Google's self-driving car was once just a pet project, but it has since bloomed into one of the company's largest endeavors. Now, Google's autonomous fleet has logged over 190,000 miles, is street-legal in Nevada, and it may or may not have gotten into its first crash. But what we don't know is just how the automated system manages to stop at a red light or keep from careening into us--until now.
Stanford University professor Sebastian Thrun, who led the development of the Google's self-driving car, and Google engineer Chris Urmson revealed the secret workings of these autonomous vehicles in a keynote at the IEEE International Conference on Intelligent Robots and Systems last month. The talk divulged how the system works, and Sebastian dn Chris showed footage of how it sees the world.
At the "heart of the system" is the Velodyne 64-beam laser range finder that's mounted on to the roof of the car. The device scans and records laser measurements to create a 3D model of the world that it correlates to high-resolution maps. From this it creates routes that avoid obstacles and obey traffic rules.
The vehicle is also equipped with radar, GPS, an inertial measurement unit, and other sensors to keep the vehicle on course and provides it with 360-degree situational awareness.
More interestingly, the car is programmed to be extremely courteous, yielding to any pedestrian crossing the street as it strictly adheres to road rules. At the same time, it can also be aggressive to other vehicles not obeying the right-of-way rules by easing into the intersection to assert that it will be turning first.
Be sure to check out IEEE Spectrum for the full story, along with plenty of photos and videos.