NASA is developing AI-powered GPS for navigating the Moon

As the visibility of landmarks on the Moon is affected by its thin atmosphere, making it hard to judge how far or how large faraway mountains or other features may be, being able to use a local map would be just as useful for a robotic mission as for a human, explained Yew in a press release.

Yew's background is in aerospace engineering, working at the Goddard Space Flight Center in hardware development for missions including Astro-H, MMS, GPM, and FASTSat, and directing various research efforts to come up with and test groundbreaking spaceflight tech.

He specifically focuses on "Guidance, Navigation, and Control components on small satellite architectures," as describes his bio. And has been involved with developing cutting-edge hardware like zero-gravity parabolic flights and early-stage ice sampling mechanisms for planetary science research.

Backing up LunaNet

Another aspect of Yew’s research is that as NASA is developing the navigational and communication capabilities for future colonies on the Moon, it looks to build systems that could serve as backups for each other.

Its LunaNet program seeks to bring internet and location services to the Moon, building up a network of lunar satellites for use by spacecraft and astronauts for use in establishing a long-term presence at or on the Moon.

NASA and other space agencies, like the ESA and JAXA, would join efforts to create LunaNet, installing the ability to communicate with each other and share data up to an altitude of 125 miles (200 kilometers) across the whole Moon. 

As explained by Cheryl Gramling, NASA’s Associate Chief for Technology of the Mission Engineering and Systems Analysis Division, in a NASA feature article:

“LunaNet will provide a new paradigm for Earth-independent navigation, assuring crewed and robotic missions can quickly and accurately determine their locations and feed that forward to their planning systems.”

Besides allowing astronauts to share information and messages, LunaNet would have other capabilities, like using space weather instruments to detect dangerous space activity, such as solar storms, and provide immediate warnings to users without having to wait for direction from network managers on Earth.

It would also aid science exploration by allowing researchers to take measurements using its radio and infrared optical communication links, and its antennas would find application in radio astronomy. 

While LunaNet’s potential is impressive, it is still a long way from being operable, and Yew believes that there are some areas and situations where his approach would likely be more applicable and serve as a needed alternative to LunaNet, especially in cases of network disruptions caused by power or signal outages.

As he explained to PopSci, “When we’re doing human expeditions, you always want [backup systems] for very dangerous missions,” adding that his AI would not be “tied to the internet, per se,” although it can be.

In the NASA press release, Yew also used the example of lunar crater exploration, where the horizon would just be the crater’s rim. Boulders and ridges are other geographic features that may be hard for a person to precisely locate on the Moon but would also be navigable by his system.

How lunar GPS would work

In his work on the lunar GPS, Yew is educating an AI on how to recreate lunar horizon features as digital panoramas.

This is accomplished by utilizing data from NASA’s Lunar Reconnaissance Orbiter, specifically the Lunar Orbiter Laser Altimeter (LOLA), while also employing GIANT (Goddard Image Analysis and Navigation Tool) to verify navigation data.

GIANT can quickly and precisely analyze images to figure out the distances between visible landmarks. 

As Yew explained to PopSci, his system uses an outline of the landscape to clearly see where the lunar ground is in relation to space.  

One application of the technology, once fully developed, would be to program handheld devices with localized terrain and elevation data to be used by lunar explorers. 

Interesting Engineering (IE) reached out to Dr. Alvin Yew for more details on his work. Here is what he said.

The following exchange has been lightly edited for clarity and flow.

Interesting Engineering: How is the AI trained and utilized to steer on the lunar surface? 

Dr. Alvin Yew: Everyone has been very zealous about AI for the technology, but we’re not quite there yet! Most of what we’ve been doing so far has been sensitivity studies and algorithm development. Pretty boring stuff, but it lays the groundwork for the future.

 IE: What are the potential applications of this technology on Earth? 

The technology applications for Earth include localizing without GPS. For example, if you were lost while hiking with no reception, or perhaps your GPS is dancing around the city due to interference from surrounding structures.

 IE: What is next for this research? How will it be tested?

Good question. For the project, the next steps include evaluating the robustness of the algorithms, deciding on constraints for use cases, selecting the hardware requirements/limitations, field testing with lunar analogs, and finally, a demonstration on the Moon.

For the last bullet, I say “good question” because my team and I are oversubscribed … so progress on this will likely be slow in the next year! I currently spend most [of] my time on In-Situ planetary and lunar sample collection and enrichment to look for trace species (i.e., noble gases, hydrocarbons, organics). 

For reference, the In-Situ Resource Utilization program involves figuring out how to produce products from materials available in space, local to their environment. It’s important to develop this program so that a sustainable infrastructure can be put in place for when NASA comes back to the Moon through the Artemis program and other space expeditions.