Google’s self-driving car is an amazing piece of engineering; it drives, for the most part, like a human driver. The Lexus RX350h has been modified with technology more advanced than what is in any other auto-driving vehicle, including laser, camera and radar systems, but there’s so much more.
The modified Lexus does more than “see” the landscape it navigates through; it decides on its actions based on the massive amount of data that the Google Self-Driving Car Project has accumulated about the roads it has traveled. The project team gathers data about new territory the car is about to drive into before it does so, including information about permanent features like lane markers, curb locations, traffic light heights and speed limits.
Andrew Chatham leads the mapping part of the project. He said recently at a press event, “We require digital maps in order for our cars to be able to drive.” He added that this information “makes the job of building the self-driving car software much simpler.” This means the vehicle already knows the stretch of road it is about to travel, which allows the software to focus on variables such as other cars, pedestrians, cyclists, construction and other variables in real time.
Software lead Dmitri Dolgov says that is the “magic of maps.” However, this means the range of the car is limited to areas that Google has data for. Chatham said, “If we have not already built our own maps in an area, the car cannot drive there.” He also said that as sensors get better, less accuracy will be required.
Chris Urmson, project director, summed up the important role maps play. “It’s also a really important part of the safety story.” He said that if the car remains only in areas that the system knows, “you can be confident that it’s going to behave well in those situations.”
To accomplish their task, the self-driving cars have lasers that rotate ten times per second, with 64 vertically stacked beams that constantly measure the distance to the objects surrounding them. The mechanism provides a 3D, 360 degree view of the area around the car. The radar and camera provide additional information, which is integrated by team members. That gives them a map to use so that the car can do the job on its own.
Dolgov says he is not worried about the seemingly massive task of gathering information about every place a person might want to go in their Google ride. He says that the cars can be driven by humans, and he can imagine a scenario where owners of the vehicles could take part in collecting data.
The laser, radar and camera systems are built into every self-driving car. If an owner wanted to drive down a street not mapped by Google, he could drive manually through the area a few times, let the car gather data, and sit back and relax on the way home.