IBM and NASA Unveil AI-Driven Lunar Map to Aid Future Exploration

NASA and IBM have joined forces to create an open-source Lunar Foundation Model that combines decades of lunar observations into one AI map.

The model uses a multimodal approach to analyze various types of lunar data, including thermal, topographic, and multispectral information. This allows researchers to study the Moon's surface in greater detail without treating each dataset in isolation.

Testing showed improvements over the SwinV2 baseline across several tasks, with a 22% reduction in error on ice-prospectivity tasks and a 19% improvement in accuracy for crater detection at 100-metre resolution. The model also delivered a 3% improvement in intersection over union (IoU) for mapping irregular mare patches.

The Lunar Foundation Model is designed to help researchers identify areas with greater potential for ice, which could be used as a resource for future missions. It can also aid in understanding the Moon's geological history and selecting landing areas.