NASA's Perseverance rover has completed its first-ever drive on Mars fully planned by artificial intelligence, the space agency announced.
The demonstration, carried out on Dec. 8 and Dec. 10 of 2025, showed that generative AI could safely plan rover routes across Mars' rugged terrain without manual input, automating a labor-intensive decision-making process typically performed by human planners on Earth.
"This demonstration shows how far our capabilities have advanced and broadens how we will explore other worlds," NASA Administrator Jared Isaacman said in a statement.
Autonomous technologies like this, he added, could help future missions operate more efficiently, respond to hazardous terrain, and "increase science return" as spacecraft venture farther from Earth. "It's a strong example of teams applying new technology carefully and responsibly in real operations."
Because Mars is an average of 140 million miles (225 million kilometers) from Earth, communication delays make real-time control impossible. For decades, mission teams have instead planned daily routes by hand; human "drivers" analyze terrain and rover status data, then map out paths using waypoints typically spaced no more than about 330 feet (100 meters) apart to avoid hazards.
Those plans are sent to Mars via NASA's Deep Space Network, where the rover executes them, according to NASA.
Perseverance's recent AI-driven test drive was led by NASA's Jet Propulsion Laboratory in California, which built and operates the car-sized rover, in collaboration with Anthropic using the company's Claude AI models.
To plan the routes, the AI analyzed the same images and data used by human planners. According to NASA, this included images captured by a camera aboard the agency's Mars Reconnaissance Orbiter as well as terrain-slope data from computer models. From this information, the AI identified key surface features such as rocks, steep slopes and boulder fields, then mapped out a route for the rover to follow.
That route included navigation waypoints, which are fixed surface coordinates that the rover is instructed to reach in sequence. In the video above from the rover's Dec. 10 drive along the rim of Jezero Crater, a waypoint appears as a blue circle. Pale blue lines trace the rover's wheel tracks, while black lines show the alternate route options the rover evaluated, NASA said.
During the two test drives, Perseverance traveled nearly 1,500 feet (456 meters), the space agency said. Before sending commands to Mars, the mission team extensively tested the instructions using a detailed "digital twin" of Perseverance to confirm the rover could safely carry out the plan, according to the statement.
"The fundamental elements of generative AI are showing a lot of promise in streamlining the pillars of autonomous navigation for off-planet driving," Vandi Verma, a space roboticist at JPL and a member of the Perseverance engineering team, said in the statement.
"We are moving towards a day where generative AI and other smart tools will help our surface rovers handle kilometer-scale drives while minimizing operator workload," she added, "and flag interesting surface features for our science team by scouring huge volumes of rover images."
