STANFORD, CA – Self-driving taxis that need no human intervention to go from one place to another will reach the market about 2030, according to a panel of 20 experts on autonomous vehicles.
That and other predictions emerge in a poll conducted by Steve Underwood, director of the Connected Vehicle Proving Center at the University of Michigan-Dearborn.
Underwood presents his findings, part of a larger study that still is being edited, during a meeting here of 300 people organized by the Transportation Resource Board of the National Academies. The board is trying to identify research questions that need to be answered before self-driving cars reach city streets and highways.
The panel of experts Underwood surveys is not named and they are unknown to each other. But he says they represent the leading researchers in the field, and some of them likely participated in the transportation board’s TRB summer workshop on the Stanford University campus.
While self-driving taxis remain a science-fiction idea for most people, the subject is being treated seriously by the automobile industry, which is developing the technologies that ultimately could appear in such vehicles.
Volkswagen, BMW, Bosch and Hyundai had speakers at the podium, and the audience was peppered with representatives from General Motors, Ford, Honda and others, led by six from Mercedes-Benz and five from Chrysler.
Auto makers have been automating driving functions for years. Some, such as cruise control, are visible while others, such as emergency stability control, are not. But since the Google car started driving people around California and Nevada with no hands on the wheel for part of their journeys, public interest in the idea has mushroomed.
Steps toward autonomous driving include using sensors for lane-keeping and adaptive cruise control for new functions, such as the combined steering and adaptive cruise control in the new Mercedes S-Class, or the traffic-jam assistant in the BMW i3 coming this fall that will automatically manage low-speed stop-and-go driving.
R. David Edelman of the White House Office of Science and Technology Policy, tells the conference the Obama Admin. is enthusiastic about the development of autonomous vehicles, and suggests cooperation by several government agencies in establishing policy in areas such as data ownership, safety, insurance, and incentives that might be needed to spur development.
One of the unresolved questions about these cars is whether they need to be connected to each other and the infrastructure by wireless communication, which is now being tested by 2,800 vehicles in Ann Arbor, MI, in a program overseen by the University of Michigan Transportation Research Institute.
The Ann Arbor experiment is meant to help the National Highway Traffic Safety Admin. when it decides this fall whether to require such communication in future cars.
Although many individuals at the conference believe connectivity is not a requirement for autonomy, NHTSA is likely to favor the idea.
“I think that the full benefits of automation can only be achieved through connectivity,” says Mike Schagrin, program manager-intelligent transportation systems safety and automation at the U.S. Department of Transportation. “You can only take automation so far with the autonomous vehicle. You take a huge leap when adding connectivity.”
Google uses no connectivity in its current cars.
“We have said before that if (vehicle-to-vehicle connectivity) is available we would use it, but we don't need it,” says Rod Medford, safety director at Google and a former deputy director at NHTSA.
Google has not made clear the business case it sees behind the development of self-driving cars. “We are involved in the development of software, perception and validation of software,” Medford says. “What we deploy and how is not certain, but we are aware that price is important.”
At the 2012 Society of Automotive Engineers conference in Detroit, Google's Anthony Levandowski said its software was “a driver that has driven millions of miles and is ready to take over at any time.”
At the Stanford conference, Medford says: “Our vision is a fully autonomous car. Google is working toward the goal of a level-4 vehicle,” which is the NHTSA definition of a vehicle in which any human occupant has no role in its driving. “We think that a lot of focus should be getting to level 4 as soon as possible.”
Underwood also polls the 20 experts about the intermediate steps auto makers must take before arriving at the complete robot car.
They say NHTSA's level 2, automation of both steering and adaptive cruise control, now is being achieved in cars such as the S-class. Level-3 technologies that let people take their mind off the road by, for example, platooning –where cars could get into line on a highway and follow each other, essentially driven only by the person in the lead car would arrive between 2017 and 2020.
Fully autonomous urban cars could appear in 2025, although there are some votes for as early as 2019, and some for a later arrival around 2030.
Automating cars step by step until they are self-driving seems to be the path the auto industry is taking, but in Europe autonomy is oriented toward public transport. The European Union has funded research that has resulted in fully robotic vehicles that transport people on demand on protected lanes; one such operation is in daily use at London's Heathrow Airport.
Europe also is funding an experiment that will start next March in five cities in which lanes will be shared with pedestrians and some cars.
“Our vehicles already have no drivers,” says Adriano Alessandrini, a researcher at the University of Rome and coordinator of the CityMobil2 project. “We are doing robot taxis in an urban environment right now.”
CityMobil2's 10-passenger shuttle vehicles are connected and not autonomous, he notes. The cars and the route they will follow have sensors to respond to their surroundings, including people and cars in their path.
Part of the project is to develop a European rule allowing such shared-use lanes. Automated vehicles today are allowed in protected lanes, such as the airport-shuttle systems or some subways in Paris, but the goal is to develop rules for a lane protected virtually by sensors and software.
“Our vehicle needs a lane, like the cable cars in San Francisco, identified as being for the vehicles, but not necessarily protected,” says Alessandrini. “In the case of higher speeds, autonomous vehicles need a segregated lane to avoid surprises.”
Regulating the technology also is needed in the U.S. In announcing its autonomous-vehicle policy in May, NHTSA urged states to go slowly in developing rules for operating the vehicles.
However, the federal government already is behind the flow of events.
California in 2012 passed a bill requiring the Department of Motor Vehicles to establish regulations for operation of autonomous vehicles by Jan. 1, 2015.
“Autonomous vehicles are going to change the way that we fundamentally work as a society,” says Bernard Soriano, deputy director of the California DMV. He says the department has a lot of work to do to finish the regulations by December 2014.
Regulators “will be mindful not to slow down innovation,” Soriano says, but the rules will require that “the person in the vehicle is able to take control of the vehicle.” That would rule out the European CityMobil2 vehicles, which have no steering wheel or driver's seat.