Let’s face it — driving is hard. Potholes, bike messengers, and everyone else’s terrible driving make for an often interesting experience behind the wheel, and that’s with a (hopefully) dynamic human mind. How will a machine react? Can a self-driving car learn all the subtle nuance of the road? Google thinks so, and has shown off their self-driving cars performance in the streets of their home town of Mountain View, California.
In the video, which you can see below, the self driving car has “learned” how to best navigate many of the scenarios we find in daily traffic. Construction zones, avoiding bicyclists, and a large truck pulled over are just some of the things Google has asked their self driving car to monitor. Google notes it may not be as difficult as we think to educate their car, saying “As it turns out, what looks chaotic and random on a city street to the human eye is actually fairly predictable to a computer.”
Google is also teaching their car to monitor the unexpected, like a rogue bicyclist entering its lane or car blowing through a red light. The humans pulling the strings have also taught their car some best practices, like staying off train tracks, or away when other cars are on them. Even turning right on a busy road with many pedestrians and bicyclists have been appreciated by the self-driving car, giving a glimpse at how Google is appreciating the intricacies of our commute.
It’s not easy, and Google has a lot of work to do before we see these vehicles on our streets. According to Google, they want to teach the car more about Mountain View’s various streets before they “tackle another town”. That suggests the self-driving car needs to be taught all about the city they drive in before being ready to hit the road.