Ohio State researchers unveil LUMINA AI to pick lunar landing sites

Researchers at The Ohio State University introduced LUMINA, an artificial intelligence model built to help mission planners choose safe lunar landing sites and design outposts, at the American Astronomical Society’s 58th annual meeting of its Division for Planetary Sciences in October 2026.

This framework evaluates lunar landing sites with attention to essential resources such as water ice as well as scientific value, bridging artificial intelligence and planetary science. LUMINA pulls together lunar surface data covering topography, terrain hazards, geology, surface properties, and comparisons between candidate sites instead of relying only on orbital imagery. With those varied datasets, the framework can identify landing zones across the Moon that offer the strongest effectiveness for science and human habitation.

Gokce Inal explained on Oct. 28, 2026, that the model works as an active decision-support system beside experts, not as a passive or fully semiautonomous machine. Earlier AI tools give simple answers to direct prompts, while this framework serves as an active scientific assistant that collaborates on complex problems.

Mission planners can use the model to pick locations rich in water ice and other critical resources. By performing complex observations and providing scientific commentary, the tool can point to regions key to survival and mission success.

The agency plans to return astronauts to the Moon in 2028 for the first crewed landing since Apollo 17 in 1972.

NASA seeks a permanent human presence on the Moon and a later journey to Mars through the Moon to Mars Architecture, its evolving strategic framework. The framework develops and tests technologies needed for humans to live on the Moon before sending them to Mars, spanning power and transportation work, habitat systems, robotics, and in-situ resource utilization. A viable outpost means managing interdependent pieces such as habitat connectivity, power generation, waste systems, and room for future modules.

The earlier model showed capabilities for describing lunar terrain and answering scientific questions, while LUMINA goes further by carrying out multistep analyses and explaining recommendations. The predecessor model, known as LLaVA-LE, was designed to analyze and describe orbital images of the Moon.

Gokce Inal stated on Oct. 31, 2026, that AI systems will become vital tools as human operators sort through the millions of images and datasets produced by coming space missions. The lead author said the work serves as a bridge between the separate fields of artificial intelligence and planetary science. She said the same day that NASA’s Artemis program gives the team valuable opportunities, since fresh lunar data can improve tools like LUMINA for planning later missions.

Alper Yilmaz stated on Oct. 28, 2026, that the group wants to broaden their map model so it works for other planetary bodies, including Mars and multiple asteroids. Few GPT-style map models exist for space exploration, so the team plans to extend the framework’s research and exploration work beyond the Moon.

Alper Yilmaz noted on Oct. 28, 2026, that planners will need tools like this model to handle fast-growing datasets, or picking the best landing or investigation sites will be extremely difficult. Yilmaz said that as datasets grow more complex, searches carried out by people alone will become impractical when choosing the best locations for crews.

Alper Yilmaz suggested on Oct. 28, 2026, that AI offers a necessary second set of eyes to check information and strengthen communication for crews touching down in harsh places. The reasoning is that planetary environments are extreme by nature, so they demand extra layers of verification and communication support.

In-situ resource utilization aims to extract local resources such as water ice and lunar regolith for crew survival, fuel generation, oxygen production, and potential construction to end Earth resupply needs.