Tesla upgrades Autopilot for better signal processing

Tesla upgrades Autopilot for better signal processing image

Text: / Photos: Tesla | posted September 13, 2016 15:29

Creates a picture of the world by using the onboard radar

In an effort to provide better signal processing, Tesla upgrades the Version 8 of their software which was previously applied on their vehicles as part of the Autopilot hardware back in October 2014.

With its upgrade, the Autopilot that previously served as a supplementary control sensor to the primary camera and image processing system now becomes a primary control sensor, however, without the need of a camera to confirm visual image recognition.

The radar has been tweaked too and now recognizes objects on the road more precisely. Back then, any object with a concave shape such as a can be detected as something big. Because of this, it will be seen by the radar as a large and dangerous obstacle creating a false alarm. This problem was considered by Tesla as they were developing the update for the Autopilot system.

The onboard radar now features a more detailed point cloud allowing more information per object. It also consists of assembling the radar snapshots into a 3D picture of the world. With this, the car can now assess whether the object is real and assess the probability of collision.

In addition to these, it also takes noteof road signs, bridges and other stationary objects, mapping the world according to the radar. Thus, even if the camera does not notice the object/s ahead, the vehicle will automatically apply mild braking which increases as its system confidence level rises.

Tesla said that this may not always prevent collision entirely but it will dramatically reduce the impact speed preventing serious injuries to the vehicle occupants.