Building the uncrashable car

Tesla’s Autopilot, Intel’s Mobileye, and a suite of other technologies are making self-driving autonomous cars closer and closer to reality.  According to Austin Russell of Luminar though, camera systems like Autopilot only solve 99% of all autonomous use cases. This 99% might seem like a lot already but vehicle safety is really about covering the remaining 1% of autonomous driving edge cases where catastrophic things can happen.   Working with safety-conscious brands like Volvo (who invented the modern seatbelt) Luminar and other experts are arguing that a 3D understanding of the environment is the key to the “uncrashable car”. With Lidar technology, 3D data can be obtained with a precision of centimeters up to 500 meters and also allows you to get the same 3D understanding of the environment in both daylight and nighttime. With such advantages though, come challenges in the massive data density associated with 3D data and in using specialized machine learning software stacks and AI to solve those problems. With Nissan scheduled to put these technologies on vehicles by 2030, the future looks bright for the uncrashable car.