What if an autonomous vehicle could drive a racing car faster than a human around a track? That’s the question the Abu Dhabi Autonomous Racing League (A2RL) is trying to answer. The series had an abortive start during its inaugural race last year, but is continuing to develop its technology towards further competition in 2025. I talked to Stephane Timpano, CEO of Aspire, which created A2RL, about what went wrong before, and what the future holds.
Rapid Road To Autonomous Racing
“We had the first race in April 2025,” says Timpano. This was only a year after the decision had been made to create A2DL, which is phenomenal considering the technological challenges involved. “The race didn’t go as planned. But this was an R&D challenge. We knew that some things would go well and others not so well. We also took a lot of risks, because there are other organizations working on autonomous racing.” Aspire is part of Abu Dhabi’s Advanced Technology Research Council (ATRC), the country’s innovation agency. Aspire has a wide portfolio of futuristic projects, which includes grand challenges if which A2DL is the most audacious so far.
“No one else has pushed the limit of putting multiple cars on an F1 circuit at the same time,” says Timpano. “We knew that by doing this, we put the cars and the teams at risk of an incident. But what we learn from this kind of incident is very helpful, because it helps us to understand how to avoid incidents in the future.” A2DL has put this learning towards future races. “The eight teams we had last year are already confirmed for this year, and we have four additional teams joining, which is expanding our global coverage. This includes teams from the US, Europe, Asia and Japan. They are covering the entire planet today but also not only the R&D world, which is the origin of all these things, but industry as well. Some of these teams are in racing. Some are in autonomy, and they want to understand how this works in complex environments like Super Formula or Formula One.”
One of the A2RL teams comes from the Autonomous Lab at Khalifa University in Abu Dhabi, led by Dr Majid Khonji. “The A2RL 2024 race had three parts: timed laps, attack and defend, and multiple vehicle challenge,” says Dr Khonji. “With timed laps, you need to complete a lap as fast as possible. Attack and defend is between two cars, which overtake and defend their position. The multiple vehicle challenge extends that to more cars. This was our first time participating, and it was surprising how technologies fail at high speed. Things that work at low speed, fall apart when pushed.”
“We crashed our cars, actually twice due to localization errors, but we managed to achieve a speed of 173 kilometers per hour, which is respectable,” says Dr Khonji. “Our lap time was a decent if not exceptional two minutes 50 seconds.” A2DL uses a Super Formula single-seater car with 550hp compared to a Formula 1 car’s 1,000hp, so lap times will be higher even with a human driver. The latest Formula E car has 469hp. Khalifa University’s autonomous system is also accomplished software outside of racing. “Our object tracker is still ranked number one on Waymo and Kitti data sets,” says Dr Khonji. “These are benchmarks that compare tracking system algorithms.” However, A2DL is more about quick decisions than the sophisticated autonomy required by urban self-driving. “In the race, we are mostly trying to follow a path. We design the path offline. The AI is blind. We don’t use any fancy AI control theory. We also need to build the hardware so that if it fails, it can recover. That’s an area that these races would really improve. Given that you really need to be very fast and react quickly, you have less time for compute, so most teams don’t use fancy ideas in a race. The compute needs to be on the car.”
Expanded Horizons For Autonomous Racing
Khalifa University will continue to compete as A2DL expands. “Since last year, we’re now planning a multi-year engagement with multiple races,” says Timpano. “We went to Japan a few weeks ago in Suzuka for a demo. One car ended up on the wall, which happens in these situations. We have a new race in April 2025, which is going to be like the one we did last year but with four additional teams. This time, we’re going to have more of a show. We’re going to bring more human aspects to it. Last year we had Daniel Kvyat, a human driver racing against an autonomous car. At the time, we had a little bit less than 10 second difference between the two cars, which was incredible.”
Timpano thinks autonomous cars can do better, however. “We’re trying to reduce the distance between human and machine, and we’re going to have multiple cars doing that at the same time,” he says. “We’re adding complexity. We’re trying to push the R&D envelope on finding new ways of doing autonomous driving in a more efficient and precise way, at higher speed in more complex environments. The second element is building up trust. This is the most important word here for the public. We need people to start being convinced that autonomous solutions bring a big value to the mobility of tomorrow. This is going to change the cities as we know them today, because they are not sustainable anymore. Too many people are concentrated into small areas. You need to have people be confident to put themselves on an autonomous bus, and to put their kids in an autonomous car or taxi. If they see Super Formula cars at 250 kilometers per hour doing well, then you start building up trust.”
“Adding the human versus the AI is where the interest sits for most people, seeing former Formula One racing champions pitted against the machine,” says Timpano. This harks back to former Grand Challenges, such as when IBM’s Deep Blue beat Gary Kasparov at chess, IBM’s Watson won the Jeopardy gameshow, and DeepMind’s AlphaGo defeated the world’s number one Go player. “Daniel Kvyat had the courage to race against a machine where you cannot control its performance. A lot of drivers have declined to do this because they don’t feel comfortable having a machine racing a few meters behind them.”
A2DL is also considering systems to augment rather than replace drivers. “We could have racing vehicles with champion drivers assisted by technology to improve the performance,” says Timpano. “We’d all love to see drivers racing at 400 kilometers per hour but today we can’t, because drivers cannot do that in the cars. Tomorrow, maybe we can make that possible.”
Autonomous Racing Drones
Single-seater racers aren’t the only vehicles A2DL is working on, either. “The league started with cars, because it’s what people like the most, and it’s visible,” says Timpano. “But we’re also organizing the same kind of competition with drones. We have a partnership with the Drone Champions League, which is one of the major pilot drone leagues. With them, we have developed a drone into which we have put autonomous kits. We are inviting teams to participate. A Super Formula car costs $100,000 each, but an autonomous drone is a few thousand dollars so it’s easier to access. We had eight teams in the first car race. We have 135 teams that are trying to be selected for the drone race and 35 of them have already done the qualifications. We have secondary qualification in January and in April. We’ll have the first race with the best teams, where we have, at this stage, both research teams and racing teams together. They will race in Abu Dhabi.”
Even if the first A2DL race was a varied experience, Timpano foresees an exciting future for autonomous racing. “Ten years from now, we won’t be surprised if there is a league with autonomous cars and human cars, and maybe even on the same team,” he says. “A lot of things need to happen between now and then, with technology, security, safety, and performance, but we believe in the journey.”
