Story originally appeared on the Faculty of Mathematics website on June 23, 2023.
University of Waterloo students on a multi-school autonomous racing team achieved a personal best speed of 173.8 kph at a race on the Monza F1 Circuit in Milan, Italy.
The race, which featured five teams with members from universities around the world, was the Waterloo students’ fifth race, and the first on a road course rather than a banked oval track.
The Italian race is the latest competition in the Indy Autonomous Challenge, a collaboration between public-private organizers and academic institutions created in 2019 to challenge university students to invent a “new generation of automated vehicle software” in order “to run fully autonomous racecars.” Every team purchases the same multi-million dollar car – the Dallara AV-21 – and then competes to create the best software to drive it.
“It’s a software race, not a hardware race,” explains Connor Kirby, a BBA and BMath student on the team. “We’re not allowed to modify the engine or tires or anything like that.”
Programming a driverless race car
Racing team MIT-PITT-RW is made up of more than sixty students from the Massachusetts Institute of Technology, the University of Pittsburgh, the Rochester Institute of Technology and the University of Waterloo. It is the only team made up primarily of undergraduate students and is funded entirely through sponsorship. The students are split into groups, each focusing on designing and programming a different aspect of the car’s operation.
Ten current team members are from the University of Waterloo: Ishan Baliyan (bachelor's in Computer Science), Evelyn Campbell (bachelor's in Computer Engineering), Eidan Erlich (bachelor's in Mechatronics Engineering), Connor Kirby (bachelor's in Business Administration and Mathematics double degree), Ryan Larkin (bachelor's in Computer Science), John Liu (master’s in Electrical and Computer Engineering), Brian Mao (master’s in Applied Math and Computer Science), Jatin Mehta (bachelor's in Computer Science), Bilal Nasar (bachelor's in Computing and Financial Management) and Andre Slavescu (bachelor's in Computer Science).
Autonomous racing removes the reflexes, decision-making and vulnerability of a human driver from the equation. Engineers and computer scientists must program the race car itself, teaching it to understand its relationship to the track and other cars, operate smoothly at top speeds and make split-second decisions. During an actual race, teams can’t make any real-time interventions. They can only step back and hope their program works.
In the previous four races, which included a competition at the famous Indianapolis Motor Speedway, the team’s car competed on banked oval racetracks. A typical race would include two phases. First, their car — nicknamed Betty — would do a solo "time attack" test. Then, pairs of cars would participate in a passing race, in which one car — the “attacker” — would have to repeatedly pass the “defender” at increasingly high speeds. Prior to the Italian race, Betty’s highest speed in the overtake of another car, set this January, was 247 km/h.
Taking to the road
Moving from the oval racetrack to a road race posed dozens of new challenges, requiring almost three months of testing and practice on the course in Italy.
“We had never even taken a right turn before!” laughs Kirby.
One of the trickiest dilemmas concerned GPS. When a car went under a bridge or overhanging trees, it would temporarily lose its GPS signal, and with it its information about exactly where it was in relation to its environment. The team had to program their car for simultaneously localization and mapping — creating a live map of its environment while also tracking its location within it.
While team MIT-PITT-RW placed fifth in the competition, they are proud of how much they have learned and improved.
“I think we did really well this weekend,” says Ryan Larkin, Waterloo's team captain. “There were a lot of technical challenges going from oval to road courses, and so it was great to see us put down a solid lap time by the end of the event. We now also have a clear direction for futher improvement and I’m excited about what we can do next year!”
Beyond just being fun and exciting, Kirby explains, this kind of autonomous racing research has numerous practical applications.
“In the past hundred years, the racing industry has been providing innovations to the commercial driving industry — such as rear view mirrors — that have saved lives,” he says. “We’re trying to bring that kind of innovation in safety and performance to autonomous vehicles. Our car goes faster than a Tesla, so we have to make our software faster too. We can make cars safer.”