-orbiting planets occur around Sun-like stars, mapping their distribution by orbital period and size with unprecedented precision.

The team found that roughly 9–10% of Sun-like stars host a close-in planet—consistent with earlier findings from Kepler mission—but with uncertainties reduced by up to tenfold thanks to RAVEN AI pipeline.

The study also delivers the first direct measurement of the so-called “Neptunian desert,” revealing that such planets occur around just 0.08% of Sun-like stars. As noted by Kaiming Cui of the University of Warwick, these results demonstrate that Transiting Exoplanet Survey Satellite (TESS) can now match—and in some cases exceed—the capabilities of Kepler in studying planetary populations.

A Launchpad for Future Discoveries

Together, these studies demonstrate how combining large-scale astronomical data with advanced AI can accelerate discovery [1]. The approach not only uncovers new planets but also boosts confidence in the results while offering deeper insights into planetary systems.

The research team has also released interactive tools and catalogs, allowing scientists worldwide to explore the data and identify promising targets for follow-up observations using ground-based telescopes and future missions like PLATO mission.

Reference:

1. https://scitechdaily.com/ai-uncovers-hidden-signals-discovering-dozens-of-new-alien-planets/

Cite this article:

Janani R (2026), AI Detects Hidden Patterns, Revealing Dozens of New Alien Planets, AnaTechMaz, pp.816