NASA and IBM Release Open-Source AI Model to Map Moon's Surface for Artemis Missions
12 Sources
[1]
NASA and IBM Launch Open-Source AI Model for Future Moon Explorations - CNET
Anna Gragert (she/her/hers) was previously the lifestyle editor at HelloGiggles, the deputy editor at So Yummy and the senior lifestyle editor at Hunker.... Read full bio The NASA‑IBM Lunar Foundation Model is one of the first of its kind: a publicly available AI model created to support
[2]
NASA and IBM open source lunar mapping tools
IBM and NASA have got together again and released an open source AI model of the Moon that could be used to make new discoveries about Earth's natural satellite. The NASA‑IBM Lunar Foundation Model has been trained on an extensive lunar observation dataset curated by researchers at the two
[3]
IBM, NASA launch AI model to help map ice, craters on Moon
Sept 10 (Reuters) - IBM (IBM.N), opens new tab and NASA on Thursday released an open-source AI model designed to help scientists analyze decades of lunar observation data and support plans for a sustained human presence on the Moon. The NASA-IBM Lunar Foundation Model is a publicly available AI
[4]
Moon's surface will be mapped in incredible detail using first-of-its-kind AI, NASA and IBM reveal
IBM and NASA have partnered to create a new artificial intelligence (AI) model capable of processing decades of lunar data so scientists can more accurately map the moon's surface one day. Decades of robotic lunar missions have left scientists with a massive, disjointed trove of data.
[5]
NASA and IBM made an AI model for exploring the Moon - Engadget
This past spring, the world experienced a rare moment of collective joy and awe when NASA's Artemis II mission, the first crewed flight to the Moon since 1972, completed its historic lunar flyby. On April 6, astronauts Reid Wiseman, Christina Koch, Victor Glover and Jeremy Hansen flew farther from
[6]
IBM and NASA release an open-source lunar foundation model
The release includes a unified lunar dataset built from nine instruments across four missions, and cuts error in identifying potential ice deposits by 23% against a general-purpose vision model IBM and NASA are releasing an AI model trained on decades of observations of the Moon, and making it
[7]
IBM is launching a new open source AI model to get NASA back to the Moon -- and making petabytes of lunar data available to study
IBM will help NASA researchers analyze decades of lunar observation data * IBM and NASA launch open source AI model to help further lunar research * Researchers will be able to better analyze petabytes of Moon data from last five decades * IBM and NASA also release dataset for public usage and
[8]
NASA wants humans back on the Moon in 2028. IBM's new AI could help get them there
The Lunar Foundation Model will analyze decades of Moon data to help researchers find hazards, resources, and changes to the lunar surface. IBM is going to the Moon. Sort of. On Thursday, IBM announced the open-source release of the NASA-IBM Lunar Foundation Model, which may help scientists tap
[9]
IBM, NASA launch AI model to help map ice, craters on Moon
The NASA-IBM Lunar Foundation Model is a publicly available AI tool designed to study the Moon. It was trained on more than 30 layers of data collected by nine instruments on four NASA missions, including the Lunar Reconnaissance Orbiter. IBM and NASA on Thursday released an open-source AI model
[10]
IBM, NASA Release Open-Source AI Model Trained on Moon Data
International Business Machines and the National Aeronautics and Space Administration released an open-source artificial-intelligence model designed for research about the Moon. The open-source model is trained on extensive lunar observation data and is designed to "help scientists turn decades of
[11]
IBM and NASA offer the Moon in an open-source AI model
IBM and NASA announce the open-source release of their NASA-IBM Lunar Foundation Model, one of the first publicly accessible artificial intelligence foundation models for the Moon's scientific exploration, now available. "Trained on a vast dataset of lunar observations curated by IBM and NASA
[12]
IBM and NASA Release NASA-IBM Lunar Foundation Model for Scientific Exploration of the Moon
IBM and NASA announced the open-source release of the NASA-IBM Lunar Foundation Model, one of the first publicly available foundation models for scientific exploration of the Moon, now available. Trained on an extensive lunar observation dataset curated by IBM and NASA researchers, the model can
Share
Copy Link
NASA and IBM unveiled the Lunar Foundation Model, an open-source AI model trained on decades of lunar observation data from four missions. Available on Hugging Face, it outperforms existing methods by up to 23% in identifying ice deposits, craters and volcanic features to support NASA's Artemis program.
NASA and IBM Unveil Groundbreaking Lunar Foundation Model
NASA and IBM released the Lunar Foundation Model on Thursday, marking a significant advancement in lunar research capabilities
1
. This open-source AI model, now available on Hugging Face, represents the first publicly accessible tool designed to process and analyze decades of lunar observation data at scale2
. The model joins the Prithvi family of foundation models, which already includes applications for weather, climate and earth observation1
.Trained on more than 30 layers of data collected by nine instruments across four NASA missions, including the Lunar Reconnaissance Orbiter, the Lunar Foundation Model addresses a critical bottleneck in lunar research
3
. Previously, scientists had to manually examine maps and images or rely on low-resolution machine learning models designed for specific tasks1
. The new model integrates observations captured in multiple data formats, viewing angles and spatial scales, eliminating the need to build custom AI models from scratch for each research question2
.
Source: Engadget
Superior Performance in Identifying Critical Lunar Features
In benchmark tests, the AI model demonstrated remarkable accuracy improvements over widely used methods. The model outperformed SwinV2-B, a Microsoft-trained vision system commonly used as a baseline for image analysis tasks, by up to 23% in identifying key features on the lunar surface
3
. For crater detection specifically, it surpassed SwinV2-B by nearly 19% while using only half the training labels4
.The model's capabilities extend beyond simple feature recognition. It can identify ice deposits in permanently shadowed regions, map craters to select safe landing sites, and analyze volcanic features called Irregular Mare Patches
1
. When estimating polar ice prospectivity within the top meter of regolith, the Lunar Foundation Model reduced errors by up to 22% compared to SwinV2-B4
. This accuracy matters because lunar ice indicates the presence of water and oxygen, resources considered essential for establishing a future moon base and producing rocket fuel for missions to Mars3
.
Source: Live Science
Overcoming Unique Technical Challenges in Lunar Data Processing
Developing the Lunar Foundation Model required solving problems unique to lunar observation. Dr. Juan Bernabé-Moreno, director of IBM Research Europe, explained that the Moon's lack of atmosphere creates extreme lighting conditions that don't exist on Earth
5
. Without atmospheric scattering to soften shadows, lunar shadows appear knife-edged and pitch black, meaning shadowed pixels carry no information5
. A single crater can look completely different depending on when it was photographed.To address this, researchers incorporated explicit metadata describing solar angles and spacecraft positions directly into the model's architecture
4
. This allows the model to recognize actual terrain features rather than being misled by shadows. The team also employed masked-token learning, a technique where the AI is shown portions of lunar tiles and must predict concealed information across millions of examples4
.Traditional training methods proved disastrous for lunar applications. Bernabé-Moreno described initial attempts as a "complete disaster" because many craters look similar when photographed from orbit
5
. The solution involved dividing the Moon into wedges like an orange and completely separating training wedges from testing wedges, providing the model with necessary consistency5
.Related Stories
First-of-Its-Kind Dataset Enables Future Research
Alongside the Lunar Foundation Model, IBM and NASA scientists compiled SomBench, the first open-source lunar dataset of its kind
4
. This dataset brings together tens of thousands of maps and images collected by instruments from NASA's Lunar Reconnaissance Orbiter and Gravity Recovery and Interior Laboratory missions, as well as Japan's Selenological and Engineering Explorer5
.SomBench organizes nearly 2 million overlapping map patches into aligned tracks, ensuring data from different instruments, resolutions or angles all correspond when capturing the same lunar tiles
4
. The dataset incorporates over 30 spatially-aligned layers, combining high-resolution optical cameras, laser altimeters, radar reflectance tools and spectrometers that measure elemental density2
."That alone is a massive scientific contribution, because put the model aside: the community now has a co-registered dataset with more than two million data points," Bernabé-Moreno noted
5
. NASA's chief science data officer Kevin Murphy emphasized that collecting data is only part of the job, stating the agency must also make data easier for scientists to explore and use2
.Supporting the Artemis Program and Beyond
The timing of this release aligns with NASA's Artemis program, which plans to return astronauts to the Moon in 2028
3
. The program aims to test new technology for a sustained lunar presence and prepare for future Mars missions. Campbell Watson, senior research manager at IBM Research, explained that by making the model and dataset publicly available, the organizations want to provide the global scientific community with a shared foundation that researchers can adapt to new questions about the moon1
.The model recently demonstrated its practical capabilities when a SpaceX Falcon 9 rocket crashed into the Moon on August 5. When IBM fed an image of the impact to the model, it correctly identified the crash site as a new crater on its first attempt, despite the impact closely overlapping with an existing crater
5
.
Source: The Register
For researchers interested in using the model, hardware requirements vary depending on the application. Most fine-tuning experiments were conducted using Nvidia A100 GPUs, though smaller-scale experiments and inference workloads may be possible on more modest hardware
2
. Watch for the scientific community to build specialized applications on this foundation, potentially uncovering discoveries about lunar ice distribution, crater formation patterns and volcanic activity that could shape future moon explorations and planetary AI applications.References
Summarized by
Navi
[2]
[4]
Related Stories
Recent Highlights
1
OpenAI's AI Math Breakthrough on Navier-Stokes Equations Ignites Fierce Controversy Over Credit
Science and Research
2
Nvidia CEO Jensen Huang Declares AGI Has Arrived With OpenAI's GPT-6 Astra Release
Technology
3
Apple debuts $1,999 iPhone Duo foldable and A20 Pro chip as new CEO John Ternus takes the stage
Technology
Recent Highlights
Today's Top Stories
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.
