NASA\/JPL-Caltech<\/figcaption><\/figure>\n
The\u00a0team for the\u00a0six-wheeled scientist used\u00a0a vision-capable\u00a0AI\u00a0to\u00a0create a safe route over the Red Planet\u2019s surface\u00a0without the input of human route planners.<\/em>\u00a0<\/p>\n NASA\u2019s\u00a0Perseverance\u00a0Mars\u00a0rover\u00a0has\u00a0completed the first\u00a0drives on another world\u00a0that were\u00a0planned\u00a0by\u00a0artificial intelligence.\u00a0Executed on Dec.\u00a08\u00a0and\u00a010,\u00a0and\u00a0led by the agency\u2019s Jet Propulsion Laboratory in Southern California,\u00a0the demonstration used generative AI to create waypoints for Perseverance,\u00a0a complex decision-making task typically performed manually by the mission\u2019s human rover planners.\u00a0\u00a0<\/p>\n \u201cThis demonstration shows how far our capabilities have advanced and broadens how we will explore other worlds,\u201d said NASA Administrator Jared Isaacman. \u201cAutonomous technologies like this can help missions to\u00a0operate\u00a0more efficiently, respond to challenging terrain, and increase science return as distance from Earth grows.\u00a0It\u2019s\u00a0a strong example of teams applying\u00a0new technology\u00a0carefully and responsibly in real operations.\u201d\u00a0<\/p>\n During the demonstration,\u00a0the team\u00a0leveraged\u00a0a type of\u00a0generative AI\u00a0called\u00a0vision-language\u00a0models to analyze existing data from JPL\u2019s\u00a0surface mission dataset.\u00a0The AI used the same imagery and data\u00a0that\u00a0human planners\u00a0rely on\u00a0to generate waypoints\u00a0\u2014\u00a0fixed locations where the rover takes up a new set of instructions \u2014 so that Perseverance could safely navigate the challenging Martian terrain.\u00a0<\/p>\n The initiative was led out of\u00a0JPL\u2019s\u00a0Rover Operations Center\u00a0(ROC)<\/a>\u00a0in collaboration with Anthropic, using\u00a0the company\u2019s Claude AI\u00a0models.\u00a0\u00a0<\/p>\n Mars is on average about 140 million miles\u00a0(225 million kilometers)\u00a0away\u00a0from Earth.\u00a0This vast distance creates a significant communication lag, making real-time remote operation\u00a0\u2014\u00a0or\u00a0\u201cjoy-sticking\u201d\u00a0\u2014\u00a0of a rover impossible.\u00a0Instead, for\u00a0the\u00a0past\u00a028\u00a0years, over several missions,\u00a0rover routes\u00a0have\u00a0been planned and executed\u00a0by\u00a0human\u00a0\u201cdrivers,\u201d\u00a0who\u00a0analyze\u00a0the\u00a0terrain\u00a0and status data to sketch a route using waypoints,\u00a0which are\u00a0usually spaced no more than 330 feet (100 meters) apart to avoid\u00a0any potential\u00a0hazards.\u00a0Then they send the plans via\u00a0NASA\u2019s\u00a0Deep\u00a0Space Network<\/a>\u00a0to the rover, which executes them.\u00a0<\/p>\n But for Perseverance\u2019s drives on the 1,707 and 1,709 Martian days, or sols, of the mission, the team did something different: Generative AI provided the analysis of the high-resolution orbital imagery from the HiRISE<\/a>\u00a0(High Resolution Imaging Science Experiment) camera aboard NASA\u2019s Mars Reconnaissance Orbiter and terrain-slope data from digital elevation models. After identifying critical terrain features \u2014 bedrock, outcrops, hazardous boulder fields, sand ripples, and the like \u2014 it generated a continuous path complete with waypoints.<\/p>\n To ensure\u00a0the AI\u2019s instructions were fully compatible with the rover\u2019s flight software,\u00a0the engineering team\u00a0also processed the drive commands through JPL\u2019s\u00a0\u201cdigital twin\u201d\u00a0(virtual replica of the rover),\u00a0verifying over\u00a0500,000 telemetry variables before\u00a0sending\u00a0commands\u00a0to Mars.\u00a0<\/p>\n On Dec.\u00a08, with generative AI waypoints in its\u00a0memory, Perseverance drove\u00a0689\u00a0feet (210\u00a0meters).\u00a0Two days later,\u00a0it drove\u00a0807\u00a0feet (246\u00a0meters).\u00a0<\/p>\n \u201cThe fundamental elements of generative AI are showing a lot of promise in streamlining the pillars of autonomous navigation for off-planet driving: perception (seeing the rocks and ripples), localization (knowing where we are), and planning and control (deciding and executing the safest path),\u201d said Vandi Verma, a space roboticist at JPL and a member of the Perseverance engineering team. \u201cWe are moving towards a day where generative AI and other smart tools will help our surface rovers handle kilometer-scale drives while minimizing operator workload, and flag interesting surface features for our science team by scouring huge volumes of rover images.\u201d<\/p>\n \u201cImagine intelligent systems not only on the ground\u00a0at Earth, but\u00a0also\u00a0in\u00a0edge applications in our rovers, helicopters, drones, and other surface elements trained with\u00a0the\u00a0collective\u00a0wisdom of\u00a0our NASA\u00a0engineers,\u00a0scientists, and astronauts,\u201d said\u00a0Matt Wallace,\u00a0manager\u00a0of JPL\u2019s Exploration Systems Office.\u00a0\u201cThat\u00a0is\u00a0the\u00a0game-changing technology we need to establish the infrastructure and systems required for a permanent human presence on the\u00a0Moon and\u00a0take the U.S.\u00a0to Mars and beyond.\u201d\u00a0\u00a0\u00a0<\/p>\n Managed for NASA by Caltech, JPL is home to the Rover Operations Center (ROC). It also manages operations of the Perseverance rover<\/a> on behalf of the agency\u2019s Science Mission Directorate as part of NASA\u2019s Mars Exploration Program portfolio.<\/p>\n For more information on\u00a0the ROC, visit:<\/p>\n https:\/\/www.jpl.nasa.gov\/roc<\/a><\/p>\n<\/p>\n News Media Contacts<\/strong><\/p>\n DC Agle Karen Fox \/ Molly Wasser 2026-008<\/p>\n
NASA\/JPL-Caltech<\/figcaption><\/figure>\nProgress for\u00a0Mars,\u00a0beyond<\/strong>\u00a0<\/h2>\n
![\"An](\"https:\/\/www.nasa.gov\/wp-content\/uploads\/2026\/01\/e2-pia26645-hyperdrive-overlay.png?w=1366\")
<\/a><\/figure>More about Perseverance<\/strong>\u00a0<\/h2>\n
Jet Propulsion Laboratory, Pasadena, Calif.
818-393-9011
agle@jpl.nasa.gov<\/a><\/p>\n
NASA Headquarters, Washington
202-358-1600
karen.c.fox@nasa.gov<\/a>\u00a0\/ molly.l.wasser@nasa.gov<\/a><\/p>\nExplore More<\/h2>\n<\/div><\/div>\n
![\"\"](\"https:\/\/assets.science.nasa.gov\/dynamicimage\/assets\/science\/missions\/cold-atom-laboratory\/cal-team.jpg\")
<\/figure>\n<\/p><\/div>\n<\/p><\/div>\n