One-by-one, driverless car test vehicles are already hitting the road, and soon, they'll start to appear by the dozen. The university town of Ann Arbor, Michigan, plans to up the ante further by putting the nation’s first networked, fleet of driverless cars—a full 2,000 vehicles—on the road within the next eight years.
The project, a collaboration between the city and the University of Michigan’s Mobility Transformation Center, is a large-scale test of how to integrate autonomous vehicles into the urban environment. It’s not about developing the advanced sensors and software that will guide driverless cars, as Google and carmakers are doing, but more about developing the communications and infrastructure needed to help these vehicles live up to their potential to change the way we drive.
"It would be addressing the question: What is the city of the future and how do we get from where we are now to there? This is a transition period," says Jim Sayer, the Mobility Transformation Center’s director of deployment. "No one is going to wave a magic wand, and this will all be deployed 100%, and everyone’s going to get a new vehicle."
The university has already started to make this a reality. For the last two years, Sayer has been leading a project called Safety Pilot that includes 2,800 volunteers from Ann Arbor who agreed to outfit their own cars with wireless radio communications devices that can "talk" to traffic signals at 25 intersections. The cars can also receive warnings when they are going too fast around certain curves.
By the time the 18-month project, done in conjunction with the U.S. Department of Transportation, is finished in a few months, Sayer estimates it will have collected 12 billion wireless transmissions from the volunteer vehicles—each broadcasting 10 times a second. This trove of data will help answer all sorts of questions, like how people respond to real-time warnings about their driving, how traffic signals can respond to the current traffic situation on the road. It will also offer insight into simpler problems, like learning the range limits that the wireless signals can reliably broadcast.
The data is needed, Sayer believes, because the Google approach of relying on sensors alone isn’t going to be enough. For example, in snowy weather or terrible visibility conditions—exactly when drivers need the most assistance on the road—sensors may have trouble "seeing" just like humans. By contrast, a broadcast device could communicate with the car ahead of it and know to give it some room. "We believe that we’re not going to reach a high level of automation without some form of connectivity," says Sayer.
The next phase in reaching the 2021 goal will be the construction of a recently approved $6.5 million driverless car test facility, spread out over 30 acres. The school will work with the wealth of auto industry R&D centers in the area, along with the state government, to put not just driverless cars, but driverless towns, to the test. There won’t be the traditional high-speed oval, says Sayer, but rather a continuous set of challenging everyday scenarios and environments to test the limits of safety in a driverless vehicle (and train the next generation of engineers to work on this kind of technology).
Michigan researchers think they can save lives if they pave the way for self-driving car technology. By 2021, the goal is to have a self-driving taxi fleet available that could be hailed by a smartphone. Taxis without the taxi drivers? As a New Yorker, I can certainly see why that might be safer.