STUDY SHOWS SELF-DRIVING CARS WON’T PREVENT MOST COLLISIONS
They may be good at recognizing hazards, but don’t respond well to unpredictability
As it turns out, the fully computerized cars we’ve long been told will be the end of automobile collisions might not save nearly as many lives as we thought. Perhaps no more than the advanced driver-assistance systems — lane-departure warning, forward-collision mitigation, et al. — we’re already using.
It’s quite a shock. After all, we’ve long been told that a robot behind the wheel is infallible.
Not quite, says the U.S. Insurance Institute for Highway
Safety (IIHS). Its latest research suggests computerization will not eliminate all car accidents. In fact, robots in and of themselves will barely reduce collisions by a third — a far cry from media’s constant refrain that driver error is the cause of more than nine out of 10 car crashes.
As the IIHS explains it, driver error is actually the “final failure in the chain of events” that causes collisions. That we knew. What has remained unexplained so far, according to the institute, is that even if those driver errors had been eliminated, two-thirds of those road accidents would still have occurred.
What gives? Well, as the IIHS explains, while “conventional thinking has it that self-driving vehicles could one day make crashes a thing of the past, the reality is not that simple.” In fact, as its research of 5,000 police-reported crashes reveals, even if all the cameras, sensors, and computers that go into making a car fully robotic “can identify hazards better than people, we found that this alone would not prevent the bulk of crashes.”
The problem is that while robots can eliminate “sensing and perceiving” errors (not seeing a hazard) and “predicting” mistakes (misjudging traffic gaps, etc.) they can do nothing to “predict” the future (someone unexpectedly walking in front of the car, like the Uber accident in Tempe, Ariz.). What the IIHS then explains — and my only contention with this excellent study is how it soft-pedals this last conclusion — is that only by eliminating “planning and deciding” errors can self-driving cars be made completely collision free.
And what are planning and deciding errors? As might be expected from an organization representing insurance companies, speeding is its major bugaboo. Indeed, the only way self-driving cars will get us to that predicted fatality-free nirvana is a strict adherence to each and every traffic law and implementing “driving strategies that account for uncertainty about what other road users will do,” such as “driving more slowly than a human driver would.”
In other words, not only might we have to surrender our driver’s licences, but we will have to spend eternity driving places. Imagine being in the back seat on the slowest taxi ride you’ve ever taken for the rest of your driving lives. Because, sure as shooting, no automaker (OK, except maybe Tesla) is going to let cars for which it’s legally responsible disobey even the dumbest of traffic laws.
Previous studies — including research by the IIHS, the AAA Foundation for Traffic Safety, the Boston Consulting Group, and University of Michigan Transportation Research Institute just to name a few — estimate that human drivers, enhanced by the latest advanced driver’s aid systems, could see about the same 30 per cent reduction in collisions as the institute predicts for fully self-driving robots.
Even the IIHS acknowledges that it’s the control over driver temperament, rather than ability, that will make the difference.