While space has been hailed as the “final frontier”, there is no fear for Anthropic’s Cloud, which has broken new ground by planning a route for a NASA vehicle to Mars.
The 400-meter excursion across Mars’ rugged terrain in December marked the first time NASA has used an AI model to determine a path for its Perseverance rover on the Red Planet.
The exercise was led by the agency’s Jet Propulsion Laboratory (JPL) in Southern California, with the cloud plotting waypoints for Perseverance – a decision-making task that NASA has described as “complex” and which is typically performed by human planners.
The Perseverance rover is a car-sized robot equipped with cameras and scientific instruments that is being used for research on Mars since 2021.
But since the planet is about 225 million kilometers away from Earth, and its surface is rocky and difficult to walk on, driving the vehicle is a challenge, with the possibility of it overturning or getting stuck an ever-present risk.
Thus, NASA typically deploys human operators to painstakingly create waypoints for the rover to follow, using a combination of images from space and onboard cams – known as a “breadcrumb trail”.
On this occasion, waypoints were provided by the cloud after JPL engineers decided to see if AI could plan a route as effectively as a human. Doing so involved providing Anthropic’s programming agent, Cloud Code, with reams of data collected over 28 years of missions to Mars, to allow it to write commands for Perseverance.
“(By) using his vision capabilities to analyze overhead images, Claude planned Perseverance’s breadcrumb trail point by point,” according to Anthropic’s account. In doing so: “It strung ten-metre sections together into a path, then repeated to refine way-points, criticizing its own work and suggesting revisions.”
Given the importance of the task, JPL engineers reviewed Cloud’s work before approving it, running it through digital twin simulations and verifying more than 500,000 telemetry variables. He found that only minor changes were needed.
In early December, commands were sent to Perseverance via NASA’s Deep Space Network, each taking about 20 minutes to transmit due to the distance between Earth and Mars. The result was encouraging: successful drives of 210 meters and 246 meters respectively.
The hope now is that the cloud’s contribution to the successful demo will pave the way for even more input from AI.
“The fundamentals of generative AI are showing great promise in streamlining the pillars of autonomous navigation for off-planet driving: perception (seeing rocks and waves), localization (knowing where we are), and planning and control (deciding and executing the safest path),” explained Vandi Verma, JPL space roboticist and Perseverance engineering team member. “We are moving toward a day where generic AI and other smart tools will help our surface rovers handle kilometer-scale drives while reducing operator workload.”
This type of AI-powered functionality could be important in future missions, including NASA’s Artemis mission to return humans to the Moon.
