NVIDIA’s Self-Driving Platform Uses AI To Teach Cars How To Move On Their Own

Machine learning could revolutionize the automotive industry. NVIDIA CEO Jen-Hsun Huang proved AI can improve the process of making driverless cars as good on the road as human drivers at the GPU Technology Conference last week. His DriveWorks platform coupled with DAVENET deep neural network reduces the cost of mapping areas frequently and training autonomous vehicles repeatedly #machinemagic #automagic

DriveWorks developer platform, set to ship this spring, starts with computer vision software and sensor fusion to work with 12 individual cameras and sensors in order to give the self-driving car accurate information about the environment it will face. Once it does, advanced machine learning helps the car learn the road and navigate safer.#machinemagic

If self-driving vehicles have relied on human-made algorithms to follow the desired route faithfully until now, NVIDIA and other companies are now investing their time and energy in making cars learn the beaten path after a series of driving sessions. As Huang shows us in the video below, after 3000 miles of supervised driving, their car – powered by DAVENET (the deep neural network) – can navigate safely on freeways and country roads, even on rainy days.

As you can imagine, NVIDIA isn’t going to concentrate on automotive solely from now on; instead, it’s going to provide this technology as part of a set of tools that makes using the Drive PX 2 necessary. DriveWorks platform is going to support maps such as those from HERE but it will encourage map development, also, with the help of cloud-based processing.