Spend enough time behind the wheel, and chances are you're going to see some pretty wild things — if you work for Google, at least.
One time, an onlooker was so excited to see one of the company's self-driving cars pass by that he ran out onto the street completely naked and leaped onto the vehicle.
Another time, the car had to slow down because there were as many as three other cars driving the wrong way up the street toward it.
There was the time a group of people hopped across the street in front of a Google car, interrupting its route with a real-life game of Frogger.
And then there was the mysterious case of a woman in an electric wheelchair chasing a duck in circles in the middle of the street.
These quirky encounters really happened, said Chris Urmson, the director of Google's driverless car project, at South by Southwest Friday.
But Urmson wasn't just in Austin to talk about strange stories involving driver automation. He also shed some new light on a recent high-profile incident involving a company car colliding with a bus.
If you're not familiar with the tale, the incident marked the first time that Google's car was "at fault" for a crash. It was a minor accident, and nobody was hurt. But the episode is instructive for policymakers and engineers, and in typical Google fashion, Urmson on Friday showed off some images of the data the car was gathering in the moments leading up to the crash. The images below are essentially what the car itself was seeing right up until it sideswiped the bus.
Here's the car's initial expected route, in green. You can see it was planning to execute a right-hand turn.
Then it saw some sandbags — those are the yellow obstacles below.
So it pulled back into traffic to avoid the sandbags. You can see in the bottom right-hand corner the car's software running a set of calculations to predict where the bus was going to wind up. But it didn't count on the bus failing to slow down.
Which is how you get this. The damage to the bus is circled in red.