During a CES 2016 keynote address, Nvidia CEO and cofounder Jen-Hsun Huang revealed that to ease the difficulties in developing self-driving cars, the company is launching a new platform for autonomous vehicles that would deliver the power of about 150 MacBook Pros.

Named the Nvidia Drive PX 2, the platform is made up of 12 CPU cores that are built onto a 16nm FinFET architecture and four chips with Pascal GPUs for eight teraflops of processing power and 24 deep learning tera operations per second.

The Nvidia Drive PX 2 is tagged as the most powerful engine for in-vehicle artificial intelligence in the world, and would allow companies developing self-driving cars to utilize artificial intelligence to solve the issues that are present in autonomous driving. Deep learning is utilized so that the self-driving vehicle will have a 360-degree situational awareness around itself, meaning that the car will be able to precisely determine its location and how to move forward in a safe path.

"Drivers deal with an infinitely complex world," said Huang. "Modern artificial intelligence and GPU breakthroughs enable us to finally tackle the daunting challenges of self-driving cars."

To illustrate the power of the platform, Huang said that the Nvidia Drive PX 2 requires water cooling to prevent it from overheating, as the platform's components are working really hard to release its all-world output. However, it will be able to perform under any conditions within any given environment.

According to Nvidia, the first automobile manufacturer that will be delivering self-vehicles utilizing the Nvidia Drive PX 2 will be Volvo, with more companies to follow.

Nvidia believes that there are currently two directions to take for the development of self-driving cars. The first path is the creation of auto-pilot systems that would require the participation of humans. The second path would be similar to the self-driving car project of Google, wherein there would be no need for humans to participate in the motion of the vehicle.

Huang said that solving the problems of self-driving cars to finally be able to release the vehicles en masse into the streets would be a great contribution to society. Self-driving vehicles will decrease the number of cars needed on the road, which would lead to more space as parking lots are transformed into other structures such as parks. In addition, autonomous vehicles will lead to fewer accidents.

The key to fulfilling this vision, according to Nvidia, is control. The company is looking for self-driving cars to be able to understand unpredictable and complex situations in real-life traffic, anticipate possible events and make the proper decision accordingly. The goal would be the creation of an end-to-end deep learning platform called a neural net. The deep learning platform is named Nvidia Digits.

The artificial intelligence that the Nvidia Drive PX 2 will enable would need to learn and train constantly, with new learnings being uploaded to the cloud as information that other vehicles will be able to access as well. This reference platform, created by Nvidia and named as Nvidia Drivenet, features a total of nine deep nested neural networks.

Nvidia demonstrated that the algorithm is able to identify vehicles and other objects on the road. In addition, it is able to detect abnormal circumstances such as roads without lines or school buses that the vehicle should not pass. Training the network takes a day due to the GPU acceleration found in the Nvidia Drive PX 2, as without it, training would require a month.

The first iteration of the Nvidia Drive PX 2, the Nvidia Drive PX, was delivered by the company last summer to over 50 car companies, suppliers, research institutions and developers that have adopted the artificial intelligence platform of Nvidia for the development of self-driving vehicles.

The companies praised the ease of development, capabilities and performance of the Nvidia Drive PX, and the company is looking to do even better with the Nvidia Drive PX 2. Other futuristic car tech will star at CES 2016, so stay tuned for more exciting news.

ⓒ 2024 TECHTIMES.com All rights reserved. Do not reproduce without permission.
Join the Discussion