US: Google’s autonomous prototype involved in another crash, company says software is improving image

Once more, one of the Google self-driving cars on the streets conducting trials has been hit by a human driver and the program is gathering more and more evidence about the obvious – human error is absolute.

It appears the impacts recorded by Google’s prototypes are supporting the rather obvious – the computers are more efficient that human drivers and the former are only getting better at it. This could be the only possible explanation for the string of rear-end collisions suffered by Google’s testing fleet – the latest happened just last month. The company has clocked more than a million miles of prototype testing in an effort to grow its database with as many real-world driving scenarios as possible.

Google’s autonomous prototypes have clocked 16 accidents, with 12 rear-end collisions and none of them the fault of the self-driving cars, according to figures provided by the company to the California Department of Motor Vehicles. The latest incident involved a driverless Lexus that approached a Mountain View crosswalk without a stoplight and detecting a pedestrian ready to cross – the autonomous car started braking and the onboard safety driver decided to intervene and brake more firmly.

This made the car get hit by a driver that was also changing lanes at the same time – the accident was his own fault because he should have allowed for more space between the vehicles. According to the program manager Chris Urmson, data analyzed afterwards showed the Google test driver followed protocol to intervene in a crisis situation but that if he allowed the autonomous brain to have its way it would have braked less violently and thus give the rear driver more space to react.