Ohio State researchers present AI that explains its lunar landing-site recommendations

Ohio State researchers have presented preliminary findings for an AI framework designed to assess lunar landing sites and explain the evidence behind its recommendations. Called LUMINA, the system is meant to help mission planners work through complex lunar data—not choose a destination on its own. The team now plans further testing with higher-quality data.

Gokce Inal, a doctoral candidate in civil, environmental and geodetic engineering, presented the findings in October at the American Astronomical Society’s 58th annual Division for Planetary Sciences meeting. Ohio State described the work on October 28, reporting early results in making complex lunar observations and producing scientifically meaningful commentary about candidate landing sites.

From describing terrain to working through a decision

LUMINA builds on the team’s earlier LLaVA-LE model. That system combines images and language to describe lunar terrain and answer scientific questions. The earlier research, published in a CVPR 2026 workshop paper, specializes in characterizing the moon’s surface and subsurface.

The added capability, according to Inal, is a framework that can plan and carry out analyses involving several steps. It is intended to evaluate potential landing locations and explain why the evidence supports its recommendations.

Inal designed LUMINA to help scientists identify hazards that future crews could face and find geologic features worth investigating. Ohio State says those capabilities could also help planners locate areas rich in water ice and other resources.

“Most AI tools are passive tools,” she said. “You ask a question and get an answer, but we’re moving toward an active scientific assistant that can work through a problem alongside the experts.”

Gokce Inal, speaking to Ohio State News

An additional set of eyes, not a mission authority

Inal describes LUMINA as a decision-support system working with mission planners, who would make the final judgments. More active analysis is not the same as authority over a mission.

Co-author Alper Yilmaz, an Ohio State engineering professor, frames the need around the growing volume and sophistication of planetary datasets. Searching them for places to land or investigate becomes increasingly difficult, he said. In a harsh environment, he sees value in another set of eyes that can question findings, verify them and help experts communicate.

The next test is usefulness to planners

The team plans to test LUMINA on new, high-quality lunar data and further evaluate how well it supports mission planning. Ohio State’s account describes the preliminary performance qualitatively, without numerical accuracy or time-saving results.

Inal points to NASA’s Artemis program as an opportunity for the research, saying new lunar data could improve systems like LUMINA. Yilmaz also wants the approach to become useful beyond the moon, including on Mars, asteroids and Earth.