How does a self-driving car work?

The early stages of driverless car technology has its flaws but, some say, it’s worth the risk.

How do these cars work?

There are variations in the technology across developers but the overview below generally addresses the common points:

Data is pulled together from radar and lasers and cameras in order to build a map that reflects the vehicle’s environment.

For example, the roof-mounted LIDAR or “light detection and ranging” device measures distances to objects by sending a laser beam into the surroundings. When the laser beam bounces off an object, a map is created of the surroundings.

Bumper-mounted radar also takes in information on roads, including pavement surface, detours and other obstacles. GPS is used to map a general route. Cameras identify lanes and signs. They pick up vehicles and pedestrians and road surfaces. Some developers use thermal imaging to identify objects.

Then, all this information needs to be sifted through to help the car respond appropriately to specific situations. In order to do this, algorithms or computerized calculations are used not only to identify objects but also to predict their movement. From this, paths are plotted that get the car through its world.

Once plotted, these paths are sent to systems that control acceleration, braking, steering, etc.

Are we ready for the robot-car? Or is the risk not worth the reward? Let’s talk.



Watch David Silver’s Ted talk here:

David Silver is from the Self-Driving Car Team at Udacity where he teaches a program that trains engineers to work in this industry. Silver has an MBA from Stanford University, and a BSE in computer science from Princeton University, and has worked in industry on robot cars.

Be the first to comment

Let's Talk