Regina Garcia Cano/AP Photo
Patrons wait to board a driverless shuttle bus on a street in Las Vegas.
Public transit systems have run driverless trains for decades. The first fully driverless train network, Vancouver's SkyTrain in western Canada, has been in operation since 1985. The Paris urban transit system has automated its oldest and busiest Metro line and plans to automate about half the subway system by 2050. Yet in the United States, transit automation is years behind Europe where driverless trains mostly connect airport terminals.
Automated buses are already operating in some communities. But when American cities have green-lighted the technology, the results have been uneven: A November accident in Las Vegas involving the first automated bus for public use in the U.S. raised still more questions about whether the technology is ready to hit the road. Some private companies launch relatively untested technology, hoping to spearhead innovation by being the first to deploy an invention. But without a tight regulatory regime in place (and in the U.S., the technology is miles ahead of the regulators in most states), the results can be disastrous.
Las Vegas's transit agency, the Regional Transportation Commission of Southern Nevada (RTC), uses technology developed by the French company Navya, one of the global pioneers of driverless bus production, and is partnering with Keolis, the public transportation operator owned by the French national railway system. But just two hours after RTC launched the American Automobile Association-sponsored pilot, a truck collided with the Navya shuttle.
The bus sensor technology responded at it was programmed to do—the shuttle stopped when it detected the truck. But the truck kept going and hit the bus. There were no injuries to the bus passengers or the occupants of the truck, and Las Vegas officials decided to continue the year-long pilot program. The accident has not deterred other cities. Small-scale pilots are underway, about to begin, or have been completed in Minneapolis, Denver, Gainsville, Florida, and San Ramon, California, while Bismarck, North Dakota, might consider a pilot program next year.
But driverless technology is still very much in its infancy. In March, an Uber driverless car struck and killed 49-year-old Elaine Herzberg, who was walking her bike across a Tempe, Arizona, road. Transportation experts blamed Uber's self-driving technology for the crash, saying that an attentive human driver would not have killed Herzberg. The average car accident death rate in the United States is one fatality per 85 million vehicle-miles traveled. (In Britain, where the roads are slower and safer, it is one in 141 million miles.) The American self-driving car industry has only logged 15 to 20 million miles traveled, and Uber driverless cars have only logged one million miles.
(Uber suspended its self-driving car experiments after the crash, as has Toyota. More comprehensive regulations would likely demand that the operators be held to higher safety standards than human drivers, rather than lower standards as is the case today.)
Nevertheless, public transportation officials continue to take a hard look at the technology. Transit authorities could run small driverless buses more frequently, especially in areas where it may not be cost effective to run large buses, such as on routes with lower ridership. How buses interact with other vehicles; how bus drivers might be reassigned; and how to respond to passenger concerns about traffic accidents and crime are a few of the concerns that the technology poses for transit systems.
Automation offers certain benefits, speed for one. In Paris, automated trains provide faster acceleration and deceleration. On the automated Metro line, average speeds improved from about 16 miles per hour to nearly 20 miles per hour, a 23 percent increase. But while driverless trains are faster than trains with people at the controls, buses have yet to pick up the pace: one driverless bus prototype, the EZ10, has a top speed of 25 mph and a cruise speed of 12 mph.
However, automated buses have significant challenges to overcome before widespread deployment can begin. Like cars and unlike trains, buses run on roads shared with other vehicles, with cars on crossing streets at intersections and pedestrian navigating crosswalks or, worse, jaywalking. Driverless buses would need regular updates about traffic rules (like temporary speed limits and road conditions) along its route as it jockeys with other vehicles and pedestrians. Driverless buses would still need to employ technology that can recognize and brake for pedestrians and cyclists. Yet while driverless buses may be able to navigate safely, they may not be able to avoid collisions with other drivers who do not know how to interact with driverless buses as the Las Vegas incident shows.
The new technology also may be at a disadvantage due to the historic, second-class status of buses. Eric Goldwyn, a postdoctoral fellow at New York University’s Marron Institute of Urban Management, argues that every new mode of transportation needs extensive associated infrastructure. Trains have railroads, cars have wide paved roads, planes have airports, high-speed trains have dedicated high-speed lines. Yet buses never have dedicated infrastructure: Bus rapid transit lanes have not been widely adopted nationwide (and some metro areas are just catching on to the benefits) which is one of the reasons that buses do not have usage rates as high as cars or subways.
While driverless buses handle acceleration and braking with relative ease, it is not clear to what extent driverless bus technology can replace the human eye. Obstacles in a road, such as a pedestrian about to cross a street or a car that suddenly changes lanes, still flummox the technology, rendering an automated bus more of a hazard than a viable means of transportation. Driverless buses may need to maintain greater distances between other vehicles on the road, and adjust speeds accordingly, which would reduce rather than improve performance: a bus that slows down on the road trying to judge whether something is an obstacle or not creates more safety issues.
Potential road hazards are not the only problem. Brian Sherlock, an international safety specialist with the Amalgamated Transit Union, notes that the technology sometimes fails when the vehicle has to enter a bus bay where passengers wait to board.
Mobileye, an Israeli company, has pioneered some of the most advanced technology driverless in use today. Even so, a 2016 study by Jung-Hyup Kim, a University of Missouri industrial and manufacturing systems engineering professor, found that that Mobileye's assisted driving technology has a reaction time of 2.8 seconds, too long to brake in some cases. Moreover, a 2017 review by the trade paper Electronic Engineering Times finds flaws in Mobileye's safety calculations and says that its propriety technology should be independently assessed, as its claims to safety are all sourced to internal company data. Indeed, this week, a Mobileye driverless car ran a red light at a demonstration in Jerusalem, which was blamed on electromagnetic interference from a camera from a nearby film shoot.
Even in Helsinki, the one of the few cities in the world with a regular driverless bus service, the buses a run in a closed, controllable environment. Jaime Garmendia, a commuter rail project coordinator at the Massachusetts Bay Transportation Authority, the regional Boston transit system, searched out the city’s automated buses during a recent visit and found that they are restricted to a closed, short route (only a fraction of a mile) at the city’s zoo. Cars cannot drive on the road where the buses run.
While labor unions have been concerned about what automation means for transit workers, these drawbacks mean that front-line bus employees are still in the driver’s seat. While some systems may choose to prepare for the transition by choosing to hire fewer drivers as veteran drivers retire, it is far from certain that driverless buses will replace conventional buses in the short term.
Sherlock of the Amalgamated Transit Union, argues that automated buses are simply not roadworthy, since current state of pedestrian-detection technology for driverless vehicles in general is “really fairly wretched.” Moreover, the Las Vegas crash shows that while the vehicles themselves may be safe, but the human drivers of other vehicles have unpredictable responses to unfamiliar technology. (While driverless subways and elevated trains perform well, for the most part, on dedicated tracks, they have not been deployed on lines with street-level crossings and traffic lights—that is, where they may have to interact with other vehicles or humans.)
Like drivers of other vehicles, passengers also face a major culture shift, accepting that a computer not a human is in control of the bus. Yet the more difficult adjustment may be riding with strangers on an unstaffed bus. People routinely share subways with strangers. Usually those trains are staffed with just one train driver, in a compartment away from the passengers and is unable to immediately respond to an incident. But bus passengers would have to take an extra leap of faith if there is no transit employee anywhere on the bus. Some bus operators employ an attendant to assuage these concerns. The Las Vegas pilot has customer service attendants.
Amelia Zaino, a biracial, 27-year-old community activist from the North Bronx, does not worry about crime on public transit. She doesn’t believe that driverless buses would be more dangerous anyway in New York because “here the buses are so big and so crowded that the driver really couldn't do much in terms of pickpocketing or small crimes.” However, while Zaino expresses some apprehensiveness about the technology, she is more worried traffic accidents than crime.
Others worry that the absence of a driver means trouble. Martha Dosztál, a 28-year-old transgenger University of London student of Hungarian and British descent says she would feel “very vulnerable in small shared driverless vehicles,” and would especially worry about safety in Hungary, where homophobia and transphobia are worse than in Britain and the United States. During visits to Baltimore, Dosztál always sat next to the driver especially when she rode buses at night when there are fewer passengers.
Employing bus attendants could reassure some nervous riders. However, turning drivers into attendants would turn a trend to reduce the need for labor into a scheme to further de-skill workers, replacing relatively well-trained and paid bus drivers with low-wage customer service or security workers. When Paris automated its Metro line, the system did not lay off any drivers. To avoid strikes, it reassigned them as platform attendants. But because bus operating costs are heavily dictated by driver wages and benefits, some transit agencies may not reassign every driver to a new position unless an agency automates buses very slowly.
Neither driverless buses nor cars have little chance of revolutionizing transportation without substantial investments in entirely new infrastructure, according to Adriano Alessandrini, a professor of civil and transportation engineering at the University of Florence in Italy and one of the lead researchers for the European Union’s CityMobil2 project which aims to integrate automated vehicles into cities.
The free-for-all approach used by Uber and other American technology companies is unwise: American tech companies' efforts focus on running driverless vehicles in mixed traffic. But CityMobil2 runs its vehicles in mostly segregated traffic in closed circuits aiming to run them in fully-mixed traffic in the future. Helsinki's driverless bus, running on a road without cars, is one example of this alternative approach.
Although the CityMobil2 project does suggest that some driverless cars and buses may be in our future, their full adoption by transit agencies may take decades. Security issues alone may restrict deployment to the busiest routes in the cities with the most extensive bus systems. The Las Vegas accident and Helsinki’s limited bus route suggest that the most pressing issue for driverless road vehicles—the need to safely navigate streets along with other cars, trucks, and buses—is not yet solved.
What is more likely is that public transit operators will continue to study transitioning to driverless fleets. But the technology media’s claims of incipient disruption of public transportation as Americans know it today are premature. American cities that want better public transit should invest in conventional bus and rail service and not base their bus fleet decision-making on a technology that is still in its infancy.
This post has been updated.
