It's been a tough road for driverless cars: We recently learned, for example, that all of the crashes involving Google's test-phase autonomous vehicles have been caused by humans. Which is one of the reasons experts have just opened up a testing center in Michigan that's trying to recreate the chaos of the human-built street in a controlled environment.
It's called Mcity: A $US10 million, 32-acre simulation of city streets, suburban roadways and everything in-between, designed by a group of researchers, government agencies and car companies in Ann Arbor, Michigan. Its moveable facades make it possible for researchers to rearrange any kind of conditions imaginable, from blind corners to odd intersections, all in the service of developing smarter autonomous vehicles. Gravel roads? Mcity has them. Paving brick? No problem. Freeway signs, graffiti, HOV lanes -- it's all recreated here, so that engineers and researchers can figure out how autonomous vehicles will react.
Mcity is the first major project of a part-governmental/part-academic/part-commercial partnership called the University of Michigan Mobility Transformation Center. With million-dollar investments from companies like Nissan, Toyota, Ford, GM, Honda, State Farm, Verizon, and Xerox, it's trying to build the testing infrastructure that will make varying levels of automation across cars more feasible.
That includes driverless cars, sure, but it also including testing systems we already knew UM was working on, like vehicle-to-vehicle connectivity that allows cars to "talk" to each other and adjust accordingly.
Teaching a driverless or connected vehicle about the world is surprisingly difficult because the world is unpredictable, and computers don't deal well with surprises. "Even seemingly minor details a vehicle might encounter in urban and suburban settings have been incorporated into Mcity, such as road signs defaced by graffiti and faded lane markings," the University explains in a release today.
It's actually these environmental eccentricities that are really challenging for driverless car tech. Sure, you can teach an autonomous vehicle to understand the way the road should look, 95 per cent of the time. What about when something unexpected happens?
UM isn't the only research centre trying to figure out how to teach a machine to react safely to unpredicted weirdness. At Google I/O this spring, Google X's "Head of Moonshots" Astro Teller, described some of the bizarre ways in which the company is testing its self driving cars, including throwing beachballs at it, buying fake birds and having them swoop down towards it, and perhaps most hilariously, having someone hide in a canvas bag in the middle of the street and jump out. Even when you work for the moonshot team at the most powerful technology company in the world, sometimes you just have to shut up and zip yourself into a bag.
Now that Mcity is open, odds are good we'll be hearing more about autonomous testing soon. Whether UMich will be using Google's patented bag-surprise technique remains to be seen.
[Michigan News; All images via University of Michigan]