Indy Autonomous Challenge Is Giving AI a High-Speed Education

• The Indy Autonomous Challenge has been testing the minds of college students and professors since its first competition in 2021.

• This isn't your parents' version of their favorite racing video game coming to life.

• It's not just a race to see who can go the fastest, it's about teaching AI to predict what the opponent will do.

The Indy Autonomous Challenge continues to turn laps and heads as the only autonomous racing series in the world.

The series made its on-track debut at the Indianapolis Motor Speedway in 2021. Since then, the series has raced six more times—at Las Vegas Motor Speedway and Texas Motor Speedway in 2022; at Las Vegas and Monza, Italy, in 2023; and this year at Las Vegas and most recently back at Indianapolis on Sept. 6.

At the Indianapolis Motor Speedway, 10 teams comprising educators and students from 18 universities in North America, Europe, and Asia once again pushed the limit of autonomous transportation and artificial intelligence.

And, no, this isn't your parents' version of their favorite racing video game come to life.

One of the newcomers on board at this past week's Indianapolis competition was a group of students and faculty members from Michigan State University. The Spartans joined forces with students and faculty team members from Politecnico di Milano in Italy on the powerful PoliMOVE team—a four-time event winner.

This time around, PoliMOVE finished second in the 10-team field. And while it was a runner-up finish on the scoreboard to the University of Virginia's Cavalier Autonomous Racing—the first American team to win the Challenge—the experience was definitely a win for the newcomers from MSU.

Daniel Morris is one of the organizers of this year's Indy Automous Challenge program at MSU.

"One of the interesting challenges in autonomous vehicles is predicting what other vehicles are going to do," said Morris, associate professor of Biosystems and Agriculture Engineering and Electrical and Computer Engineering at MSU. "I think that's a problem we have when we drive. We want to drive safely, and that's when it's not just our vehicle on the road. There's a lot of other vehicles on the road, and to drive safely we need to predict where other cars are going to be—especially if they intersect with our trajectory."

That's where the Indy Autonomous Challenge becomes the ultimate test kitchen—it's not just a race, it's about teaching AI to predict what the opponent will do.

"That's our problem for the autonomous race car," Morris said. "How do we know where our opponents are going to drive so that we can avoid them or maybe block them—or drive safely in any case? We need to sense the other vehicles and get in the mind of the other driver or, in this case, the other vehicle."

MSU has been involved with the program for the past year. It missed out on competing at Las Vegas in January because the PoliMOVE car was being upgraded, but the team nonetheless got to experience the race day.

The Italian-led team tested the car mainly at Kentucky Speedway throughout the year. MSU students and instructors were able to view a few of the test sessions.

"A lot has happened in the last year," Morris said. "The teams have been refining their vehicles, and now they are a lot more reliable. We're still building up our team and assuming we're going to get more and more involved this coming year. We'll have a person working on the project full time starting in about a week."

The event itself consists of two parts—time trials and a head-to-head passing competition.

Time trials are just that—one car going out at at time and having seven minutes of track time to set its fastest lap. Cavalier and PoliMOVE finished 1-2 in the time trials, both amazingly topping 170 mph despite challenging wind conditions. PoliMOVE checked in at 170.651 mph and Cavalier a tick over 171.

The second level of the competition is two-car, head-to-head passing that really puts AI to the test.

"There's a lead vehicle and an attack vehicle," Morris said. "The lead vehicle is given a speed to run at, and the attacker vehicle's job is to pass the lead vehicle within one lap. Then they switch roles, and now the attack vehicle becomes the lead vehicle."

The back-and-forth starts at a modest 80 mph, and the speed is increased to 100 mph and then 10 mph increments thereafter until one car is unable to make its pass.

Once the car leaves the pits, AI calls the shots. The artificial intelligence programmed for the cars "drives" the competition.

"No, no, no," Morris said when asked if this is just some giant video game with students manning joysticks in the pits. "It's all fully autonomous. We're not controlling the vehicle at all once it leave the pits. Between each lap we can set parameters—tell it how much faster it needs to go—then we have to let it go all around the course on its own. Everything has been preprogrammed into it.

"You want to plan ahead how your car is going to race, both as a lead vehicle and as an attacker vehicle. And you don't know how your opponent is going to race. Your vehicle needs to be smart enough to adapt to whatever the other vehicle does—and smart enough to make it harder for the other vehicle to pass."

Morris said that there's a lot of excitement among students at MSU and other universities around the country to get involved with the Indy Autonomous Challenge. The PoliMOVE team's racing Dallara made an appearance on campus at MSU in May, and that really got people talking.

"We've got both undergrads and graduate students interested the program," Morris said. "We've got computer science students, mechanical engineering, electrical and computer engineering, quite a few on the wait list."

One of the students on board for MSU this year was Mk Bashar, a PhD student from Bangladesh who signed on to the team as part of his computer science engineering research. Bashar said that he wasn't really a race fan before the project but now has taken at least a little interest in NASCAR, Formula 1, and other forms of racing.

Here's a look at all the action from autonomous race day at Indy:

The biggest difference, obviously, between the Indy Autonomous Challenge and, say, F1, is that there's no drivers in the equation—only data.

"We're getting data from six cameras, other light sensors and radar sensors," Bashar said. "Sensors help us get a partition around the car for 360 degrees. Now our car knows where it is precisely, and also it knows the position of the opponent. So, now the AI is making decisions based on your opponent as to how you get past the other car."

Bashar said that more than once, opposing cars made what he described as "aggressive decisions" out on the Indy racetrack.

"In the practice session, one of the cars goes very close to our car and got involved in an accident," Bashar said. "Our car was repaired, and we were kind of surprised why other teams were taking these aggressive decisions in the middle of the race.

"The passing competition final was a very tough game—actually kind of a nerve-wracking competition. It was so amazing."

Bashar said that the entire experience and the technology involved is a look into the future of autonomous transportation. No one expects the Indy Autonomous Challenge to one day replace the 33 cars—and drivers—on the grid for the Indianapolis 500.

That's not the goal.

"The Indy Autonomous Challenge, the goal is pushing the autonomous driving technology to the next level," Bashar said.