Self-driving cars might have less impact on safety than expected

A new study from the independent Insurance Institute for Highway Safety shows that self-driving cars may prevent only a third of all road crashes. That’s a far cry from the almost complete elimination of crashes and road-related deaths that we often hear discussed when it comes to a self-driving future as zealots look forward to Vision Zero.

Here’s how the researchers came up with their results. They examined a sample of 5,000 crashes in which emergency responders were called to the scene and at least one of the vehicles involved had to be towed away. They found that the mistakes that led to the accidents fell into one of five categories (in different combinations, that’s why the total percentage is higher than 100%):

1. 24%: “Sensing and perceiving” errors included things like driver distraction, impeded visibility and failing to recognize hazards before it was too late;
2. 17%: “Predicting” errors occurred when drivers misjudged a gap in traffic, incorrectly estimated how fast another vehicle was going or made an incorrect assumption about what another road user was going to do;
3. 39%: “Planning and deciding” errors included driving too fast or too slow for the road conditions, driving aggressively or leaving too little following distance from the vehicle ahead;
4. 23%: “Execution and performance” errors included inadequate or incorrect evasive maneuvers, over compensation and other mistakes in controlling the vehicle;
5. 10%: “Incapacitation” involved impairment owing to alcohol or drug use, medical problems or falling asleep at the wheel.

They then concluded that self-driving cars would eliminate the 10% of road crashes that result from incapacitation and the 24% of crashes that result from sensing and perceiving. Their logic being that self-driving cars won’t drive drunk or fiddle with their phones and their eyesight and sensory skills will far exceed our own.

When it came to predicting errors, however, the researchers pointed out that autonomous technology is yet to prove that it will be able to predict unexpected movements any better, or even as well, as human drivers. They point to the Uber self-driving car that struggled to identify a pedestrian in Tempe, Arizona, back in 2018. The technology fitted to the Volvo XC90 couldn’t figure out what that pedestrian was about to do, as most drivers probably would have done: this woman is crossing the road and I need to brake or swerve to avoid her. The self-driving car failed and the woman died.

The researchers also looked at the 39% of planning and deciding errors where cars are driving too fast or too slow and suggested that if humans can influence how their self-driving cars drive, this category of accident may be equally vulnerable. If we’re in a rush, for instance, surely we’ll ask our self-driving cars to hop on it and push the speed limit. In this case, as the IIHS claims, it could have really dire consequences.

Based on the article in tu-auto.com.