NASA has successfully utilized the Claude large language model to autonomously plan a driving route for the Perseverance rover on Mars. This represents the first application of AI in planetary exploration. NASA has not previously used an AI system to operate a robotic vehicle on another planet.
This advancement could transform the planning of future missions to deep space. The successful test conducted in December 2025, during which Perseverance traveled approximately 400 meters across the Jezero crater, demonstrates how AI can enhance the efficiency of space agencies.
What is NASA doing to make space exploration better with AI?
Adding AI to NASA’s mission operations is a big step forward in technology. People had to be very careful when they drove the Perseverance rover in the past. You need to carefully check every command sequence to make sure the rover doesn’t slide, tip over, or get stuck on the other side of Mars.
According to Engadget’s first report, operators have always planned each drive by creating a “breadcrumb trail” of waypoints and combining rover camera images with orbital images to avoid loose sand and big rocks.
NASA’s Jet Propulsion Laboratory (JPL) engineers provided Claude Code, Anthropic’s specialized programming agent, with years of relevant mission data for the test. The AI was tasked with planning its route prior to deployment.
It developed the drive path in ten-meter segments, continuously adjusting and evaluating its suggested waypoints to create a route suited for rocky and challenging terrain.
How Claude and NASA Built the Martian Drive
NASA engineers ensured that all systems were functioning properly by following their standard procedures before sending any commands to Mars. They utilized JPL’s high-fidelity simulation software to verify the accuracy of Claude’s route.
You can also use this program to verify the operational status of all Perseverance’s standard commands. The human team reviewed the AI’s work to confirm its safety after completing this critical step.
The engineers indicated that the final route had only “small changes.” One of these changes was due to the team incorporating images taken from the ground that were not part of the original dataset provided to Claude.
The results of this successful test will have a big impact on how NASA works in the future. The agency’s engineers think that using AI this way could make planning a route take half as long and make drives more reliable.
This increase in efficiency means that scientists and engineers can spend less time planning and more time analyzing data and making new discoveries. The source material says that this tool will help us “learn a lot more about Mars.”
NASA also wants to make this partnership bigger. They say that self-driving AI systems could be very helpful for controlling probes in the outer parts of the solar system, where communication is slow.
