We hear a lot about how self-driving cars will make driving safer by taking human error out of the equation. Whether it’s car makers, self-driving tech firms, or network vendors touting enabling technologies like 5G and C-V2X, the message is the same: autonomous cars will reduce and even eliminate road fatalities.
That might be true for the people in the cars. For pedestrians, not so much.
Obviously, a key component of autonomous car safety is the car’s ability to detect obstacles in the road and avoid them – all within seconds, in real time, regardless of road or weather conditions.
Many cars today have advanced driver-assistance systems (ADAS), which don’t enable fully autonomous driving, but use a lot of the same technology (sensors, lidar, radar and cameras) to warn drivers of things like an impending collision. Many cars also come with automatic emergency braking (AEB) systems that (as the name implies) enable the car to hit the brakes in such situations.
For the most part, stats show AEB systems work reasonably well when it comes to avoiding or minimizing the impact of car collisions. But they’re not so good when it comes to avoiding pedestrians. In fact, not all AEB systems even have pedestrian detection capabilities. And according to a new study issued last week by the American Automobile Association (AAA), the ones that do don’t work very well.
The AAA study [PDF] tested four cars with pedestrian detection capabilities in various road scenarios of increasing difficulty:
An adult crossing in front of a vehicle traveling at 20 mph and 30 mph during the day and at 25 mph at night.
A child darting out from between two parked cars in front of a vehicle traveling at 20 mph and 30 mph.
A vehicle turning right onto an adjacent road with an adult crossing at the same time.
Two adults standing along the side of the road with their backs to traffic, with a vehicle approaching at 20 mph and 30 mph.
The results: if you’re an adult crossing in front of a vehicle traveling at 20 mph during the day, the odds of you getting run over depends on the car, but generally they’ll avoid you – around 40% of the time.
That’s the best result. In every other scenario, the chances of a collision go up considerably – the general rule seems to be that if any car is going faster than 30 mph, or if it’s night, the AEB won’t activate in time, or at all.
Look around you
Autonomous car evangelists may point out that ADAS technology is still evolving, and that once self-driving cars become commercially viable, the technology will have improved considerably. Which is true. But that doesn’t mean pedestrians are going to be any safer.
In a recent article in The Guardian, Peter C. Baker points out that pedestrian detection has been around since Volvo introduced it in 2010. The technology has been improving ever since, and is more widespread – yet the number of pedestrians killed by cars has gone up every year.
One reason for this is that most drivers don’t know how to use ADAS systems, or even what they do. And one reason for that is car makers are muddling the market with different names for the same basic functions, according to a separate AAA report released earlier this year. AEB systems alone are marketed under 40 unique names, according to NPR, and the name doesn’t always describe clearly their actual capability.
Another reason, Baker writes, is that the discussion of road safety tends to focus on individual behaviour – the driver was distracted, the pedestrian didn’t look both ways before crossing, etc – and not road design and planning.
This matters because there are a multitude of environmental factors in pedestrian deaths, such as speed limits, lighting, guard rails and clearly marked pedestrian crossings. However, the goal of urban road design is to facilitate a constant steady flow of vehicle traffic as fast as possible – and typically, pedestrian safety is not so much an afterthought as a general attitude that it’s the pedestrian’s responsibility to not get run over.
This is the environment self-driving cars must navigate, and Baker expects it to get worse as both urbanization and car ownership increases worldwide – to say nothing of the fact that the rising rate of pedestrian deaths isn’t alarming us nearly as much as it perhaps should. Statistically, in just about every country on Earth, it’s more dangerous to be a pedestrian (or ride a bicycle) than to drive or ride in a car.
The question is whether better technology alone will reverse that. Baker’s argument is that it won’t – we also need to rethink our approach to urban road design. A road system designed for constant and fast traffic flow is not conducive to pedestrian safety, no matter who is (or isn’t) driving.
Consequently, everyone in the autonomous-car value chain might want to lay off the utopian sales pitch of safer roads and saving lives. Improved sensor technology, smarter AI-powered cars and faster wireless networks can only get us so far, and pretending otherwise only distracts us from the bigger picture.