The Perseverance rover, which landed in Jezero Crater on Mars in 2021 to examine evidence of ancient life and sample rocks, has achieved another milestone by NASA. In December 2025, the six-wheeled rover made its first visit to Mars using routes entirely with artificial intelligence. A vision-enabled AI examined high-resolution images and terrain data to map out safe waypoints, enabling Perseverance to travel across the Martian terrain on its own.

AI-powered navigation During an experiment at the Jet Propulsion Laboratory, engineers had to plan Perseverance’s Dec. 8 and Dec. 10 drives using generative AI, according to reports.

The vision-language model analyzed orbital images and terrain data to detect hazards (cliffs, sand waves, steep slopes) and generate a path with safe station points for the rover. Perseverance then drove two AI-planned routes, each covering a distance of 210 meters (246 ft). Engineers tested the AI-generated commands in the rover’s simulated digital twin to confirm more than 500,000 variables were correct, and that’s why it was safe to deploy the commands on Mars.

Implications and future exploration According to NASA, Mars is 225 million kilometers away, which means that joystick control is not possible in real time; In practice, rover missions are planned by hand using a sequence of waypoints. Administrator Jared Isaacman said AI-based planning could help make missions more efficient and increase science returns at such distances. JPL engineers say generative AI could reduce the planning process and eventually allow rovers to drive on the kilometer scale with minimal operator intervention.

This development opens the door to smart robotic systems on the Moon and Mars that will help in future exploration and human missions.