Welcome to “Driving Detroit,” a new Roadshow series taking a high-fidelity look at the cultural and economic revolution under way in the Motor City. Exciting new mobility and transportation startups are appearing every day, and transformed, re-energized auto industry stalwarts are bringing new technology and fresh ways of thinking to bear in Detroit. Modern America was built on the back of Motown, and this city’s nascent industrial and cultural revitalization has all the makings of America’s greatest comeback yet. Join us as we take a closer look.
In theory, autonomous cars are simple to use. You hop in, buckle up, speak your destination and off you go, free to check your email, finish that movie or just watch the world go by. In fact, the latest self-driving prototypes are already remarkably close to that reality. But the road to our driverless future is infinitely more complicated, and the stakes couldn’t be higher. Even though we’re already mastering the technology, the intangibles are proving far more difficult. How will a self-driving car know what to do when a school crossing guard waves it through an intersection? Will it reliably detect a siren and flashing lights and know to pull to the shoulder for a passing emergency vehicle?
This isn’t about just convenience, it’s a life and death issue for motorists and pedestrians, as well as automakers, suppliers and the government. Unlike with most other new vehicle technologies, car companies can’t afford to introduce half-baked solutions to the market and iterate over time. Autonomous capability isn’t like a new navigation system or transmission, where a glitch might amount to an inconvenience. Even a minor flaw could trigger a potentially disastrous — and high-profile — failure. For each company introducing an autonomous vehicle, the technology has to be right the first time. As James McBride, founder of Ford’s autonomous-car program warns, “Coming to market early with an ill-conceived product could poison the well.” Said another way, if early self-driving cars yield accidents or even very unsatisfactory consumer experiences, it could turn motorists, and indeed legislators, off to the very idea of self-driving cars. Testing is key.
To work through these intangibles and more, the University of Michigan has created a $10 million mock urban streetscape, called Mcity. Located on U of M’s Ann Arbor campus, about 40 minutes outside of Detroit, Mcity is a linchpin in the resurgent region’s efforts to define the next steps in transportation, a 32-acre stake in the ground against the encroaching might of Silicon Valley. It’s a future-mobility petri dish, not only for the world’s automakers and their suppliers, but also for the lawmakers, academics, telecoms, city planners, tech companies and even insurance firms who are trying to envision what a landscape of self-determinate vehicles looks like.
Mcity isn’t just about providing a physical space in which to test connected and self-driving vehicles, it’s about cultivating an environment in which both problems and opportunities can be ferreted out and tested. It’s about attracting new engineering talent to Southeastern Michigan and to the university itself. It’s about trying to predict the effect that connected and self-driving cars will have on the economy and society. Perhaps most uniquely, much of the resulting data from these tests is being shared among all these participants with a vested interest in the future of the global transportation network, fostering a mutually beneficial environment for all.
An autonomous Ford Fusion prototype tests at MCity.
Opened just six months ago, Mcity, part of the university’s Mobility Transformation Center, has been designed to test every possible driving scenario to evaluate connected and autonomous cars. Be it a roundabout, a meandering gravel road, a railroad crossing, a tree-lined street that blocks GPS line of sight, or a high-speed freeway merge, Mcity’s four-plus miles of roadway allows testers to repeatedly test, observe and shape the behavior of connected and autonomous vehicles in a way that can’t safely be carried out with real-world driving. As U of M Associate Professor Ryan Eustice puts it:
We get to be ‘Maximum Evil’ to the technology within Mcity. When a car drives here, we can purposefully set up scenarios that are difficult corner cases that you find in the real world. Individually, you as a driver have to drive a lot of miles to encounter these types of events, but in aggregate, these events are pretty common.
There are so many things to think about: How will a self-driving car behave in a galvanized metal tunnel — will its radar and lidar sensors get confused by the reflective surfaces and sudden darkness? How will one perform in a whiteout snowstorm?
In fact, Michigan’s seasonality is a major advantage: You can’t test for all weather scenarios in Cupertino. Engineers have actually been rescheduling their Mcity test sessions to intentionally coincide with climatic events like snow and rain.
This 3D map depicts how an autonomous car’s sensors see the world.
Mcity’s roads, fixtures and buildings have been expressly designed to replicate as wide a variety of road environments as possible, including various road surfaces and lane markings, streetlights, even curb heights and styles. And the building facades themselves (which are made of real brick and glass) can be moved to be closer to the sidewalk or farther away to simulate different city environments.
The goal is to curb accidents by 90 percent, eliminating all those crashes caused by human error. Yet as it turns out, most experts agree that developing a self-driving car that doesn’t crash isn’t the hard part — it’s getting computerized vehicles with essentially ideal, 100 percent legal behavior to coexist harmoniously in an imperfect legacy road network populated by unpredictable, human-driven cars and equally unpredictable pedestrians.
Think of it this way: Self-driving cars can’t just be unceasing law-followers, dutifully riding the speed limit, coming to full and complete stops at every intersection with a ceaseless “after you” policy of motorway courtesy. That might work in a world of exclusively autonomous cars, but in today’s mixed driving environment, self-driving vehicles quickly become the outlier and a liability on roads where human-piloted interstate traffic flows steadily above the posted limit, or in an urban cityscape where cut-and-thrust driving is the accepted norm.
Street signs and lighting are all real-deal, Michigan Department of Transportation-spec pieces.
An October 2015 study by the University of Michigan readily illustrates the need for a facility like Mcity. Over the course of 1.2 million miles, prototype autonomous cars from Google, Audi and Delphi testing on California roads were involved in twice as many accidents as their human-piloted counterparts. This despite the fact that they have “thus far [been] driven only in limited (and generally less demanding) conditions (e.g., avoiding snowy areas).” Yet even though self-driving cars have suffered double the crash rate, none of them has been found at fault for these accidents.
Some of the incidents can be attributed to these self-driving cars’ relentless law-abiding nature disrupting the natural flow of traffic. To be safe and find acceptance, autonomous cars will need to adapt to not only the legal framework of a given transportation network, but also the expectations of the individual. Not all drivers want their vehicle to behave the same way — even if they aren’t actually behind the wheel.
Naturally, programming a car to keep up with traffic, even if it means breaking the law, triggers the inevitable question of responsibility should an accident or citation result.
As John Maddox, assistant director of the university’s Mobility Transformation Center, says,
With today’s transportation system, the whole concept is built around individual driver responsibility, and insurance follows that. What happens when you take the driver out of the loop? Who do you insure? Do you ensure the vehicle, the manufacturer [or] the person who owns the vehicle?
John Maddox, assistant director, Mobility Transformation Center.
Perhaps most of all, Mcity is about fostering an environment in which stakeholders can gather to ask questions about how a nation that moves autonomously changes not just mobility, but everything else. “The data is the most common thread amongst these companies from very different sectors,” says Maddox. “Everybody wants the data.”
And that data isn’t just for traffic analysis or vehicle communications. It’s about predicting how the cascading consequences of autonomous mobility are likely to disrupt the private-ownership and insurance models of today. It’s about how autonomy could exponentially increase car sharing, change parking systems and trigger a seismic shift in commercial trucking. Self-driving cars will almost certainly greatly increase the mobility of the elderly, the disabled and the very young, which will profoundly change the complexion of transportation. And that’s just for starters. Downstream, what will such developments mean for society as a whole?
For accurate testing, it’s important for Mcity to include every last detail, including “street furniture” like real mailboxes and benches.
This sort of potential has created an unprecedented pan-industry collaboration at Mcity. The site’s Leadership Circle partners — companies that have pledged an investment in the facility of at least $1 million over three years — gather on a monthly basis to discuss their findings. After all, automakers may be able to relatively easily expand their existing vehicle proving grounds to create Mcity-like environments, but they cannot re-create its interdisciplinary depth of perspective and data. As Andrew Christensen, senior manager of technology planning at Nissan’s Farmington Hills, Michigan-based Technical Center North America, says, Mcity has “allowed us to come together in a precompetitive area and start to have these discussions. Research projects have been initiated that were voted on by the group to try and touch on everyone’s interests moving forward.”
Mcity is also great news for the University of Michigan, one of the state’s biggest employers. With 15 founding members of the $1 million Leadership Circle and nearly 40 affiliate members who have each pledged $150,000 over three years to be involved, the $10 million facility figures to make for good business while burnishing the university’s engineering reputation and strengthening its existing research ties with long-standing partners like Ford and Nissan.
Just as importantly, as U of M’s Eustice notes:
A lot of this comes down to trying to track down the brightest and smartest people to work on these really hard problems. Having a facility like this that is unique in the world allows us to really excite a lot of the next generation of engineers and get them in here. It really helps us in recruiting that type of talent.
Ryan Eustice, associate professor, University of Michigan
The University of Michigan and participating companies we spoke with declined to give specific estimates on the job growth Mcity will bring to Ann Arbor and Greater Detroit, but Mcity’s presence will doubtlessly help attract top young talent in the field and leverage the region’s existing engineering and manufacturing expertise. It, along with deals between traditional automotive powerhouses and tech companies, will help Detroit further cement its role at the center of the future of transportation — and fend off the further advances of those Valley-based tech titans.
As Maddox notes, “The auto industry has built up a competence and an incredible confluence of activity around the verification and validation and design of automobiles. It’s not a simple task. Detroit — Southeast Michigan — is really the most comprehensive place in the world where all of this activity and expertise comes together.”
Mcity may be the first of its kind, but it’s unlikely to be the last. “We’ve even had requests to copy Mcity in Korea, China and Europe, so we think that others see the value of such a facility,” says Maddox.