Future
Bridging virtual and real-life racing to train AI racing cars
by Samarth Kanal
5min read
Artificial Intelligence could have a host of applications in motorsport, and one of those is to further self-driving technology by using the track as a testbed. A2RL is doing just that - but taking its efforts to the virtual world to bolster its real-world findings.
Sign up for a newsletter and we'll make sure you're fully up-to-date in the world of race technology
In A2RL - the Abu Dhabi Autonomous Racing League - this ‘loop’ includes a virtual rendition of the track to further teams’ software development.
.jpg?cx=0.5&cy=0.5)
Racing in the cloud
The series has even hosted ‘man versus machine’ spectacles involving ex-Formula 1 racer Daniil Kvyat, who has bested the AI cars with ease.
This is a cloud-based simulation environment supplied by a US-based company that recreates A2RL’s single-seater racing cars, tracks and various conditions to help the teams strengthen their machines and their learning to handle real-life racing.
Ex-F1 driver Daniil Kvyat racing an AI-driven car in 2024 at Yas Marina Circuit
A2RL’s head of sporting management explains to Raceteq that the virtual platform can be run locally by teams and they have free rein to practice on fictional circuits or Yas Marina Circuit in their own time - even hosting their own races against other teams as one would on a racing game or simulator.
And that’s not including the fuel, tyres, and mechanics that A2RL supplies.
A2RL showed its autonomous machines to the world in 2024 as its Yas Marina season finale was broadcast live around the globe. The cars struggled to complete a lap of the circuit, sometimes stopping in their tracks when another rival was in the vicinity.
A virtual rendering of the A2RL AI cars at Yas Marina Circuit
He explains: “So the sim is meant to shorten that learning curve… it's helping the team spend more time to go slowly and understand how other vehicles behave on track because every algorithm is different. It's like every driver is different - they have different braking styles, they have different levels of risk taking, some are more aggressive, some are more passive.”
“It provides a low-cost and accessible route into the sport - similar to how karting feeds into Formula 1. Eventually, we would like to open the competition to a wider community of participants, including fans, independent developers, and aspiring teams worldwide. It would provide a structured pathway for talent to enter the league and contribute to the future of high-performance autonomy.”
Fireworks light up the sky at the inaugural A2RL season finale in 2024
How has A2RL changed in its second year?
The A2RL car is a 2023 Super Formula SF23 car, capable of hitting 200 miles per hour, but it didn’t hit those speeds in 2024. Another series, the Indy Autonomous Challenge, holds the speed record (as of August 2025) for an autonomous vehicle: 197.7mph, set by a Maserati MC20 at Florida’s Cape Canaveral in March 2025.
That includes new bespoke technology that can better handle the G-forces, heat, and vibrations on track.
The Indy Autonomous Challenge’s record-breaking AI-driven Maserati MC20 at Cape Canaveral
Right now, every team has the same hardware, but that could change.
“Teams will be able to make set-up adjustments - such as aerodynamic balance, tyre pressures, and other race-critical parameters - just as they would in traditional motorsport,” says Roles.
With one lap of Yas Marina Circuit generating 200 gigabytes of sensor and telemetry data - and a single full-day test capturing up to 24 terabytes of data - there’s lots to be learned from running an AI machine on track.
.jpg?cx=0.5&cy=0.5)
