NASA's Perseverance Rover Creates History By completing First Autonomous AI-Planned Drive On Mars

Artificial intelligence is now helping a Mars rover decide where to drive. In a new test, NASA allowed a smart computer system to plan safe paths on the surface of Mars, showing how space exploration is slowly becoming more independent from humans on Earth. For this six-wheeled rover, the team used vision-enabled AI technology to safely navigate the Martian surface without the involvement of human route planners, reported NASA.

NASA's Perseverance Mars rover has completed the first artificial intelligence-planned drives on another planet. Conducted on December 8th and 10th and led by NASA's Jet Propulsion Laboratory in Southern California, the test used generative AI to determine Perseverance's waypoints, a complex decision-making process normally performed manually by the mission's human rover planners.

According to NASA Administrator Jared Isaacman, this demonstration demonstrates how far NASA's technological capabilities have come and also shows how humans will explore other planets in the future.

He stated that such autonomous technologies could help missions operate more efficiently, tackle difficult terrain, and enhance scientific achievements at increasing distances from Earth. He described it as a powerful example of the responsible and careful use of new technology in real missions.

During this experiment, the team used a form of generative AI called a vision-language model to analyze data from JPL's surface mission dataset. The AI used the same images and information that human planners rely on to create waypoints. These waypoints are the precise locations where the rover receives a new set of instructions, allowing Perseverance to safely navigate the difficult Martian terrain.

The project was conducted from JPL's Rover Operations Center and partnered with Anthropic, which utilized the company's cloud AI models.

Mars is located approximately 140 million miles (225 million kilometers) from Earth on average. This great distance creates significant communication delays, making real-time remote control of the rover, known as joysticking, impossible. For this reason, during many missions over the past 28 years, the rover's routes have been planned and executed by human drivers.

These drivers study terrain and situational data to create routes through waypoints. To avoid potential hazards, these waypoints are typically spaced no more than 330 feet, or 100 meters, apart. These plans are then sent to the rover via NASA's Deep Space Network, which the rover then implements.

However, for Perseverance's drives on mission days 1,707 and 1,709, known as sols, the team adopted a new approach. The generative AI analysed terrain slope data from high-resolution orbital images and digital elevation models obtained from the HiRISE camera on NASA's Mars Reconnaissance Orbiter.

After identifying key geological features such as bedrock, outcrops, dangerous boulder fields, and sand waves, the AI generated a continuous route with waypoints.

To ensure that the AI instructions were consistent with the rover's flight software, engineers also tested the drive commands through JPL's digital twin, a virtual replica of the rover. More than 500,000 telemetry variables were then checked before sending the commands to Mars.

On December 8, Perseverance traveled 689 feet, or 210 meters, using waypoints determined by the generative AI. Two days later, the rover covered 807 feet, or 246 meters.

Vandi Verma, a JPL space robotics expert and member of the Perseverance engineering team, said the fundamentals of generative AI show great potential in simplifying key aspects of autonomous navigation for off-planet driving. These include recognising landforms, locating one's position, and determining and implementing the safest route.