Planning a lesson NASA’s Perseverance rover140 million miles to Mars, it’s a lot harder than setting a driving route here on Earth, where we can punch an address into Google Maps and be on our way in seconds. The rover’s course is usually planned by NASA’s Jet Propulsion Lab team to describe the terrain, obstacles and potential hazards, in case the rover is overrun or damaged.
For the first time, NASA’s JPL used AI to program persistence studies, and it seems to have succeeded.
Two demonstrations, which took place on December 8 and 10, were organized by Claude AI of Anthropic The models were also tested twice by JPL to ensure that the AI did not accidentally drive the rover into the crater. Endurance drove less than 1,500 feet on both drives with no documented problems.
NASA took the same approach by planning waypoints as it would with human operators. Claude was given the same satellite image and data from NASA’s Mars Reconnaissance Orbiter for JPL scientists to use, and asked to plot waypoints that Perseverance could safely handle.
The resulting path was slightly modified by NASA and sent to Perseverance, which then conducted the path independently.
“This demonstration shows how far our capabilities have advanced and is expanding the way we will explore other worlds,” said NASA Administrator Jared Isaacman. “Independent technology like this can help missions become more efficient, respond to challenging terrain and increase the return of science as the distance from Earth increases. It’s a strong example of teams applying new technology with care and commitment to real-world operations.”
You can watch the December 10 flyby on NASA’s YouTube channel, shortened to a 52-second video.
The route planned by Claude is shown in magenta, and the actual route taken is in orange. NASA scientists had to make small changes to the AI approach.
The most efficient way to do it
Although AI is best known as a provider of slop, which has been accused of rapidly degrading people’s internet knowledge, it can be useful in other areas of science. It takes time to analyze years of photos and data, plan Perseverance waypoints, and use them.
According to NASA, the waypoints are rarely separated by 330 meters, which means that Perseverance is exploring the red planet one football field at a time. Take its historic climb out of Jezero Crater in 2024. The journey took Perseverance 3.5 months and, all told, the rover climbed a total of 1,640 vertical meters. As of December 2025, the rover has traveled a total of 25 miles in nearly four years.
The goal, according to JPL roboticist Vandi Verma, is to allow Perseverance (again other Mars rovers) to travel a long distance while “reducing the user’s workload.”
Verma also notes that AI could be used to flag interesting features on Earth, saving time for human science teams by eliminating the need to manually scan “large rover image formats.”
“This exhibition shows how far our capabilities have advanced and expands how we will explore other worlds,” said Isaacman. “Independent technology like this can help missions become more efficient, respond to challenging terrain and increase the return of science as the distance from Earth increases. It’s a strong example of teams applying new technology with care and commitment to real-world operations.”
