IT’S 2035, AND THERE ARE SCIENTISTS WORKING ON THE MOON. They stay for weeks or even months at a time in a small cabin near the south pole, powered by both solar and nuclear energy, which gets them through the freezing, two-week-long lunar nights. They’re experimenting with growing plants, mining the moon’s water ice, and extracting oxygen and metal from moon rocks. This will be made possible with a new supply-chain wave — space logistics.
While the ultimate goal is to become self-sustaining enough to get to Mars, the equipment, tools and resources they need are still mostly shuttled from Earth. Private companies are launching rockets with increasing regularity from spaceports all over the world, which means that the price of ferrying cargo to the moon is decreasing. And AI is playing an important, and still growing, role in this nascent lunar logistics industry.
This sci-fiesque scenario is speculative, and fairly optimistic. We don’t yet know exactly what will be achieved within this timeframe, and some experts think it will be a little longer before a moon base is established. But it’s based on plans that are already underway; this isn’t merely the realm of fantasy. Both China and the U.S. are leading international missions that have the ultimate aim of establishing moon bases, with new ground being broken all the time.
In late 2022, the chairman of China Aerospace Science and Technology Corporation, Wu Yansheng, announced intentions to establish its International Lunar Research Station by the 2030s. Meanwhile, NASA is planning to land humans on the moon for the first time in half a century in 2025. After that, it plans to operate a ‘Lunar Gateway’ space station that will orbit the moon, and to then ultimately establish its Artemis moon base. This will allow scientific advances, as well as serving as a stepping stone for a future crewed mission to Mars.
These things are much easier said than done, of course, and plans keep getting delayed. But concrete progress is being made towards these goals. “The 2030s is probably when we’re going to see that base on the moon start to solidify,” says Laura Forczyk, executive director of space consulting firm Astralytical. “We will start to see mining operations happening. We might start to see lunar telescopes being built on the far side of the moon.” For many of the processes involved in this activity, “automation will be necessary” she adds. “One way that AI is being used already is in landing on the surface of the moon. This is going to be important, both for cargo delivery and passenger delivery.”
Another area in which AI will be useful is science and construction. Metalysis is a British company that has won contracts with the European and UK space agencies to figure out how to turn moon dust into oxygen and metal powder, in situ on the lunar surface. This powder can then be 3D-printed and used as a building material. “AI is a critical tool for our technology to function, long-term, on the moon,” says the company’s MD, Ian Mellor. Machine learning and AI, he says, “can help with operating the plant more efficiently,” and it could autonomously intervene to help fix problems with machinery as they arrive. “I think without AI, [the moon base] won’t happen.”
Then there is the design and operation of the vehicles that will actually do the transporting of cargo to the moon. “Artificial intelligence will probably help engineers design better and more efficient rockets,” says Ian Crawford, Professor of Planetary Science and Astrobiology at Birkbeck, University of London. He points out that several teams are currently unlocking new possibilities for transportation of cargo to the moon.
This includes the European Space Agency’s Argonaut lunar lander, which aims to deliver scientific instruments and cargo to the Moon by the end of the 2020s. SpaceX is developing its own Starship lunar lander, which will be a game-changer due to the high masses it will be designed to carry. Then there are the private companies that have been awarded contracts as part of NASA’s Commercial Lunar Payload Services (CLPS, pronounced “clips) programme. It’s been announced as part of this that the companies Astrobotic, Intuitive Machines and Masten Space Systems will deliver payloads to the moon in 2023, with further deliveries scheduled from Astrobotic, Firefly Aerospace, Intuitive Machines and Draper in 2024 and 2025.
Most of these deliveries include multiple items sharing space inside a rocket’s cargo hold. For example, Firefly’s first Blue Ghost mission will carry 12 commercial and government payloads, including 10 that are part of NASA’s CLPS initiative. They will be taken to Mare Crisium, a low-lying basin on the moon’s near side. There, Firefly will operate the payloads for a lunar day – about 14 Earth days – and into the freezing dark of a lunar night.
The tools that will be delivered are intended to carry out research and demonstrate various capabilities. This includes testing samples of moon rock, finding out the tolerance of computing to radiation, and looking into the mitigation of lunar dust. The data captured will also provide insights into how space weather and other cosmic forces impact Earth, among other research. It will be the first of many Firefly missions, the company’s communications team says, to the Moon and beyond.
Astrobotic’s Peregrine Lander will carry 28 payloads from eight nations, including 14 for NASA’s CLPS. This encompasses an array of scientific instruments and exploration devices, including Astrobotic’s own shoebox-size CubeRovers, which will explore the Moon and transmit real- time HD images and data back to Earth. The lander will also deliver a more whimsical payload: the DHL MoonBox: a pod full of mementoes from people all over the world, which will serve as a lunar time capsule.
Christoffel Doorman is a research engineer at AI company InstaDeep, who has a Bachelor’s and Master’s Degree in Theoretical Physics, as well as a Master’s in Machine Learning. He’s working on a project that could have particular relevance for missions like Astrobotic’s Peregrine 1 and Firefly’s Blue Ghost. It uses artificial intelligence to design optimal solutions for packing items into a container. “This may sound easy,” he says, “but there are so many possibilities for packing a container with a set of items, you just can’t compute the best configuration. You can’t even compute a fraction of the possibilities.” Items being carried to the moon, which are often fragile and awkwardly shaped, can be particularly difficult to pack efficiently.
What InstaDeep’s AI-Powered 3D packing software DeepPack does is simulate being a human packing a container, learn from its actions, and iterate this process over and over. “This is a really new area of research,” Doorman says. “You have so many constraints to take into account. AI and reinforcement learning are able to pack these difficult sets of items into a difficult space millions of times, and then figure out the patterns to come up with a really good packing solution that surpasses human capabilities. In the end, that may result in either having smaller spacecraft, or even better, having to have less frequent flights to the moon, to get the cargo there. Decreasing the volume of the load is really important.”
This is just one small way that AI and machine learning can play a role in this new era of lunar exploration, Doorman says. Others include “denoising” video recordings from the moon and planning a vehicle’s specific path towards the lunar surface before liftoff. In the next couple of decades, he predicts, AI-powered systems that make recommendations to astronauts “will increase gradually, starting with small tools that assist humans during spaceflight.” Looking into the 2040s, he says, “the autonomy of spacecrafts themselves will gradually increase.” Once capabilities for full autonomy are reached, “things will go really rapidly, and the need for humans to be in space will decrease drastically.”
Getting humans into space is expensive, difficult and risky, but Prof. Ian Crawford hopes that, however advanced AI becomes, support for human spaceflight continues. “We’re not using little robots to try to explore Antarctica,” he points out, “and if we did, they’d last a few days, break down, and you’d have to send out a human engineer to fix them. I think this would be the same in space.”
He continues: “Obviously, we would use robots where we can, and the more intelligent we can make the robots, the better. But if you try to take humans out of the loop completely, then I think we’re going to be worse off, from a scientific point of view.” From this perspective, it’s encouraging news that not one but two international coalitions of space agencies are racing to land humans back on the lunar surface, and ultimately to build bases there. Once that takes place, the even more audacious goal of setting foot on Mars is one step closer. As Laura Forczyk says, “There is a fascinating future ahead of us.”
