Google offers inside look into driverless car collision with human driver

While news reports about minor collisions involving Google's self-driving cars are getting more common, the company has for some time now tried to explain that nearly all the accidents are the result of human error, and not caused by software or other technical glitches. Chris Urmson, the head of Google's driverless car program, has written a new blog post on Medium that offers a detailed, inside look at a recent collision, and how the self-driving Lexus performed exactly as it should have.

The incident took place on July 1st, and saw Google's vehicle rear-ended by a human driver at a traffic light. The intersection was experiencing heavy traffic, so the self-driving car, along with two human-driven cars directly in front of it, remained stopped during a green light. This was to avoid getting stuck in the middle of the intersection, and causing further congestion for others.

Google released a recreation video showing exactly what happened next:

The car behind never braked, and hit the driverless car at 17mph, most likely because the driver only saw the green light, but wasn't paying enough attention to see the other cars were stopped.

Google's car performed exactly as a human should have at the intersection: stopping to avoid further traffic problems. As Urmson points out, it was the human driver behind that made the mistakes.

As you can see from the video above, our braking was normal and natural, and the vehicle behind us had plenty of stopping distance — but it never decelerated. This certainly seems like the driver was distracted and not watching the road ahead.

In other words, if another human-driven car had been in the Google vehicle's place, it still would've been rear-ended. Urmson adds that the sensors on Google's car allow it to have "360 degree visibility and 100 percent attention out in all directions at all times," whereas human drivers can be prone to things like checking mobile devices instead of pay attention to the road.