“We know where the cars are in the moment, but we have to do better than that. We have to predict what’s going to happen.”
The LIDAR system consists of a cone or puck-shaped device that shoots out lasers capable of creating a high-resolution map in real time.
This allows self-driving cars to “see”, even if an object wouldn’t be in their camera’s field of view. The system is sophisticated enough to keep track of the movements of multiple objects simultaneously, even if the circumstances are out of the ordinary.
In addition to driving in the real world, Google’s self driving cars drive log millions of simulated miles in the lab every day. These tests tune systems like LIDAR to help the car make decisions humans wouldn’t be able to make. During his talk Urmson gives an example of a bicyclist who blew a red light. Normal drivers wouldn’t notice the cyclist’s daring move, but the Google car did, even though it wasn’t in its camera’s field of view. Thankfully the cyclist avoided the traffic, but it was proof that Google’s algorithms allow its cars to see what drivers don’t.
Since giving the talk Google’s self driving cars have faced their first accident. Although it was a minor accident, it’s still proof that driverless cars will require much more tweaking before they’re road ready. Urmson says his goal is to have the cars operational by the time his 11 year old son takes his driving test, so that he won’t need to.
Here’s Urmson’s presentation about the fascinating technology that powers Google’s cars: