Federal officials say they intend to aggressively shape the emergence of driverless cars, increasing their role well beyond the traditional recalls of cars when they prove defective.
Transportation Secretary Anthony Foxx spelled out that determination Monday in issuing a long-awaited policy paper that details 15 points he expects automakers to comply with as they rush to put autonomous cars on the road.
The specifics previewed Monday ask manufacturers to document for the National Highway Traffic Safety Administration (NHTSA) how and where they expect their vehicles to operate, how they will interact with other cars and the roadway, how they validate their testing, how they intend to protect privacy and prevent hacking, and how they would share data collected by onboard computers.
Foxx said his department will work with automakers to “ensure that safety is appropriately addressed in the front end of development.”
“As technology races forward, the government could sit back and play catch-up down the road, or we can keep pace with these developments, working to protect public safety while allowing innovation to flourish,” Foxx said in a conference call with reporters.
Long awaited by automakers and state officials grappling with the issue, the federal government on Tuesday plans to lay out in greater detail its vision for revolutionizing the way in which people get from one place to the next.
“We are looking at what type of information related to events, incidents, crashes, et cetera, need to be recorded, and we’re actually looking at ways for that information to be shared so that you would provide a way for all autonomous vehicles to learn from those issues,” said NHTSA administrator Mark Rosekind.
An expanded federal role seems likely to diminish that played by states. For example, one issue is the question of who’s in charge.
Until now, states have regulated drivers — issuing licenses, requiring that they register and maintain their vehicles to state standards. The federal government has been responsible for auto safety, pushing for seat belts, air bags and backup cameras and recalling defective vehicles.
Google is testing vehicles that have neither a steering wheel nor floor pedals. Is there a “driver” in that car for states to regulate, or is there just a manufacturer who should be regulated on the federal level?
“Part of what we’re doing with this policy is saying when the software is operating the vehicle, that is an area where we intend to regulate,” Foxx said. “When a human being is operating that vehicle, the conventional rules of state law would apply.”
Foxx said federal authorities plan to engage automakers as well as members of Congress and other policymakers in conversation about explicit powers to measure vehicle safety before cars hit the road, a concept he described as “pre-market approval.”
“You have to remember that the United States’ approach to automobile-safety regulation is a self-certification approach, so automobiles today are supposed to meet our federal motor-vehicle-safety standards [and] the manufacturers self-certify that they do,” Foxx said in an interview after the conference call. “We’re putting out the idea that in this emerging arena, there should be conversation about pre-market approval, the idea that additional authorities could be given to NHTSA to evaluate a technology and to essentially have to approve its use before it’s put on the marketplace.”
With test vehicles already on the road in several states, and the app-based service Uber inaugurating commercial use in Pittsburgh, a critical piece of the federal guidance is aimed at keeping states from creating a patchwork of laws that might be at odds with other states’.
“Right now we have multiple states in the country — California, Nevada, Florida, Michigan, Pennsylvania and the like — each setting up their own rules and regulations governing testing and deployment of self-driving vehicles,” said Raj Rajkumar, a professor at Carnegie Mellon University who has been a pioneer in driverless car innovations. “This could be a nightmare, if you drive across a state border in an autonomous vehicle.”
In December, California released draft policy language that would require “safety certifications from both the manufacturer and a third-party testing organization” to demonstrate that autonomous vehicles are ready to carry the public, according to the state Department of Motor Vehicles.
Some carmakers have balked at such a “third-party testing” requirement.
Earlier this month, the Michigan Senate passed legislation that would explicitly allow self-driving cars on roads in the state. It also would allow “on-demand, automated” car networks to run throughout the state, like those envisioned by General Motors, Ford, Uber, Lyft and others. And the legislation would “prohibit a local unit of government from imposing a fee, registration, franchise, or regulation on an on-demand automated motor vehicle network through December 31, 2022,” according to a state legislative analysis.
“Some companies will conclude that existing law is perfectly flexible and accommodating for whatever they want to do,” said Bryant Walker Smith, a law professor at the University of South Carolina who heads the Emerging Technology Law Committee of the Transportation Research Board. “Some companies may be more conservative and want [legal] structures upfront.”
But just when those companies will want regulations in place also is open to debate.
“Developers of these systems will not want any legislation until they are pretty sure what technologies they are going to deploy,” Smith said, “and at that point they will have crafted proposed regimes that are most conducive to their particular vision, and will enthusiastically push for adoption of those particular regimes.”
Many transportation and technology industry executives are planning to deploy self-driving cars first in particular environments. One goal of the guidance is to bring some thoughtfulness to that process.
On Sunday, John Zimmer, president and co-founder of ride-hailing firm Lyft, offered a detailed account of how he sees the rollout occurring.
“At first, fully autonomous cars will have a long list of restrictions. They will only travel at low speeds, they will avoid certain weather conditions, and there will be specific intersections and roads that they will need to navigate around,” Zimmer said. “Hypothetically, Lyft could initially have a fleet of autonomous cars that completes rides under 25 miles per hour on flat, dry roads. Then, we could upgrade the fleet to handle rides under those same conditions, but at 35 miles per hour. And so on and so on, until every kind of trip can be completed by an autonomous car.”
In the parlance of Google, and now federal regulators, there are particular “operational design domains” the cars are designed to handle.
“We believe that fully autonomous technology will most likely be deployed incrementally, starting with locations that have distinct geographic, weather, roadway type, and other features, and progress to general deployment,” Google argued in comments to the federal government. “We believe that our fully self-driving vehicles should be able to successfully demonstrate competency in a variety of reasonably foreseeable traffic situations.”
There are a range of views as to how aggressive federal and state governments should be in overseeing these different domains, or categories.
The Department of Transportation opted to describe the document as “guidance” for the industry and state officials, both out of caution and for a practical reason. Even as most automakers are rushing toward completion of systems that will relieve drivers of many burdens, and eventually lead to cars that need no driver, there are myriad technological advances still to be made.
Federal officials worried the premature regulation of that technology might inhibit its development. They also realized the ponderous process to produce an official federal rule — it’s not uncommon for it to take five years — could leave them with something already outdated before it saw the light of day.
Finally, they are as aware as anyone that not everything developers see in their crystal balls will turn into reality, and the technological revolution of the past two decades will continue to leapfrog forward, creating unforeseen opportunities for the auto industry.
Another Analysis
Self-driving cars were a lot simpler when they were still confined to science fiction. But Uber’s recent rollout of semiautonomous taxis in Pittsburgh and the tough questions Tesla faces after a death involving its Autopilot feature are strong reminders that fact is quickly closing in on fiction. In an effort to keep the two from colliding, the U.S. Department of Transportation (DoT) on Tuesday made good on its promise earlier this year to give Silicon Valley start-ups and Detroit incumbents alike guidance to help them build safe self-driving vehicles.
The DoT’s new Federal Automated Vehicles Policy includes a 15-point safety assessment to help autonomous vehicle makers evaluate their technology in areas including data recording and sharing, vehicle cybersecurity, and how the vehicles interact with human drivers and passengers. The guidelines state that reporting these safety assessments to the National Highway Traffic Safety Administration (NHTSA) is voluntary, but could become mandatory as the DoT develops formal autonomous vehicle regulations.
The safety assessment is significant because it gives tech companies and automakers some much-needed guidance as they hurtle into a murky future. “The public is asking themselves if, once these regulations are followed, are autonomous vehicles going to be safe?” says Nidhi Kalra, a senior information scientist at research institute RAND Corp. and co-director of its Center for Decision Making under Uncertainty.
Right now the answer is “maybe,” Kalra says. Whether autonomous vehicle makers will ever be able to answer “yes” to that question is complicated—there is no test to determine whether an autonomous vehicle is safe nor is there even a consensus as to how safety is defined. “Some people will not accept mistakes from machines, even if those mistakes occur at a lower rate than they do with humans,” Kalra says. “Others will say that if the vehicles are safer than the average human driver, then we should allow them on the roads.”
The DoT’s guidelines clarify the roles that federal and state governments will have in regulating self-driving cars. The federal government will continue to be responsible for the safety of the vehicles themselves—including the software that allows them to operate autonomously—while the states remain in charge of regulating human drivers when they are at the wheel, according to Secretary of Transportation Anthony Foxx. At a press conference announcing the policy, Foxx also noted that the DoT will put in place an exemption process that companies can take advantage of if they want to depart radically from current vehicle designs—such as eliminating steering wheels and brake pedals—if they can prove that the new design improves vehicle safety.
In March the DoT acknowledged that current federal motor vehicle safety standards did not directly address automated vehicle technologies. The newly proposed regulations are an effort to remedy this problem, as carmakers and tech companies hurry to develop and sell new self-driving technologies. The main focus is on highly automated vehicles that can take full control of the driving task in at least some circumstances. Portions of the guidelines also apply to lower levels of automation, including some of the driver-assistance systems already being deployed by automakers today.
The DoT is walking a fine line as it tries to clear a path for autonomous vehicles while ensuring those vehicles meet the safety goals at the core of the agency’s mission, says Ragunathan Rajkumar, a professor of electrical and computer engineering in Carnegie Mellon University’s CyLab Security and Privacy Institute and a veteran of the university’s efforts to develop autonomous vehicles—including the Boss SUV that won the Defense Advanced Research Projects Agency (DARPA) 2007 Urban Challenge. As a result, the policy “will be seen as a milestone and also debated widely,” Rajkumar says. The DoT built some flexibility into its guidance by stating that the policy will be updated annually as technologies and circumstances develop.
The DoT has long positioned self-driving cars as the key to improving automobile safety, reducing the country’s dependence on oil and addressing the gridlock that increasingly chokes major cities. The new policy “provides the path toward a fully driverless car,” NHTSA Administrator Mark Rosekind said at Tuesday’s press conference. His agency released preliminary data on July 1 showing an estimated 35,200 people died in motor vehicle traffic accidents last year, up 7.7 percent from 2014.
The DoT hopes its current approach—encouraging both autonomous vehicle development and safety—will bring those numbers down significantly. “They see an enormous opportunity for improving driving by reducing mistakes that humans make,” RAND's Kalra says.