news

Autonomous vehicle tech is the biggest computing problem of its kind: Nvidia

The world’s leader in graphics processor unit manufacturing and the supplier of both software and hardware to the automotive industry for autonomous vehicles, says the computing challenge facing the industry is daunting.


Speaking to the media at today’s Consumer Electronic Show (CES) in Las Vegas, Nvidia boss Jensen Huang, said “building a computer for autonomous vehicles is at a level of complexity the world has never known.”

“In order to make autonomous vehicles possible, you have to solve this incredible computing problem, the largest scale computing problem of its kind from the bottom all the way to the top. This is a processor architecture problem, an algorithm problem, a system problem, a cloud computing problem, a car system and sensor problem," he said.

“The number of challenges necessary to solve, in order for the auto industry to bring autonomous vehicles to the world, is utterly daunting.”

Huang said that Nvidia has taken the first step by producing the world’s first processor built specifically for autonomous vehicles, called the Nvidia Xavier drive.

According to Huang, the hardware and software required to run autonomous vehicles needs to be “always on and never fail”, running software that “the world has never known how to write.”

According to Nvidia, the Xavier drive processor is the largest R&D effort the world has ever made for a processor.

The chip encompasses nine billion transistors and took eight thousand engineering years to design and engineer. It has multiple layers of built-in redundancy, eight cores, and dual execution so that each computation is carried twice for accuracy.

Nvidia’s recently announced Volta GPU (125 teraflops-per-second of computing power), is capable of recording every video input at full high-dynamic range in real time, while processing 1.5 gigapixels of imagery per second.

Nvidia says the chip will be used for level three and four autonomous vehicles, with the system to go into production before the end of this year.

So far, the likes of Volkswagen, Hyundai, Baidu and ZF have partnered with Nvidia to use its autonomous vehicle and artificial intelligence architectures. Uber will also become a big partner to Nvidia as it seeks to create an autonomous vehicle fleet for its services.

As for the holy grail, which is level five and fully autonomous driving capability, Nvidia is building its “Pegasus” system, which the company claims has enough processing power to allow the vehicle to face situations for which it has never faced before and operate safely in real time.

In order to test its hardware, Nvidia is using an automotive simulation system that allows its processors and their associated software to run in a virtual environment using real-world data (high definition video recordings as well as simulated environments). This allows engineers to retest the vehicle in the same scenario with updated hardware or software.

Nvidia says that it will aim to build a systems that is capable of Automotive Safety Integrity Level D (ASIL-D), which means it has the highest classification of initial hazard (injury risk) defined within ISO 26262. Otherwise, it’s the most stringent level of safety measures to apply for avoiding an unreasonable residual risk in the event of a system failure.

The GPU and chip manufacturers foray into the world of autonomous vehicles is only strengthening as time goes by. Companies from Toyota to Mercedes-Benz and Tesla all use Nvidia processing units for their autonomous vehicle systems, with the company positioning itself at the very forefront of the autonomous vehicle revolution.

Chat with us!







Chat with Agent