NASA tests AI-driven autonomous navigation on Mars with Perseverance rover

NASA has taken a major step toward autonomous space exploration after successfully testing artificial intelligence-driven navigation on Mars, allowing the Perseverance rover to plan and execute its own routes without direct human guidance.

The demonstration, carried out on December 8 and 10, marked the first time a Mars rover completed fully automated drives using AI-generated planning rather than step-by-step instructions from mission teams on Earth. The tests were led by NASA’s Jet Propulsion Laboratory (JPL) in Southern California, which manages Perseverance’s operations.

Mars lies an average of 140 million miles (225 million kilometres) from Earth, a distance that creates long communication delays and limits real-time human control. To overcome this challenge, researchers deployed vision-language models, a form of generative AI, trained on JPL’s extensive surface mission datasets.

Also read: Sam Altman suggests biometric ‘proof of personhood’ as a long-term fix for bot-driven social media

Using the same images and environmental data that human planners rely on, the AI system analysed the Martian terrain and autonomously generated waypoints, predefined locations where the rover pauses to reassess its surroundings and plan its next move. This allowed Perseverance to safely navigate complex landscapes filled with rocks, slopes and surface ripples without continuous human intervention.

NASA administrator Jared Isaacman said the test highlights how rapidly space exploration capabilities are evolving. He noted that autonomous systems will play a growing role as missions venture farther from Earth, where delays make traditional control methods inefficient.

According to Isaacman, AI-enabled navigation can help spacecraft respond more effectively to unexpected terrain, operate with greater independence and ultimately deliver higher scientific returns.

The work was coordinated from JPL’s Rover Operations Center, which is operated by the California Institute of Technology on NASA’s behalf and serves as the command hub for Perseverance’s daily activities.

Also read: EFF targets Apple, Meta and Google in new push for end-to-end encryption by default

Indian-origin roboticist Vandi Verma, a key member of the Perseverance engineering team at JPL, said the results demonstrate the potential of generative AI in off-planet mobility. She explained that the technology shows promise across the core elements of autonomous navigation, including perception, localisation, and planning and control, all essential for safe movement on alien terrain.

NASA sees this experiment as an early glimpse into how future robotic and human missions could operate with far greater autonomy, especially as exploration targets extend beyond Mars to more distant and challenging environments in the solar system.