How a 192-MPH Self-Driving Race Car Is Accelerating Autonomous Technology

Autonomous race cars are not the gimmick you might think they are. No one is trying to replace human drivers at the track or the mechanical horse races you find at the casino. Instead, the Indy Autonomous Challenge pits the world's top universities against each other in a race to accelerate the development of autonomous vehicle technologies. The decision to develop the technology with high-speed racing, as exciting as it is, tackles a specific and under-researched area of consumer vehicles. It is a monumental undertaking that requires full-time dedication from dozens of top researchers in the field, millions of dollars, government support, and the shared vision of a safer future. And although testing autonomous vehicles on a closed course is safer than using public roads, crashing high-tech race cars with six-figure price tags is still a major setback.

Humble Beginnings

In 2005, after lessons learned from an unsuccessful inaugural year (no one finished), the DARPA Grand Challenge shifted autonomous vehicle (AV) development into high gear with a competition that sounds straight out of a movie: a $1 million prize to the winning team of a 132-mile race through the Mojave Desert. Picture the plot ofRat Race, but the setting is the Mint 400 coverage fromFear and Loathing in Las Vegas. Admittedly, the Grand Challenge was not that. But if you're not a tech nerd, the impact it had on advancing AV development was as big as the impact that film obviously would have had on pop culture.

After being appointed director of the Stanford Artificial Intelligence Laboratory in 2004, Sebastion Thrun led Stanford's entry, Stanley, to victory in the 2005 race. Thrun was already recognized internationally for his contributions to the field of robotics and would go on to co-develop Google Street View, launch the Google Self-Driving Car Project (now known as Waymo), and co-found Udacity.

Mark Miles, president and CEO of the Penske Entertainment Corporation (Indianapolis Motor Speedway, Indycar, and IMS Productions) set Indycar on a path of growth after taking over as CEO in 2012. Looking to expand the impact of the resources at his disposal, he invited Thrun and Riley Brennan (executive director of Stanford's Revs Institute automotive research program) to brainstorm ideas at the Indy 500 in 2017. The race served as inspiration to solve edge-case highway driving scenarios by subjecting AV technology to the perils of open-wheel racing. The software in an AV mostly relies on making decisions based on data from situations that have predictable outcomes. Edge cases represent the unknown scenarios and require the software to make decisions on the fly like a human driver would—a capability still in its infancy. The Indianapolis Motor Speedway's history of innovation made it the perfect place to host such an experiment. It's credited with the first use of a pace car (1911), first use of four-wheel hydraulic brakes (1921), first mandatory use of helmets (1935), and first use of crash-data recorders (1993), and it pioneered the use of the SAFER Barrier in its four turns, introducing them to IndyCar in 2002.

See All 5 Photos5

From Mountain View to Track-Side View

With fresh thoughts of self-driving race cars swarming through the racing capital of the world, Miles secured the support of the state of Indiana and the funding to bring the idea to fruition. Thrun hosted a small gathering in Mountain View, California, to test the waters, and on May 23, 2019, over 20 universities and research organizations were invited to IMS for a formal exploratory workshop that would conclude with the 103rd running of the Indy 500. That last bit might be what initially got some participants there (who would turn down an invitation to the Indy 500?), but this meeting was the de facto event that led to the official beginning of the Indy Autonomous Challenge (IAC) in November 2019.

The plan was to develop a high-speed AV competition to serve as a successor to the DARPA Grand Challenge. Through this effort, the IAC would solve the edge-case scenarios of autonomous technology, innovate new technologies and methods, and advance public acceptance of AVs—after all, the societal challenges are proving to be just as big as the technological hurdles. A familiar and exciting racing competition is a way for people to witness the advancements without being directly involved if they aren't ready to jump in themselves.

If You Build It, They Will Come

Many of the May 2019 workshop participants had participated in the early DARPA competitions. The failure of the 2004 DARPA Grand Challenge was in part a testament to the state of technology at the time, but it was also widely attributed to rules that were too strict. This was a key lesson that the forefathers of the IAC would not forget. A lot of the buy-in was because of the opportunity to be involved in designing the competition and setting the rules. Being at the bleeding edge of research and development meant the universities knew the limits of the current technology, and letting them set the ground rules was a unique opportunity to design the competition for attainable yet still ambitious goals.

A one-off race at IMS was planned for 2021, with prize money and funding for vehicle development raised by the Indiana-based Lilly Endowment, one of the world's largest private philanthropic foundations. An open invitation was released, and at best, the IAC hoped for a dozen universities to respond. To everyone's shock and amazement, over 40 universities from around the world wanted in.

Without a vehicle to work with, the initial challenges focused on software development and simulation. Even if teams were confident enough to head straight to the track, the COVID-19 pandemic would have halted that effort in 2020 anyway. Cloud computing and a simulated IMS development environment by Ansys allowed teams to continue working while IAC looked for a chassis and hardware for the teams.

See All 5 Photos5

Project Deep Orange

In 2020, the IAC contracted the Clemson University International Center for Automotive Research (CU-ICAR) to lead the creation of a vehicle to host the software being developed by the teams. Through Project Deep Orange 12, CU-ICAR unveiled the Dallara AV-21 at the 2021 Consumer Electronics Show. Based on a Dallara Indy NXT IL-15 (known as Indy Lights until 2023) chassis, the AV-21 is equipped with six cameras, four radars, three lidars, and an RTK GPS controlled by an Intel Xeon E 2278 GE 3.30 gigahertz CPU with a single Nvidia Quadro RTX GPU, and 64 gigabytes of RAM to control the Mazda-AER 2.0-liter turbo I-4 and six-speed sequential gearbox.

Remarkably, that's commercial equipment available to the public. Gaming computers often have more memory and processing power. This highlights the complexity of modern software-defined vehicles (SDV). Over 40 teams of 10 to 15 doctoral students worked nearly full-time for a year and a half tosimulateautonomously piloting the 450-hp Dallara—a task typically handled by a single human driver. There's a reason our reality bears no resemblance to decades-old predictions from researchers or pop culture: The devil is in the details, and the details are endless lines of perfectly executed code replicating complex calculations we intuitively make in our heads just to walk around, let alone to replicate the intuition and reaction time of an IndyCar driver. Remember, the 2005 DARPA Grand Challengerevolutionizedthe state of AV technology, but in 2023 we're still a long way from freely roaming public roads with full autonomy.

See All 5 Photos5

Skin In the Game

The cars carried $ 1 million price tags, and although IAC donors subsidized some of the costs, the field narrowed to just 10 participating teams. Most of them involved collaborations between multiple universities both for brain power and financial support. Late in the summer of 2021, the AV-21 made its first laps at Lucas Oil Indianapolis Raceway Park, and by October the teams had their cars running single-car time trials and slaloms at speeds up to 140 mph at IMS. The IAC brought 170-mph head-to-head racing to Las Vegas Motor Speedway at CES in January 2022, and by April the same year, the winning team PoliMOVE (Politecnico di Milano and the University of Alabama) set a new AV land speed record of 192.2 mph at the Kennedy Space Center. Returning to Las Vegas for CES in 2023, PoliMOVE defended its 2022 win by reaching competitive speeds up to 180 mph, an AV speed record on a racetrack.

See All 5 Photos5

The Futureof the IAC

The competitions so far have been on oval tracks, where the cars are guided mainly by GPS. The radar and lidar sensors allow the cars to make the subtle moves required to compete in the defender/attacker-style racing where the cars run in pairs, reminiscent of the "battle" portion of a drift event. But instead of following the leader as close as possible, cars try to make a pass, and the winning car is the one that overtakes the defender at the highest speed. Through these competitions, teams have begun to master high-speed maneuvers to the point where cars are able to avoid crashes and maneuver out of a spin—the exact scenarios that can't be safely studied and developed on public roads.

The AV-21 will make its road course debut for an exhibition race at the Monza F1 circuit during the Milan Monza Motor Show (MIMO) in June. With sections that cross under bridges and out of GPS communication, the nine current teams have their work cut out for them. But guiding cars out of a dependance on GPS and documented roads is the path from Level 2 to Level 4 autonomy, the current focus of automotive manufacturers and regulatory legislation. As of 2023, only Level 2 driver support systems are on sale across  the United States, requiring drivers to keep their attention on the road and intervene when necessary.

See All 5 Photos5

After MIMO 2023, the IAC, in a new partnership with Continental and their fifth-generation 4D ARS540 Imaging Radar, will launch the updated AV-23 at Las Vegas Motor Speedway during CES 2024 and hold an official competition at Manza.

In production since 2021, Continental's 4D imaging radar has almost 1,000 feet of range (300 meters), and can determine an object's distance, direction, relative velocity, and height to assemble an image of the surveyed environment. Previous automotive systems with 3D radar could not detect height, which makes identifying specific objects challenging. The ARS540 is currently used in production cars in Germany and will make its debut in a North American-market vehicle this year. The challenges the IAC faces with increasing speeds and GPS dependance will allow Continental and the next generation of AV researchers to safely develop Advanced Driver Assistance Systems (ADAS) off public roads while putting on a show that's far more entertaining than watching a minivan navigate suburban strip malls.

See All 5 Photos5

Images provided by Indy Autonomous Challenge and Penske Entertainment