Page 32 - EE Times Europe Magazine – June 2024

32 EE|Times EUROPE

Racing on Algorithms

here to compete with Formula 1,” Mitchell asserted. “We are using this
platform as a way to accelerate the testing and validation of hardware
and software, and to give students a learning lab experience that they
otherwise can’t get.”

TRIALS AND TRIBULATIONS
IAC progress has been impressive. Initial time trials between univer-
sity teams took place at the Indianapolis Motor Speedway in 2021,
with more time trials and two-car track racing following on other U.S.
oval speedways in Las Vegas and in Fort Worth, Texas. “When you’re
on an oval, you may never drop below 150 mph ... and path-planning,
overtaking—these vehicle dynamics are extremely hard,” Mitchell said.
“Can all systems operate and make decisions continuously? This is the
challenge.”
Simulations, which include details of the track and car, have featured
heavily during software development to test AI driver performance
before the real race. And along the way, university teams have broken
autonomous world record after autonomous world record, achieving The PoliMOVE team won the second annual Autonomous
highest land speed (192.2 mph) on the Kennedy Space Center runway Challenge at CES in Las Vegas in 2023, reaching a top speed
as well as top track speed (180 mph) and fastest on-track head-to-head of 180 mph—a new world record for autonomous speed on a
overtake (177 mph), both on oval tracks. racetrack. (Source: Business Wire)
In June 2023, the IAC veered from its usual oval circuit onto a
Formula 1 road track at the Monza Circuit in Italy for a single-vehicle sharply, and so you need to know within a couple of centimeters where
time-trial competition. This was one of the biggest challenges yet, said you are on the track. With the GPS [at Monza], sometimes the car didn’t
professor Sergio Savaresi, a leader of the Politecnico di know where it was within 10 meters,” he said.
Milano–University of Alabama PoliMOVE team, which holds the land To counter the intermittent GPS, university teams revised their algo-
speed and other records and recently added Michigan State University rithms to use a combination of data from RTK-GPS as well as additional
to its roster. cameras and sensors, including LiDAR, while training the software on
An autonomous race car uses its localization software, which largely digital maps of the racetrack environment. “This time, we were also
relies on real-time kinematic positioning with RTK-GPS sensors, to relying on visual localization and object recognition of lines at the
navigate safely as it pelts around the track. Savaresi and cohorts had curb, trees, walls—essentially developing map-matching [algorithms],”
mostly developed their localization software for the open-sky oval Savaresi said.
circuits, but on certain stretches of the winding Monza road track, with In the end, each team conquered the road track. PoliMOVE scored
trees, overpasses and tunnels, GPS dropped out. the fastest lap, reaching a top speed of 169.8 mph/273.4 kph, with TUM
“Being GPS-denied on some parts of the Monza circuit was difficult, Autonomous Motorsport coming second, hitting a high of
as it was like being blind,” Savaresi said. “It’s one thing to localize at 167.7 mph/269.9 kph. But issues remain. As Hoffmann pointed out,
50 km per hour in an urban autonomous vehicle, but doing this at “Velocity is quite constant on ovals, but on road courses we are decel-
300 km per hour is much more tough; any tiny error in localization can erating from around 250 kph to 50 kph within a few seconds ... These
immediately take you off the track and into a wall.” fast decelerations also make it more challenging to try and predict the
Hoffmann concurred. “To be fast, you need to take the corners behavior of other cars.”

PoliMOVE passes TUM during the competition at the Las Vegas Motor Speedway in 2022. (Source: Indy Autonomous Challenge)

JUNE 2024 | www.eetimes.eu