On the Road to a Driverless Future

On the Road to a Driverless Future

When will we see fully autonomous vehicles on the road? And who will be making them?

In 1885, inventor Karl Benz took the world's first automobile out for an inaugural spin -- and promptly crashed it into a wall. Chris Urmson, until recently the head of Google's self driving project, used that fun fact from history as the icebreaker for his March 2015 TED talk, "Google's Vision for Driverless Cars."

Ever since that day 130 years ago, Urmson told the crowd, car manufacturers have been "patching around the problem" of human error in automobiles by introducing piecemeal improvements, from seat belts to driver-assist technology. He likened automakers' approach to driverless vehicles to saying, "If I work really hard at jumping, someday I'll be able to fly."

Google, on the other hand, has a nobler mission, Urmson essentially said: disrupting the inefficient, lumbering automotive industry by creating "fully self-driving cars" and realizing "what they can do for the world." Among the altruistic ends Urmson saw: relieving Americans of 6 billion unproductive minutes per day spent sitting in traffic, helping blind people become more independently mobile and saving the equivalent in lives of one 737 airplane falling out of the sky each day.

No one's arguing that Google doesn't have vision in spades with its friendly, R2D2-like driverless car (aka the "bubble car"). But autonomous technology is still very much a work in progress. For instance, the LIDAR remote sensing technology that Google and others use to provide a 360-degree view of any obstacles around the car doesn't work well in inclement weather.

Fully autonomous vehicles that are capable of operating under all conditions--that's probably not going to happen until at least sometime in the 2030s.

— Sam Abuelsamid, Navigant

"The images produced by those systems as they scan around, if it starts to rain or snow, that all turns into noise," says Navigant's Sam Abuelsamid, a mechanical engineer-turned-industry-analyst who worked on control systems for General Motors and Ford. "The signals bounce off of every raindrop and snowflake, and the sensors basically become blind at that point. One of the reasons manufacturers have been doing most of their development of autonomous systems in California is they don't have to deal with the weather out there."

Some predict autonomous vehicles will be out on the road full force by 2025; for others, it's 2030 and far beyond. Vehicles with no steering wheels and no pedals "are a long way off," says Abuelsamid. "We may start to see those vehicles in the 2020 time frame in certain restricted environments -- college and business campuses and perhaps some central urban areas [like heavily congested Paris] where regular vehicles are not allowed. But in the longer term -- and it's going to be a lot longer -- fully autonomous vehicles that are capable of operating under all weather conditions, all driving conditions, that's probably not going to happen until at least sometime in the 2030s."

Google's Chris Urmson (R) shows a Google self-driving car to U.S. Transportation Secretary Anthony Foxx (L) and Google Chairman Eric Schmidt (C) at the Google headquarters in February 2015 in Mountain View, Calif. Justin Sullivan/Getty Images

Urmson's comments aside, automakers (and auto suppliers like Delphi and Continental) are making their own gains. Among the OEMs, the frontrunners in autonomous vehicle technology appear to be Audi, Mercedes-Benz, BMW and General Motors. Ford and Toyota are in the middle somewhere, and Fiat Chrysler is at the back of the pack, still licking those bankruptcy wounds and counting on Jeeps to bring in profits. Academic institutions are part of the race, too -- Carnegie Mellon University, Stanford University, Ohio State University and University of Michigan are all testing or developing autonomous vehicle technology.

The biggest vote of confidence for the auto industry's efforts recently came from none other than Google. In September, Google brought in John Krafcik, a Stanford trained engineer and former CEO of Hyundai Motors America, to direct its self-driving car project. Krafcik now is technically Urmson's boss, and Urmson has shifted from project lead to technical director.

No matter how long it takes to get to driverless cars, one thing's for sure: Getting there is going to be a fun ride.

Not That Far Apart

Googlers (that's slang for Google employees) have been tooling around the Santa Cruz mountains in their autonomous bubble cars since 2013, past palm trees and sunsets, a Jonathan Franzen novel in one hand, perhaps, and a venti latte in the other. Auto industry research on driverless cars hasn't been nearly as imagination-capturing. Any venturing out onto public roads is happening quietly, with an eye toward minimizing risk and not tipping a hand to the competition. A ribbon cutting for a new testing lab, MCity, is about the most attention-getting feat. MCity is a necessary tool -- researchers need controlled environments to recreate conditions to discern what the variables are -- but it's geek without the chic.

Really, though, Silicon Valley and the auto industry aren't all that far apart, and collaboration between the two is already happening, says Richard Wallace, director of the Transportation System Analysis group for the Center for Automotive Research. GM, whose history of autonomous vehicle research dates back to the 1950s, has a pilot program in Shanghai where it's testing a fleet of vehicles, "which are in concept very similar to the Google bubble cars," he says.

OEMs have the complex manufacturing know-how and experience dealing with heavy government oversight and regulation, but are at a disadvantage "as far as the automated systems go," says Wallace. Tech companies can buy the chassis and body, the mechanical parts of the car, the power train. "They can buy kits and put their technology on top of it. That's where I wonder if they won't start with acquisition. Google or Apple could buy Tesla -- though neither of them seems to have wanted to. They could have bought Chrysler."

Wallace wouldn't be surprised "if at least one of the Silicon Valley companies takes a dive into being an automaker -- or contracts with an automotive supplier." Yet the OEMs and the tech companies "do each have their respective sweet spots. Maybe collaboration is the best way to go."

In early September, Wallace said that Google and Apple were very close to partnering with auto companies, that they've "signed collaborative agreements with one or more automotive companies." Within two weeks, in addition to Google hiring Krafcik, an article in the Wall Street Journal reported that Apple, which had already begun cherry-picking some engineers from Tesla, had plans to triple the size of its electric car development team to 1,800.

Although tech companies are constantly merging and acquiring each other, Wallace has yet to see an auto company try to acquire some of the smaller and more promising tech companies. "It's probably happening in small doses. I think it might have to do with the Detroit auto companies coming out of bankruptcy. They don't want to be seen as spendthrift, and so they focus instead on their core business. The next few years you might see more surprising business arrangements come to bear."

Some collaboration has happened another way -- by tech companies raiding automakers R&D departments. In 2008 or 2009, says Abuelsamid, Google got serious about the self-driving project, bringing in many researchers who worked on the various Department of Defense DARPA robotics challenges and autonomous vehicles -- researchers from Volkswagen, GM and Stanford University, "a lot of really smart people who worked on the fundamentals of how to do autonomous vehicle control."

Abuelsamid expects to see more cross-pollination. "I wouldn't be surprised if at some point in the next 12 months, Google didn't open up an engineering facility in Michigan," he says. "And I would be shocked if they aren't recruiting a lot of experienced engineers from the auto industry in various locations, here and in California to join their program already.

Mercedes-Benz's autonomous research car, the F 015, reimagines the vehicle's interior as a "luxury lounge" with seats facing each other as the car drives itself.

If the tech companies decided to manufacture vehicles themselves, he speculates, they would do well to set up engineering facilities in Michigan in order to have better proximity to the automotive supply chain. It's just easier if they're physically located in the same area," he says.

He also wouldn't rule out a Google acquisition "of either Fiat Chrysler or perhaps Telsa. I think Tesla would be a better fit for Google just because they're focused on electric vehicles." And FCA, while CEO Sergio Marchionne has been itching for a merger with someone -- anyone, "is in a lot of market segments that Google probably doesn't' have any interest in, and they don't have a lot of strength on the electric power train area, which I think is something that Google would want. I think Tesla would be the better fit for a partnership or an acquisition."

Safety, and More Safety

Google's engineers, with the help of data collected from its self-driving cars, have long been working on  algorithms to avoid accidents in the most random of situations, including, according to Urmson's TED talk, a woman driving in circles in the middle of the road on an electric wheelchair, chasing a duck.

The point all the players "are reaching now," Wallace says, "is once you get past the fundamental problems of controlling these systems, then you start to get into what at first seemed minor details but in fact are actually more difficult problems to deal with in terms of making the systems robust to work in all different types of conditions."

Bastiaan Krosse, program manager for automated driving at TNO, an applied research institute in the Netherlands that is doing automated vehicle testing mainly around truck fleets in heavily congested areas, says the autonomous technologies are there for the most part.

Krosse believes safety -- and the public trust in it -- will be the biggest issue in autonomous vehicle acceptance. "It's more about 'It's 99% safe -- how do you make it 99.1% safe, and 99.9%? How do we deal with these exemptions that don't occur very often?' That's a big challenge."

Another challenge is around price -- the autonomous components are a luxury item, too expensive for your average car buyer, for now. "Everything you do increases the price," says John O'Dell, senior editor for Edmunds.com. "Do they charge full price for it right away, or do they subsidize it and hope that volume of scale will make their money back?" He's seen a study that estimated full autonomy to cost about $5,000 to $7,000 per car.

Google is proud of the fact that its cars have been driven (as of June 2015) fully autonomously for more than 1 million miles without causing an accident. That might sound like a lot, says Wallace. "But I think that can't be evaluated very easily yet. We get one fatal crash in the U.S. for every 100 million miles. They haven't even gotten to the point where they should start to expect one."

"Transitions are always messy and risky and you know someday in the future whatever year you want to call it, vehicles will all drive themselves and promises are great," Wallace adds. "In between now and then, it's small percentages and growing."

Hide comments

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish