Navigating a microgravity environment is challenging even for trained astronauts. However, this poses a much greater challenge for autonomous robots, limiting their use in places such as space stations.
Now Stanford researchers have demonstrated the potential for more autonomous future space missions by using artificial intelligence to control a free-flying robot on the International Space Station (ISS).
How does it work?
Using NASA's cube-shaped Astrobee robot, the Stanford research team demonstrated how a machine learning system can plan safe routes through the ISS's complex and crowded modules much faster than existing methods. This development overcomes a long-standing hurdle for space robots: the challenge of moving quickly and safely in this extreme technological environment with limited computing power and minimal human input.
Lead researcher Somrita Banerjee found that the station's labyrinthine jumble of equipment and experiments made movement planning particularly difficult. That's because algorithms that worked well on robots on Earth were slow on older, radiation-resistant computers approved for space travel.
To overcome these limitations, Banerjee and his team started with a standard optimization approach that breaks down a complex motion planning problem into many smaller steps. They then trained an AI model on thousands of previously calculated paths. In this way, the system can begin each new planning with a “warm start” based on empirical values, instead of calculating from scratch.
Banerjee explains this approach: “Using a hotstart is like starting with a route that real people have taken before, rather than drawing a straight line on a map when planning a trip. You start based on experience and optimize from there.”
This approach significantly reduced the actual computation time while allowing for rigorous security checks before test runs. According to the researchers, in tests conducted at the station, the calculation of routes generated by the AI-powered hot start was about 50 to 60% faster than traditional plans.
“This is the first time that artificial intelligence will be used to control a robot on the ISS,” Banerjee said. “This shows that robots can move faster and more efficiently without compromising safety, which is critical for future missions where humans cannot constantly take the lead.”
The foundations for future operations are being prepared
The team notes that similar AI-driven planning would allow robots to take on inspection, logistics and scientific tasks on future missions to the Moon, Mars and beyond, allowing astronauts to focus on higher-priority tasks.
“As robots move further away from Earth and missions become more frequent and cost-effective, we will no longer be able to control them from the ground through constant teleoperation. Autonomy with built-in safety measures is not only helpful, but also critical to the future of space robots,” Banerjee added.
Bir yanıt yazın