A2RL Racers Close the Gap to Human Drivers to 0.85 Seconds

Every car in both races ran the same Dallara-based chassis, the same lidar and the same compute stack. The only thing left to test was the code.

The short version

Nine university teams sent identical AV-24 race cars, built on a Dallara chassis with a Honda engine, Luminar lidar and an Nvidia GPU for compute, onto WeatherTech Raceway Laguna Seca on Sept. 3, 2026, for the Indy Autonomous Challenge's first road-course passing competition, according to The Robot Report. Unimore Racing, from Italy's University of Modena and Reggio Emilia, won the head-to-head format; Purdue AI Racing finished second after setting an autonomous lap record of 1 minute 27.731 seconds. Two days later, at Imola, five teams in the separate Abu Dhabi Autonomous Racing League ran EAV-25 cars built on a Dallara Super Formula SF23 platform, and Team Kinetiz won from fourth on the starting grid, beating Constructor Racing by 10.963 seconds, Fox News reported.

Why the hardware is not the story

Both series are run as spec series: every entrant gets the same chassis, the same sensors and the same compute, so the only variable a team can change is its own driving software. That is a deliberate design choice, not an accident of budget. The Indy Autonomous Challenge's own announcement for the Laguna Seca event quotes Janam Sanghavi, the organization's technical lead, saying a road course adds "corner variations, braking zones, limited visibility, and elevation shifts" that a team's software has to handle while also reacting to a competing car, on top of everything an oval already demands.

The road-course passing format itself was new for the Sept. 3 event. On IAC's prior six passing competitions, held on ovals since 2022, teams raced for the fastest single lap with one car on the track at a time. At Laguna Seca, a trailing car had two laps to attempt an overtake, after which the two cars swapped positions and the field's engine power limits were raised, according to The Robot Report's account of the rules. That format change, more than any single lap time, is the harder engineering claim: it requires two autonomous systems to negotiate space with each other in real time, not just drive a clean line alone.

Teams also had to clear a new gate before they were allowed on the physical track at all. The Robot Report reported that IAC required, for the first time, that teams qualify through a simulation-based competition run on D-Space software before their cars were permitted to run at Laguna Seca. That is a way to filter out a control algorithm that cannot handle traffic before it gets anywhere near the physical chassis, and it pushes more of the actual competition into code written and tested away from the track entirely.

How close autonomous cars now run to human drivers

The clearest number to come out of the Imola weekend is not who won. During testing there, the fastest autonomous lap ran within 0.85 seconds of a benchmark lap set by a Super Formula driver, per Fox News. A year earlier, at the same series' previous venue in Abu Dhabi, former Formula 1 driver Daniil Kvyat's own lap was only 1.58 seconds faster than the autonomous car racing against that benchmark.

Venue	Season	Gap to human benchmark
Abu Dhabi	prior season	1.58 seconds
Imola	2026	0.85 seconds

Those two numbers are not a clean trend line. Abu Dhabi and Imola are different circuits with different benchmark drivers, so the gap narrowing is not proof of a steady, measurable rate of improvement the way a chip's tokens-per-watt curve would be. What the two numbers do establish, on their own terms, is that autonomous race software is now landing within a couple of seconds of a professional driver on two different tracks a season apart, which was not true when IAC's passing competitions were confined to ovals.

The counter-case: nobody has a driver to root for

Fox News's own framing of the story is the honest counterweight to the lap-time numbers: "Part of the fun of racing is having someone to root for. Fans follow drivers, get to know their personalities and pick favorites." Take the driver out of the car, and a race becomes a systems competition that a general-admission crowd has less reason to watch than a lap-time chart does. IAC has not published attendance or broadcast-viewership figures for either the Laguna Seca or Imola weekends, so there is no data yet on whether removing the driver actually cost the format an audience, only Fox News's argument that it might.

What IAC's format has produced is a pipeline into commercial autonomy work: The Robot Report's reporting on the event notes that alumni of past IAC teams have gone on to found driveblocks and Autonoma AI, two companies now working on autonomous-driving software outside racing. IAC itself is run through Aidoptation BV, whose chief executive, Paul Mitchell, is also IAC's own chief executive. It is the same research-to-product path already showing up in warehouse humanoid deployments and in the scramble to give industrial robot hands a working sense of touch: a narrow, measurable competition first, a commercial pitch second.

What would change this read

The number to watch is not a lap time; it is whether IAC or A2RL runs the same two circuits again with the same benchmark drivers, which would turn Abu Dhabi's 1.58 seconds and Imola's 0.85 seconds into an actual measured trend instead of two isolated data points. Neither series has announced its next race date as of this writing. Until one circuit hosts a second season under identical rules, the honest read is that autonomous race cars are close to professional drivers on at least two tracks, not that the gap is closing at any particular rate.