Imagine how carefully you would drive if there were young children running around on either side of the road. Knowing that they could dart suddenly in front of your vehicle, you'd have to go slowly enough to stop in time.
That sort of caution, that deference to people on foot, is what Adam Millard-Ball sees in future cities full of driverless cars.
"This has the chance to change the norms," says the California academic, who has analyzed the emerging technology's possible effects and believes it has the potential to change the way people live and walk in a city. "If you're a pedestrian and you know the car is going to stop, the calculus changes. You can cross with freedom."
As cities ponder how to protect the most vulnerable road users, autonomous vehicles (AVs) are offering the prospect of re-empowering pedestrians nearly a century after motorists seized control of city streets, inventing the concept of jaywalking and pushing foot traffic onto the sidewalks.
The technology is no longer in the realm of science fiction. Car makers and technology firms are investing heavily and expect AVs to hit the market in large numbers as early as next decade. Self-driving taxis are being trialled by Uber in Pittsburgh. The firm's self-driving freight company, Otto, performed its first truck delivery in Colorado in October. In November, Ontario announced its support for AV testing in the province.
The potential impacts are seismic. Less road congestion, coupled with fewer deaths and injuries, are among the best-case scenarios envisioned by academics and futurists.
With the technology starting to appear, cities are looking at how the vehicles will change transportation patterns and fit into plans to make the roads safer for all users. In Vancouver, planners are working from the assumption AVs will use roads more efficiently, opening up extra space for pedestrians and cyclists.
But there are also fears of technological failures and societal pushback. Say, for example, pedestrians come to expect that they can step safely in front of an autonomous vehicle.
Will they cross wherever they like and aggravate motorists by slowing them down? If so, will motorists insist that pedestrians be kept back, fenced off from the road, to let traffic move freely?
If AVs are programmed to go slowly enough never to hurt someone, will they have any appeal to motorists?
Will manufacturers keen to sell cars program their vehicles to balance speed against safety, judging a certain level of risk to other road users to be acceptable?
As car design has improved and traffic laws have been better enforced, the number of people who die in collisions has dropped dramatically. Transport Canada says that the average annual death toll, which stood at 4,220 in the late 1980s, had been cut in half by the first half of this decade. A few thousand people are still killed on the roads in Canada every year, though, and the benefits of safety improvements have not been shared evenly, with pedestrians and cyclists the most at risk.
"Road-crash injury is a social-equity issue," argues an Ontario coroner's report into 2010 pedestrian deaths that showed people on foot face extra risk. The report called for "equal protection to all road users."
In recent years, several Canadian cities have begun increasing safety efforts to address the imbalance. The concern is particularly acute for seniors, who are more likely to be hurt when hit by a vehicle and form a disproportionate share of the victims. Canadian society is aging, and the increasing proportion of the elderly is expected to bring a corresponding rise in dementia, adding urgency to the issue.
Driverless cars, which are often billed as a way for seniors to retain their mobility, could also help keep them safe while walking. But the focus on a solution that is still years away is not without controversy.
"One of the things that concerns me about this conversation [about AVs] in general is it's allowing people to imagine themselves past the current problems," said Maureen Coyle, who sits on the steering committee of the advocacy group Walk Toronto. "There is that desire to sidestep the immediate problem of how we solve the road-safety issues facing us so pressingly."
And even though most analysts believe AVs will make far fewer mistakes than human drivers, how safe they will be for pedestrians and cyclists remains to be seen.
A study in the journal Science showed the majority of consumers would not buy an AV unless it was programmed to put its occupants' safety above that of others. An executive at Mercedes-Benz made headlines last year when he was quoted saying – in a new report promptly disputed by his firm – that their AVs would prioritize saving their passengers' lives over those of nearby pedestrians. And Google raised eyebrows by patenting a sort of human flypaper for their cars, designed so that when pedestrians are hit, they stick to the vehicle and avoid the further injury of being thrown.
The most critical safety factor, however, will always be speed. In a position paper released last summer, the National Association of City Transportation Officials – a primarily U.S. group which counts Montreal, Toronto and Vancouver as members – called for AV speeds to be restricted in urban centres. The group wants a limit of 40 kilometres an hour. Even at that speed, the typical small vehicle requires approximately 13.5 metres of road to brake to a full stop. But being hit at 40 km/h or less allows most pedestrians to survive.
The association's plea mimics a U.S. safety campaign in the 1920s to restrict vehicles to 25 miles an hour (40 km/h) in cities, a push that was thwarted by motoring advocates. Keeping speeds down continues to be controversial today. Even though some cities, including New York, have already dropped their default speed limit to 25 mph in order to protect pedestrians, Toronto is among the municipalities that have resisted this change.
Complicating matters for vulnerable road users, early AV testing has shown that some of them struggle to understand the behaviour of those on foot or riding a bicycle. And there is research suggesting that humans don't know how to interact with AVs, either.
Michael Clamann, a senior research scientist in the Humans and Autonomy Lab at Duke University in Durham, N.C., recently ran an experiment to explore how an AV could communicate with pedestrians.
A van was fitted with a display board that showed either the vehicle's velocity or a "walk" or "don't walk" image. People on foot were told this was an AV (though it actually had a human driver) and were told to assess the displays and determine when it was safe to cross. Interviews after the experiment showed that 76 per cent of people saw the displays, but only 12 per cent used them to inform their decisions on whether to cross.
"What we ended up finding was … they didn't pay attention to it," said Dr. Clamann, calling for more work on figuring out how an AV can make its intentions known to someone nearby.
The Swedish company Semcon has floated the idea of having an LED smile or grimace on a vehicle's grille. A 2015 academic paper prepared for Toronto's Transportation Services department raises the prospect that all road users might have to embrace technology to replace subtle human signals. It suggested "as a safety measure, pedestrians and cyclists may one day passively announce their presence to connected vehicles via mobile apps or wearable technology."
Sorting out this communication is crucial because a lot of the interaction now between drivers and those nearby is non-verbal, based on eye contact and small gestures.
Autonomous vehicles operating in a closed environment – a port or a mine, for example – are relatively easy to create. More difficult is programming them for modern cities. Operating safely in an urban environment where drivers, cyclists and pedestrians are all acting unpredictably is the hardest job of all, and achieving this is key to the mass acceptance of these vehicles.
In September, the U.S. Department of Transportation released a 116-page policy paper on autonomous vehicles. One of the "behavioural competencies" specified is that AVs must be able to "yield to pedestrians and bicyclists at intersections and crosswalks."
This is a key point, and one that Dr. Millard-Ball believes could lead to a profound change in how roads are used. For a paper published in October in the Journal of Planning Education and Research, the assistant professor of environmental studies at the University of California, Santa Cruz, applied game theory "to develop a new model of road-user interactions" between pedestrians and drivers.
In the paper, he argues that pedestrians often don't assert their right to cross, because they fear the oncoming vehicle won't stop. But a pedestrian who is certain the vehicle will yield will be more assertive. Even before AVs become ubiquitous, he argued, human drivers who encounter newly assertive pedestrians will learn to be more careful. And their new caution could spark a self-perpetuating effect, making walking safer and prompting other pedestrians to assert their rights as well.
"At a crosswalk … driverless cars potentially might be stuck for hours," MIT professor of robotics emeritus Rodney Brooks posits in a recent blog posting.
"Will people take pity on them as they do on human drivers? To take advantage of this, the cars would need to understand human social signals of giving them a turn, but without a reciprocal signal [from the vehicle] it is going to be confusing to the generous pedestrians and they may soon decide to not bother being nice to driverless cars at all."
Of course, this sort of pedestrian assertiveness requires faith that the technology will be 100-per-cent reliable – a level of confidence that remains a sticking point for many.
"There's hope, certainly, that [AVs] can improve road safety, and that certainly would be welcome" said Erin O'Melinn, spokeswoman for the Vancouver advocacy group HUB Cycling. But she added "there's so many variables on the road," citing children and differently-shaped types of bicycle as complicating factors.
Even though most analysts agree that computers will do a better job than human drivers, it remains unclear how much better an AV will have to perform to be good enough in the public mind.
A study released in January by the consulting firm Deloitte suggested that people are open to being convinced, but many currently have little faith in AVs. The report looked at six countries and showed that the belief AVs will not be safe ranged from a high of 81 per cent in South Korea to a low of 62 per cent in China. Canada was not included, but fully 74 per cent of U.S. respondents did not believe AVs would be safe.
Last year's U.S. government policy paper on AVs even raises the concern that a firm might try to cheat on the safety testing, much as manufacturers have been accused of fiddling their emissions results, and urged for strong oversight.
Brad Templeton, a consultant who has worked for years in the AV field, including time at Google, believes that no company will dare release an autonomous vehicle unless their lawyers are satisfied it's safe.
But he adds a sobering followup: "People will die. Perfection is not possible."