Cars are getting smarter and more capable. They're even starting to drive themselves, a little. And they're becoming a cause of concern for European and American safety agencies and groups. They're all for putting better tech on the road, but automakers are selling systems like Tesla’s Autopilot, or Nissan’s Pro Pilot Assist, with the implied promise that they’ll make driving easier and safer, and a new study is the latest to say that may not always be the case. More worryingly, drivers think these systems are far more capable than they really are.
Euro NCAP, an independent European car safety assessment group (similar to the Insurance Institute for Highway Safety in the US) has just released the results of its first round of tests of 10 new cars with driver-assistance technologies. It also published the results of a survey of over 1,500 car owners in seven countries, asking them what they believe these cars are capable of.
“Seventy percent of people believe you can buy autonomous cars," says Matthew Avery, head of research at the UK’s Thatcham Research, a Euro NCAP member. Eleven percent said they’d be tempted to have a nap, read a paper, or watch a film while using one of the highway-assist features available today, even though every automaker peddling the tech requires drivers to pay attention to the road at all times. “It’s really worrying that consumers are believing the hype.”
And it's the hype that's the problem. Driver assistance systems are a blend of technologies, but the two at their core are adaptive cruise control, which uses a radar behind the front bumper to slow down when the car in front slows down, and lane keeping, which uses cameras to spot white lines and adjusts the steering to stay within them. Automakers aren't saying that their cars drive themselves—they often say the opposite—but they are using buzzy terms like semiautonomous and enjoying the perception that they're technologically advanced.
To see what the cars can actually do, Euro NCAP tested a Tesla Model S (which 40 percent of survey respondents believed could drive itself), along with a BMW 5 series, Audi A6, Mercedes C Class, Volvo V60, and Nissan Leaf. There were also a few cars that US buyers might not know or think of as advanced, but which get semiautonomous abilities in Europe: the DS 7 Crossback, Ford Focus, Hyundai NEXO, and Toyota Corolla.
The engineers put the cars through their paces on a track, testing how well they can avoid collisions in simulated highway driving, when unexpected things happen. It stopped short of rating the cars—it plans to do that in the long run—and instead gave each a report card with notes about what it sees as good and bad.
“The systems on vehicles now, for assisted driving, are really good if you use them correctly,” Avery says. “But they’re not infallible, and the driver has to maintain alertness and be in the loop.” That's why the systems that Euro NCAP says are safest aren’t the most capable, with flashy hands-off features, but rather ones that work with a driver, cooperatively, without ever seeming to take over.
One crucial test Euro NCAP preformed was a look at how reliably adaptive cruise control would brake when a car encounters a stationary object ahead. This can happen when the vehicle in front suddenly changes lane, revealing a parked fire truck for example—something that Tesla’s Autopilot has had problems with. The testers used dummy, deformable, remote controlled cars, which look real to sensors but fall apart harmlessly when hit.
None of the cars did well at this tricky test, which isn’t surprising, as it pushes systems to their current limits as they try to figure out what’s a real obstacle and what’s a harmless road sign or trash can. The computers and sensors in these cars are great at detecting large metal objects, but if they brake for every stationary one, they'd be constantly slamming to a halt, making them near useless, so automakers have to find a balance by often ignoring things that aren't moving. That's a problem if it's a stopped vehicle.
Then engineers tested the cars' lane-keeping abilities through an S curve. They subjected them to a pothole test, where a driver would try to add some extra turn of the steering wheel to avoid something in the street. This test highlighted the difference in approaches to driver assistance between the manufacturers. Tesla’s steering assistance doesn’t let the driver add anything. It automatically handled the S curve very well, even slowing down to make the turns, but if the driver tugged on the wheel, Autosteer disengaged. Audi, Mercedes, and Volvo all allow the driver to gently turn the wheel a bit more and work cooperatively with the computer input. That feeling of being supported, instead of replaced, may not feel as futuristic, but it’s better for safety, Thatcham says, because it’s less likely to lead to complacency.
And that’s the reality of every system on sale now: They are designed to work in partnership with a driver, who has to stay focused and ready to take over. “If you use the Tesla correctly and understand its limitations, it’s actually a very good system,” Avery says. “However, if you misuse it, and it’s open to abuse, it will lead you into trouble.”
That’s because some drivers don’t get those limitations (despite the warnings on the car’s screen and in the manual) when it feels like they can take their hands off the wheel and the car is capable of steering around any object that pops up. It’s not, as shown when a Model X on Autopilot was involved in a fatal highway barrier crash in northern California in March).
“The other vehicles never let you feel that you’re not in control,” Avery says.
Euro NCAP and its member research organizations consider this an important new safety subject and are keeping an eye on it. NCAP plans to have a rating systems in place by 2020, looking at not just the systems on the car but also the manuals and advertising materials, to call out how manufacturers are promoting and selling these systems. (The IIHS is developing a similar ranking system in the US, working with the likes of Thatcham.) Avery says some manufacturers are already taking the findings onboard. Nissan says it’ll review how it sells its Pro Pilot feature.
There is an opportunity here. One of the survey questions asked prospective customers if they’d be willing to watch a training video or do an online course to better understand the functions and limitations of a new car. Seventy-eight percent said yes, which is good news. Because these systems do have the ability to make driving safer by reducing rear-end crashes, lane drifting, and sideswipes. But only if they’re marketed, sold, and used properly.