NASA has successfully tested an autonomous navigation system for the Perseverance rover on Mars, relying on artificial intelligence for route planning.

Autonomous Navigation on Mars

The Perseverance team used AI to generate the rover's waypoints. On two separate days, Perseverance traveled a total of 456 meters without human control, based on the AI-generated points.

This approach is necessitated by the approximately 25-minute delay in communications between Earth and Mars. Currently, technicians on Earth examine images and elevation data to program a series of waypoints, usually no more than 100 meters apart. These plans are then sent to Perseverance via NASA's Deep Space Network (DSN).

AI Implementation Details

In this demonstration, the AI model analyzed orbital images from the Mars Reconnaissance Orbiter's HiRISE camera, as well as digital elevation models. The AI, based on Anthropic's Claude AI, identified obstacles such as sand traps, boulder fields, rocky outcrops, and generated a safe path. Subsequently, Perseverance's auto-navigation system took control, processing images and driving plans while in motion.

Before transmitting the waypoints to Perseverance, NASA's Jet Propulsion Laboratory (JPL) used a twin model of the rover, called the "Vehicle System Test Bed" (VSTB), located in JPL's Mars Yard. This engineering model allows for problem-solving and testing new features in a controlled environment.

The Future of AI in Space Exploration

NASA is working on a system where Perseverance can use AI to re-localize itself autonomously, overcoming the uncertainties that increase as the rover moves further without human assistance. AI could be trained to compare orbital images with ground-level images.

AI is set to play an increasingly important role in planetary exploration, with concepts including swarms of flying drones controlled by AI to expand the explorative reach on Mars. NASA's Dragonfly mission to Saturn's moon Titan will also make extensive use of AI for autonomous navigation and data management.