Just under the cloud-flecked Ann Arbor water tower, a tiny city is being built. Curving asphalt streets are the only hint of the project, but once the asphalt hardens, buildings will be raised, stoplights strung up and simulated pedestrians will be programmed. These 32 acres at the edge of North Campus will be home to the Mobility Transformation Center’s Test Facility, a complex built to develop automated, driverless vehicles.
Scheduled to officially open in Spring 2015, the center’s test facility is one of the first of its kind.
“In actuality, there’s probably nothing else quite like it in the world,” said Jim Sayer, the MTC’s deployment director in an interview with The Michigan Daily.
The University has quietly been moving itself to the forefront of the study and development of automated and connected vehicle technologies. The two terms are similar, but distinct: Connected vehicles will warn the driver if traffic is slowing down up ahead, while automated vehicles take that information one step further and actually decrease the car’s speed.
Cars are “connected” to other cars and to infrastructure through wireless signals. Traffic lights, for example, might constantly be broadcasting whether they’re at green, yellow or red. The connected car would then let the driver know that it’s not worth slowing down because the light is about to change. The automated car would hit the brakes.
A fleet of nearly 3,000 connected cars, trucks and buses were deployed for use on Ann Arbor streets in 2012 but testing the full potential of this technology has proved tricky.
Sayer said that first deployment made it clear that there was “a lot of movement working toward automated vehicles,” but that there was no good way to test them without raising safety concerns.
Automated cars not only react to other connected vehicles and infrastructure, but also sense their surroundings with the help of cameras, lasers, radars and other image sensing technologies. These are constantly surveying and calculating so that a vehicle can react in unexpected situations, like someone suddenly crossing the street. But testing to see if the car will do that reliably isn’t something that can be done at the intersection of State Street and South University Avenue.
Two years ago, when Sayer and colleagues were sketching bare-bone plans for the facility, researchers in the field were closing down parking lots on campus to test their technologies.
Engineering Associate Prof. Ryan Eustice said his work on autonomous vehicles for the Ford Motor Company allowed him to use their test track. He said the industry test track looks like an “asphalt lake”— flat, without anything for an automated car to recognize and react to. Eustice said cars need to be tested to their limit, so the testing space should be one where a car can safely spin out when they reach it. Researchers lacked a way to test automated vehicles in real world situations, without actually putting them in the real world.
So Sayer and his team started planning, trying to figure out how to pack as many variations of traffic landscapes, signals and scenarios into 32 acres as possible.
“When folks come out and see it, they’re really surprised by how much we really did think about,” Sayer said. “How much we were able to cram in and the level of flexibility that’s really going to be able to be afforded.”
Sayer said the Mobility Transformation Center worked closely with a committee that included researchers from the College of Engineering, like Eustice, as well as representatives from the Michigan Department of Transportation, like Matt Smith, program manager for the DOT’s Intelligent Transportation Systems program. The group came together to create a “wish list” of roundabouts and traffic signals, which was then handed off to a design firm that worked through the details.
“It was a very iterative process because we started off with just kind of a white board so to speak,” said Smith. “An open slate with whatever you want and heck, by the third or fourth iteration of the process — lo and behold you had a pretty detailed design of what the facility is going to look like.”
Sayer said that design combines common driving scenarios with anything that would be “particularly challenging” for connected and automated vehicles to deal with: roundabouts, unpredictable pedestrian crossings and railroad grate crossings. Traffic signage is one of the most difficult things for the technology to contend with simply because of the amount and variety of postings.
“Is it mounted on a pole that’s on the opposite side of the road that you’re trying to cross? Is it strung across the roadway diagonally? Is it mounted on a pole right at the stop-bar where you’re told to stop?” Sayer explained.
The challenges are compounded by the fact that traffic signals are sometimes written right onto the road or whipping around on a particularly windy day. Sayer said that putting this facility in Michigan meant that vehicles would also have to deal with all kinds of inclement weather.
The MTC worked closely with the Michigan Department of Transportation to ensure the model city would be representative of common traffic situations and ensure all elements of the facility would be built to roadway standards. This meant following MDOT’s construction codes, though not necessarily the most up to date ones.
“We had to kind of convince the architects that we did not want to do that,” Eustice said. “We actually wanted to build things that were not actually code anymore in some cases.”
Real roads have infrastructure built to all kinds of different codes. This historical layering is what gives a city its character. The MTC’s city, however, will be able to switch characters. Researchers will have their choice of multiple sets of signs — American or international, pristine or weathered.
“I got to spend time digging through a recycle bin at a local municipality salvaging signs that were no longer good enough for them to have on the street,” said Sayer. “Either because they had faded or they had graffiti all over them or they had been hit and were bent.”
Like a giant lego set, the only thing that won’t move are the streets — even the buildings can be adjusted. They aren’t up yet, but Sayer explained that they will essentially be giant theater facades. They might be crammed right up next to the street to simulate an old European city or placed 15 feet away from the curb like homes in suburbia.
“These are all things that automated vehicles are going to have to be able to detect in the future,” said Sayer. “Currently there is a lot of variation in the way this information is conveyed to the driver. The driver can differentiate it, the question is how well the automated vehicles are going to be able to detect and make sense of the variety of traffic control signal configurations and signs that are out there.”
That’s the question the facility hopes to help answer come 2015. For individual researchers, however, they’re just itching to get in.
“I’m one of the guys that’s been in these coned-off parking lots for the last couple years testing their stuff,” said Eustice. “I’m ready to be user number one.”