NASA’s Perseverance rover completed a full marathon distance on Mars on June 14, 2026, five years and four months after it began driving. Opportunity needed 11 years and two months to cover the same 42.195 kilometres.

It is tempting to call this a victory for the first self-driving vehicle on Mars. The success is real, but that description compresses a more interesting story. Perseverance was not the first Mars rover to drive with some autonomy. It is the rover that has made autonomous driving faster, more continuous and far more central to everyday exploration.

I think that distinction matters because “self-driving” can make the rover sound as if it receives a destination and handles everything alone. It does not. Human planners still decide where it should go and what science it should do. The rover takes over much of the detailed work of moving safely between the instructions they send.

A marathon is evidence of a working mobility system

A NASA image release from June 2026 records the rover as a tiny green speck west of Jezero Crater, with its wheel tracks visible from orbit. NASA said Perseverance reached marathon distance on its 1,890th Martian day, or sol.

I recently wrote about what that marathon means within the wider, increasingly fragmented Mars programme. Seen specifically as a driving result, the time comparison with Opportunity is striking. It does not mean Perseverance drove twice as fast every day, because routes, terrain and science stops differ. It does show that its mobility system has supported a sustained working mission across a large distance.

That is a better measure of success than top speed. Perseverance can reach only about 152 metres per hour on firm, flat ground, and real drives are often slower. The point is to spend less mission time waiting for detailed movement instructions and more time reaching useful rocks, outcrops and sampling locations.

What AutoNav actually does

Earth cannot steer a Mars rover in real time because radio commands take minutes to cross the changing distance between the planets. Rover planners therefore send goals and route instructions in batches.

Perseverance’s onboard AutoNav system uses pairs of navigation-camera images to build a three-dimensional view of the ground ahead. It estimates the rover’s position, identifies rocks, steep patches and other hazards, then selects safe arcs around them while continuing to move. NASA calls this “thinking while driving”.

JPL’s technical account of the system explains that Perseverance can dedicate a separate computer to surface navigation. Its cameras expose images faster than those on earlier rovers, while its software can process the terrain and choose a path without stopping after every short segment.

The destination still comes from people.

Autonomous driving on Mars began before Perseverance

NASA has operated wheeled vehicles on Mars since Sojourner arrived in 1997, and autonomous hazard avoidance has developed across several generations. Opportunity could calculate safer local paths. In 2013, Curiosity began using autonomous navigation adapted from the Mars Exploration Rover programme.

Perseverance’s advance was to make that process substantially quicker and more capable in rough terrain. During a drive on sol 200 in September 2021, it covered 175.15 metres. After the first eight metres were used to establish a terrain map, AutoNav handled the remaining 167 metres, then the longest autonomous drive completed by a Mars rover in one sol.

A NASA account by JPL robotic-operations engineer Vandi Verma also described Perseverance reaching a narrow gap near a ridge and finding its own safe way through. These are modest distances by road-car standards. On Mars, where one poor wheel placement can end a multibillion-dollar mission, cautious progress is the intended behaviour.

The first AI-planned drives were a separate step

In December 2025, the mission moved part of the planning process from human route planners to generative AI for two demonstrations. A vision-language model analysed orbital images and elevation data, identified features including boulder fields, bedrock and sand ripples, then generated waypoints for the rover.

This AI was not sitting inside Perseverance and steering live.

According to JPL’s January 2026 announcement, engineers ran the proposed commands through a digital replica of the rover and checked more than 500,000 telemetry variables before transmission. Perseverance then drove 210 metres on December 8 and 246 metres on December 10. JPL described them as the first drives on another world planned by artificial intelligence.

I made a similar distinction in an earlier article about what AI currently does aboard operational spacecraft. The useful systems are narrow. They recognise a target, select a path or compress a decision that would otherwise consume scarce communication and staff time. They are not replacements for a mission team.

Why I would still call it a success

The marathon does not prove that every kilometre was autonomous, and two AI-planned drives do not establish that route planning can now be handed over wholesale. Both achievements sit inside layers of human selection, testing and supervision.

Even with those limits, Perseverance has shown that a rover can repeatedly inspect the ground, avoid local hazards and keep moving while its wheels turn. The newer planning tests pushed autonomy one level higher without pretending the safeguards were optional.

For future missions farther from Earth, that gradual transfer of routine decisions may matter more than any single speed record. Perseverance’s success is not that Mars now has an independent robotic driver. It is that carefully bounded autonomy has become ordinary enough to help a rover cross a marathon of unfamiliar ground.