Speeding to Hockenheim: How students are building an autonomous race car
In a forward-looking move, the annual ‘Formula Student’ design competition recently introduced a “Driverless” category. With its eyes firmly on the prize, the student team from Munich’s Technical University is currently developing a car worth more than half a million Euros.
Marie Beller gleams with excitement - even though her schedule sounds more exhausting than that of many CEOs. “For Christmas, I went home to see my parents for a few days, but other than that I have spent pretty much every day this year in this workshop,” says the 24-year old. The workshop is a two-storey garage in Garching Forschungszentrum, the Northern research campus of Munich’s Technical University (TUM). It is the headquarters of TUfast, TUM’s team for Formula Student, a yearly engineering competition centred on a series of car races culminating in a final contest on Hockenheim race track in early August. “This year we will enter two separate competitions”, explains Beller, who heads the team of 63 students from Munich, “one for electric cars and the other for autonomous racers.”
100 metres in four seconds
In 2017, Formula Student introduced the new category “Driverless”. About 15 teams from different universities participated and this year sees that number almost double, as autonomous cars continue to boom. The competition will consist of three separate challenges. The first is Acceleration, in which cars have to drive 100 metres on a straight road. “Here we hope to reach the finishing line in less than four seconds,” says Beller ambitiously. The second contest, Skidpad, is a figure-of-eight racecourse. Finally, for the Trackdrive challenge, the cars have to do ten laps of an unfamiliar 500-metre course. “The latter two challenges are more about finishing reliably without veering off track”, says Beller. “But at TU in Munich we have always been willing to take chances, so we still aim to run the races at an average speed of 50 kilometres per hour.”
In order to reach such goals Beller’s team still has to improve the car’s understanding of its environment. To explain how difficult that is, she walks across the workshop, passing a drill press and a shelf scattered with electrical parts, to the test stand. Here, her colleagues mount the autonomous race car, which measures just under 2 metres in length, weighs only 175 kg and is worth more than half a million Euros. Stan Guerassimov, the team’s Technical Manager, is now installing a new system for the car’s perception of the environment. “In the field of autonomous driving many components are updated every six months,” he explains, “and our system will consist of two crucial components, which are both brand new.”
Leading with Lidars
The first is a Lidar, or laser scanner. “These sensors are controversial in the industry because they are so expensive that some manufacturers try to do without them”, says Guerassimov. “But one of the lessons of last year’s races was the approach of the winning team from ETH Zurich”. In contrast to its competitors – who used either no Lidar or several such lasers, each with a limited range of vision that required special software to combine all the data – the team from Zurich relied on a single 360-degree Lidar. It was built by a company based in Silicon Valley. “For this year we wanted that Lidar too, just like several other teams,” says Guerassimov. So his team sent a request to the producer in California and was successful: it obtained one of the Lidars as part of a sponsoring deal.
To complement the 3D-data produced by the Lidar, Guerassimov will further install two high-speed cameras. They film at frame-rates of up to 300 pictures per second and still produce high-resolution images. “The faster our car drives, the faster our camera has to check what is coming ahead”, explains Guerassimov. “Now the main challenge will be for our computer vision team to process all this data quickly enough.”
That is the responsibility of Lukas Wüsteney, a computer scientist and TUfast’s Head of Driverless Electrics. “I am working on so-called trajectory planning,” he says. “Basically, I try to prepare our car for as many situations as possible and develop reaction patterns so that the car will not have to assess each situation anew.”
But he is pressed for time. In just a few days, he and his teammates will lift their car into a large van and drive to a test track near Munich. Over the next four weeks his team then hopes to organize about 20 test runs, before they take off to their first race in Zalaegerszeg, a new automotive proving ground for autonomous vehicles in Hungary. “And we still need to do a lot of testing”, sighs Wüsteney.
Jockeying for the job fair
The opportunity to experiment with cutting-edge hardware and new algorithms is one of the factors that motivates him to forgo holidays and spend his weekends in the workshop. “Here we exchange ideas without any hierarchical boundaries”, he says enthusiastically. “So we try things that most manufacturers might not dare, because we almost want to make mistakes as that is how we really learn and progress.”
This experience however, also makes people like him attractive for future employers, explains Team Manager Beller with a smile. “To be honest, the event in Hockenheim is not only a car race”, she recalls from the previous year when she was already on the team. It is also an opportunity to personally meet the sponsoring companies of which TUfast itself has 150. “So the event is also a big job fair and a perfect ticket to our future job market.”