Last year, Nvidia voiced their interest in the automotive industry as they demo-ed their Tegra-powered dashboard. At CES 2016, they have moved from merely powering dashboards to fully powering self-driving cars.
In today’s Nvidia press conference, Nvidia President and CEO Jen-Hsun Huang went on stage to present the company’s solution to the self-driving cars problem. That solution is the Nvidia Drive PX2, the world’s first in-car AI supercomputer. It is powered by 12 CPU cores, 4 chips with Pascal GPUs and 8 teraflops of processing capability. The device itself is roughly the size of a lunchbox and requires water cooling.
The key feature of the Drive PX2 is the fact that the AI is capable of Deep Learning. Deep Learning is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using multiple processing layers with complex structures, or otherwise composed of multiple non-linear transformations. In layman’s terms, the Drive PX2 will learn as it drives, making it more efficient over time; or at least, that’s the concept behind it.
Couple this self-learning AI with superhuman image recognition technology that Microsoft and Google has, and self-driving cars aren’t as farfetched as previously though. A demo shown of on stage includes a car driving down the road while highlighting potential hazards such as cars and humans, as well as street signs. Cameras are placed at almost all sides of the car, minimising the amount of blind spots the car has. The radar will also map out the car’s surrounding, representing them as pixels
According to Nvidia, Audi is currently using Nvidia GPUs in their cars and the feedback from them is encouraging. Nvidia also announced that Daimler and Ford will also be utilising this tech in their future cars. For now, the first car manufacturer that will utilise the Drive PX2 will be Volvo.
Source : Nvidia
