Google found an unexpected source of safety problems with its driverless car: robotic precision.
The latest incident involving Google’s self-driving car took place a month ago, during a set of road trials conducted by the company in Austin, Texas.
While approaching a crosswalk, the car slowed down in order to allow a pedestrian to pass. Its safety driver applied the brakes, causing the sedan riding behind to crush into the Google car. Fortunately, no one was seriously harmed, but the incident did highlight a counter-intuitive source of safety problems for the driverless car.
Namely, that its perfect abidance to the rules of traffic makes it awkward when sharing the streets with human drivers. A human doesn’t always slow down when approaching a crosswalk. But the Google car does, causing the driver behind to get dangerously close to the vehicle.
A similar incident occurred at a traffic light. When the car’s sensor detected a driver coming from the opposite direction at high speed, it assumed the other car was about to run a red light. This prompted it to take violent evasive maneuvers, that shocked its occupants. Needless to say, the human driver involved was just a late breaker. But the LADAR on top the Google car failed to notice that.
When driving, humans rely as much on instinct and feeling as they do on the letter of the law. The driverless car’s safety problems stem primarily from its inability to account for that. “The real problem is that the car is to safe…[the cars] have to learn to be aggressive in the right amount”, says Donald Norman an expert in autonomous vehicles from the University of San Diego.
This lack of aggression is best highlighted by a 2009 incident, when the Google car failed to get through a four-way stop. Its programming required for the other drivers to come to a complete stop. Restless to get ahead, the human drivers didn’t conform.
Google’s driverless cars had been involved in 16 traffic accidents since 2009. According to Google, every one of them involves a human driver. If we were to interpret traffic regulations as strictly as the Google car does, every instance but one had a human driver to blame.
Robotics experts speculate that self-driving cars will bring an age of unprecedented road safety. But until all members of traffic will operate with machine-like precision, the Google car would have to adapt to humans’ imperfect manner.
Photo Credits: wikimedia.org