Designers of self-driving cars are putting their creations on racetracks to learn more about how to improve them.
We're still years away from seeing self-driving cars dominate the highways, but already some of the chief minds behind the technology are setting autonomous vehicles on racetracks to compete against one another.
At first glance, it may seem like a rather puerile use of the tech, but in fact, having driverless cars speed around a tortuous circuit is one of the best ways to figure out how to improve them, even for navigating the snoozy avenues of the average suburb.
For now, at least, autonomous car designers are mainly interested in testing their developments on racetracks, rather than actually racing the cars, because the miles of open road allow for experimentation that's generally out of the question on public streets. The first and biggest event so far was the Autonomous Track Day, held at California's Thunderhill Raceway last month.
In an interview with Translogic, organizer and Silicon Valley entrepreneur Joshua Schachter called the event a unique opportunity to "try out new things in a less strenuous environment than the street, with pedestrians and traffic and that kind of thing."
Among the attendees was George Hotz (best known for being the first person to carrier-unlock the iPhone), who was there to learn the limits of his autonomous 2016 Acura ILX, which Motherboard showed in action last December. The racetrack afforded the opportunity to learn how to improve his car's performance on sharp turns and incidents requiring quick adjustments.
"If you can understand how to race, you can also understand what to do in a lot of evasive safety critical situations," Hotz says in the Translogic video.
Importantly, though, the cars at Thunderhill generally weren't learning the routes entirely on their own. Instead, a human driver completed a few circuits around the track in the car as a person normally would, and then the cars used that data to create an optimal route. As Robert Hambrick of AutonomouStuff said, they took the data from their driver's runs and "ran it as fast as we could without the human driver."
The "racing" at Autonomous Track Day amounted to speed trials at best, rather than an actual competition. Schachter was doubtful that a traditional race with autonomous cars would be entertaining, telling Translogic that "If it's just robots driving perfectly, that's not exciting."
But the International Automobile Federation (FIA) certainly thinks otherwise. During the 2016-2017 Formula E championship beginning in September, the organization will host a "Roborace" at each of the circuits used for the main championship, featuring real-time competition among ten teams.
The Roborace pushes the approach seen at Thunderhill to its limits. Instead of wildly different cars, each racer will have the exact same car to work with—a fully driverless model powered by Nvidia's Drive PX 2 supercomputer. Each of the cars will have the track data, but the teams will show their mastery by honing the artificial intelligence to calculate the optimum speed for curves in order to achieve victory. We'll have to wait to see how that works in action, particularly when the cars get close enough for collisions.
In the meantime, ogling the Roborace cars is fun in itself. The slick vehicles, pictured at the top of this article, are designed by Daniel Simon, known for his vehicle work in movies like Tron: Legacy and Captain America: The First Avenger.
"My goal was to create a vehicle that takes full advantage of the unusual opportunities of having no driver without ever compromising on beauty," Simon said in a statement to Formula E. "Racing engineers and aerodynamicists have worked with me from the beginning to strike that balance … The Roborace is as much about competition as it is entertainment."
But beauty or no, would it actually be entertaining? Even with fast, sleek physical cars in play, this removal of the human driver could take away some of the thrill. At the Roboraces, we'd still be ultimately be watching a robot do its thing.