A2RL Suzuka Crash Reveals the Limits of Autonomous Racing - But For How Long?

Alex Harrington | Dec 29, 2024

A2RL team posing for photos at Suzuka / Lydia Mee

What is it that drives our passion for motorsports? Is it the raw speed? The symphony of engines and the unmistakable scent of race fuel? Or is it the 60 kilograms of human courage sat in the cockpit, risking its life for the sake of our entertainment? Motorsport has always been a celebration of man and machine, but what happens when man is removed from the equation?

I went to my first F1 race in 2017, a season defined by the fierce rivalry between Lewis Hamilton, who at the time had three championship titles to his name, and four-time champion Sebastian Vettel who was fighting for his life in a Ferrari. I was perched on the sun-warmed concrete of Luffield Corner, surrounded by fans draped in Union Jacks listening to the cheers for the Briton evolve into boos and hisses when the Scuderia shot by.

When Vettel lost it at Luffield, the jeering was louder than the car itself. Maybe, in that case, it’s not about the cars or the drivers specifically; instead, maybe it’s our deep, childlike obsession with the concept of heroes versus villains.

I'm afraid this question doesn't have a simple answer, but it's something the Abu Dhabi Autonomous Racing League is very interested in exploring.

What is the Abu Dhabi Autonomous Racing League?

A2RL autonomous car waits in the pits / Alex Harrington

You may already have heard about these guys from earlier in the year when they unveiled their ideas to the public for the first time at the Yas Marina Circuit. Four autonomous cars, each built atop the same Dallara Super Formula chassis, set out on the league's first-ever autonomous race.

It was, simply put, a disaster, as the cars struggled to show any form of pace let alone strategy or racecraft. What it did do, though, is show us a project with potential.

The Abu Dhabi Autonomous Racing League, or A2RL for short, stems from Aspire, a company whose main aim isn't to create robot race cars but rather to develop the best and most innovative technology the world has ever seen by utilizing the expertise of the most intelligent minds from across the globe. This, unsurprisingly, comes with a rather large price tag. But there are plenty of UAE investors willing to ensure the country doesn't run out of money when oil is no longer the commodity it currently is.

Autonomy is just one of several development clusters they're building. Their other focuses include healthcare, food and agriculture, aerospace engineering, and of course, sustainability. In fact, A2RL acts simply as a marketing vehicle to get their technology in front of more eyes. There is a business model behind it, but we'll get to that later.

The sport alone involves 9 countries, such as the USA, Japan, Germany, France, and the UAE, and wants to grow rapidly. If they're to do this, then they're going to need to break some eggs. And that they did.

A2RL Suzuka Man VS Machine Showdown

A2RL had two cars in Japan / Alex Harrington

I attended their latest showing at Japan's famous Suzuka Circuit where I was able to chat with the team, get up close and personal with the cars, and watch what was supposed to be the ultimate showing of Man versus Machine.

For now, let me tell you about the machine portion of this equation.

The Cars

The Autonomous stack of the A2RL cars / Alex Harrington

Each vehicle is built on the Super Formula Delara EAV24 chassis, much of which is identical to the cars that ran throughout the race weekend at Suzuka which was also hosting the Super Formula finale. One difference is the power unit: a turbocharged K20C1 internal combustion engines which has been turned down a little and in this form, produce around 550 horsepower.

The main difference is what's sat in the cockpit. Interestingly, very little has been done to the car to allow it a level of autonomy. Instead, a computer, weighing in at about 90kg, directly replaces the driver. This 'autonomous stack', built by TII (Technology Innovation Institute) connects all the extra paraphernalia that helps it not only learn a track, but drive it at a level close to race pace.

The stack consists of seven Sony IMX728 cameras providing 360-degree coverage, four ZF ProWave radar sensors, three Seyond Falcon Kinetic FK1 Lidar devices, and a Neousys RGS-8805GC high-level computer with a powerful GPU. This equipment works together to create a perception system that allows the car to 'see' and develop situational awareness. The cameras provide visual input, the radar sensors detect speed and distance of moving objects, and the Lidar devices create detailed 3D maps of the surrounding area.

The suite of sensors sat on the cockpit / Alex Harrington

All this data is processed by the onboard computer, which uses algorithms to make real-time decisions on steering, braking, and acceleration. The autonomous stack also includes a drive-by-wire system that replicates human inputs using actuators for steering, braking, and gear shifting.

How this stack processes the data is down to each team, and this is where the competition of the racing league resides. Khurram Hassan, the Director of A2RL, tells me poetically, "the culture can be seen in the code".

Saturday's Demonstration

Saturday saw both cars on track. / Alex Harrington

Getting to work on Friday, myself and a few other journalists were able to take a closer look at the cars. But Saturday saw the machinery take to the track on what was a mildly chilly Japanese day. I had two takeaways from this quick demo. Firstly, the team didn't have a lot of confidence in their cars as they turned the wick down to reduce the chance of them meeting a sticky end. After all, the main event was Sunday, where one autonomous car would race former F1 driver Daniil Kvyat - someone who has been working with A2RL from the start of the project as the bridge between racing driver and computer.

My second takeaway was a severe limitation of the car and its software. Despite it collecting and analysing 1 TB of data per minute, including data on its suspension positioning, tire slip, and rubber temperatures, it was unable to swerve back and forth to heat up its tires. The cars also opted not to take a page out of Kimi Räikkönen's book and rely on its brakes to heat the tires from the inside out. This lack of tire temp' would prove to be more of an issue on the Sunday.

Dr Giovanni Pau, chief of the TII team, admits that swerving to get some heat in the tires is much harder than it looks. “It looks simple for a human, but for an AI driver, it’s really crazy," he admits to us.

From the balcony of the media room, the cars looked slow. Their four-cylinder power units screamed down the straight, but it was evident that they weren't at full throttle just yet. And with it still pulling data from the track in a mission to optimize its racing line, the corners were taken at a snail's pace in comparison to the drivers who had gone out before it. "Sunday will be faster," I was told, as the team watched data cascade down the multiple screens they had setup in their garage.

Sunday: AI Versus Human, or is it?

Sunday was to be the highlight of the weekend, and Russian driver Kvyat was in the pits assessing his car ahead of the race. There was a distinct chill in the air, but we'd been told the day before that there is a big red stop button that the team can press if they see something go wrong. See, the car knows what it's going to do a couple of seconds ahead, so if the team sees something they don't like in the code, or if something goes wrong with the car itself, they can slam on the brakes to minimize any damage ahead of it going wrong.

But, like any other form of motorsport, a couple of seconds is a lifetime.

I'm interested in getting a professional driver's take on the project. Kvyat seems mostly at peace with the idea as he struggles with jetlag, but he admits to me that this isn't going to be the race we all expected it to be.

"The AI can't 'see' me," he tells me and Anthony Cuthbertson from the Independent. He'll have to take a wide berth then, but the team has been more than open about the performance of their cars. At this point in development, even if it could track Kvyat's car, there's no chance it can beat a human in a similar machine. Pau, the team principal tells us it's 90% there, a figure I can believe.

The cars take their place on the starting grid, but minutes go by and there's little movement. The tires were warmed to their optimum temperature., but with every second that goes by the rubber loses heat and any remaining grip slips away as we watch the track officials ready themselves. Finally they begin and Kvyat hits the accelerator pedal. Even then, the autonomous car remains still, and it's a good ten to twenty seconds before we hear its engine breathe some life into its gearbox before it disappears from view around Turn 1 on what we could class as a formation lap.

We watch on the monitors as it slowly navigates the track, not yet reaching race pace. It's a narrow course, but from what we'd seen the day before, they're very able to make their way round in such a scenario. I'm pulling focus on my camera to capture the cars as they turn the final corner before the straight, but look up as a chorus of voices erupt around the track.

The autonomous car had crashed. I rushed to the monitor to see the aftermath and there's one thing that's clear: this was the end of the weekend for A2RL.

While the actual incident wasn't filmed, I was told that the cold tires matched with a rapid loss of pressure in one of the rears forced a spin, and despite being armed with a big red button, it happened so fast that the team watching couldn't do anything to stop it. Kvyat returned to the pits followed by a recovery truck carrying the carcass.

So what's next for A2RL?

According to Hassan, the racing league is a year away from not only putting in qualifying laps fast enough to rival that of a driver, but also operate with a level of racecraft that could see them compete on a similar level, too. I don't share his optimism.

Driving fast is one thing, but understanding when to lunge, when to pile on the throttle, and when to pull back and save yourself is a whole different kettle of fish.

Hassan tells me something I very much believe, however. One possible future of this story lies in gaming, or maybe a mix of gaming and real-world racing depending on whether the team can solve the obvious latency problems. They already have VR software setup to watch from the car's 360-degree camera feed. It's very impressive. But with what they're learning here, this data can be fed directly into the likes of Gran Turismo to improve the realism of the virtual sport.

Interestingly, someone with a Polyphony Digital badge was also walking around the pits with us.

The Verdict

What do I take from this weekend? It took me a long time to uncover the angle of this story. Was it the crash, the teams, or the tech? No. I don't believe so. I believe the story resides in the spirit of racing. The intangible element that makes motorsport what it is. By replacing the driver, does A2RL remove the heart of the beast?

It's too early to tell. Only when the cars can match every other variable - the leading duo being performance and raceraft - will we be able to fully understand what's missing. So I return to what I believe motorsport boils down to: heroes versus villains.

When the robot crashed, the crowds took a sigh of relief as their hero pilot Kvyat kept his hands tightly on the reigns of the sport defined by its drivers just as much as their machines. A2RL aims to to pit algorithms against instinct. Code against courage. But humanity, it seems, has won this round.

But for how long can good prevail over evil?