Public support for autonomous cars remains at historic lows, a year on from a spate of fatal accidents. Chris Middleton reports
When an autonomous Uber test car struck and killed 49-year-old Elaine Herzberg in Tempe, Arizona, a year ago (18 March 2018), commentators blamed the safety driver for not paying attention to the road. Some even blamed Herzberg herself, a homeless woman, for crossing the freeway at night.
Yet a preliminary report from the US National Transportation Safety Board (NTSB) revealed apparent failures in Uber’s driverless system and onboard sensors.
According to the NTSB, the car’s radar and LiDAR systems registered Herzberg six seconds before she was hit. However, Uber’s software first classified her as an unknown object, then as a vehicle, and then ‘as a bicycle with varying expectations of future travel path’. Herzberg was wheeling a bike across the freeway, just the kind of situation that autonomous cars would have to deal with every day.
1.3 seconds before impact, the self-driving system realised braking was imperative, but the Volvo’s own emergency and driver-assistance technologies were disengaged by Uber’s software. Moreover, the system apparently wasn’t designed to alert the driver.
So while the safety driver was to blame for not grabbing the wheel quickly enough, Uber’s system failed to identify a woman with a bicycle, failed to avoid the collision, and failed to alert the driver of imminent danger. Yet it was functioning normally, said the NTSB. That’s hardly a vote in favour of the system’s design or safety.
But is it fair to live in fear of driverless cars as the popular media would have us do? The facts say otherwise. Indeed, they reveal that human beings are a much bigger problem.
Nearly 95 percent of the world’s 1.2 million road deaths every year are caused by driver error, such as distraction, unsafe manoeuvres, tiredness, alcohol consumption, looking at smartphones, or simple mistakes.
Human fallibility is a real design challenge for technologists, therefore, but it’s one that applies to driverless cars and those with driver-assistance systems just as much as it does to traditional vehicles.
The point was emphasised less than a week after the Uber crash, when Apple employee Wei Huang died in a Tesla Model X in Mountain View, California, as his car struck a concrete barrier. The vehicle was running on Tesla’s Autopilot software, but Wei was distracted and didn’t keep his hands on the wheel, as Tesla advises its drivers to do.
Nevertheless, he trusted the system to keep him safe – a human being abdicating responsibility for his own safety, in the belief that technology would protect him. Uber safety driver Rafaela Vasquez made exactly the same mistake as she streamed reality show The Voice before her car mowed down Elaine Herzberg. This risky behaviour is one of the challenges in developing autonomous and driver-assistance systems.
Wei and Vasquez weren’t alone in making that mistake. In May 2016, the driver of a Tesla Model S died when it drove under an 18-wheel trailer while Autopilot was engaged. Investigators found that the driver had time to avoid the collision but, again, was distracted; he was apparently watching a Harry Potter movie.
We can infer from both Tesla accidents that the name ‘Autopilot’ was, and is, misleading in the system’s current form, and the manufacturer must take responsibility for that. But we can also infer that people’s addiction to mobile devices is putting lives at risk.
But while each of these deaths was an avoidable human tragedy, it’s fair to say that each was a drop in the ocean compared with the colossal loss of life in ‘traditional’ road accidents last year. Indeed, the slaughter that we have all become accustomed to is what autonomous systems are designed to avoid.
Yet smart vehicle crashes keep grabbing the headlines – even minor ones, which spurred Tesla’s Elon Musk to tweet: “A Tesla crash resulting in a broken ankle is front page news and the ~40,000 people who died in US auto accidents alone in [the] past year get almost no coverage.”
It’s an excellent point: deaths in cars running under software control are extremely few and far between compared with the carnage on our roads, where one person dies for every 1,000 traditional cars in use worldwide. But there are comparatively few truly driverless cars out there, so like-for-like comparisons are impossible.
Either way, it’s clear that the journey to a world of autonomous services summoned by apps has at least one major obstacle: the messy human world we all live in, which demands vast amounts of data to model and predict more accurately. ‘Crossing the chasm’ from autonomous and driver-assisted cars sharing the road with humans to a world of mass autonomy among millions of connected vehicles is a challenge.
That chasm is a cultural one, as well as practical and technological: the US and other Western countries are deeply attached to the concept of private transport, which sustains millions of jobs. That said, replacing properties with services – swapping units for bits – is something we’ve all become accustomed to in other aspects of our daily lives.
Either way, private car ownership in ageing cities full of ageing people is not a sustainable solution as cities grow; smarter, connected, frictionless transport options will be essential. Most analysts accept that a mix of on-demand services, from autonomous cars and electric bikes to delivery bots and even flying taxis, is the direction of travel.
In the meantime, however, are human beings placing too much trust in too little technology too quickly? Not anymore, according to the American Automobile Association. Just weeks after 2018’s highly publicised fatalities, the AAA released a report saying that support for autonomous transport had plummeted in the US.
That report found 73 percent of American drivers saying they would be too afraid to ride in a self-driving vehicle, up from 63 percent in late 2017. Meanwhile, nearly two-thirds (63 percent) of US adults said they would feel less safe sharing the road with a self-driving vehicle while walking or riding a bicycle.
Of particular concern was the finding that the biggest fall in confidence was among younger adults. The percentage of millennials who would be too afraid to ride in a self-driving car soared from 49 percent to 64 percent last year.
Coming so soon after deaths on the road – not to mention the media’s obsession with Uber’s and Tesla’s management – the findings were perhaps not surprising. So a year on from Tempe and Mountain View, has support returned to pre-2018 levels?
Not yet, according to the AAA. Its latest sentiment-tracking report published this month found that 71 percent of Americans still don’t trust self-driving cars, a marginal increase in support.
“It’s possible that the sustained level of fear is rooted in a heightened focus, whether good or bad, on incidents involving these types of vehicles,” said Greg Brannon, AAA’s Director of Automotive Engineering. “Also it could simply be due to a fear of the unknown.”
Countering that fear is the key, therefore, especially in a world of memes, social media, and so-called fake news. In January, Waymo – the driverless division of Google’s parent Alphabet – joined forces with GM’s Cruise Automation and 22 other companies and organisations to form the Partnership for Automated Vehicle Education (PAVE).
The new industry group plans to offer self-driving test rides for consumers and policymakers, as well as to teach lawmakers, consumers, and educators about the technology.
There is more good news for the future, found the AAA in its survey. More than half of those surveyed in its report (55 percent) believe that most cars will have the ability to drive themselves by 2029. So a decade of transformation lies ahead.
- In related news, analyst firm Future Market Insights has published a detailed report on the global automotive sensors market. It forecasts that the space will witness a compound annual growth rate (CAGR) of 7.6 percent from 2018 to 2028. The market was worth $17 billion in 2017 and is projected to hit $39 billion by the end of 2028, according to the analysts’ figures.
A second report on the automotive pressure sensors market, published by 360 Research Reports, says that this subsector will hit $6.6 billion by 2023, growing at a CAGR of 8.1 percent.
Be part of a discussion and connect with like-minded leaders in your sector at our forthcoming trade event – The Sensor Show – next year.