25+ Companies Developing Level 5 Robotaxis on NVIDIA CUDA GPUs

[btarunr]

NVIDIA today unveiled the world's first artificial intelligence computer designed to drive fully autonomous robotaxis. The new system, codenamed Pegasus, extends the NVIDIA DRIVE PX AI computing platform to handle Level 5 driverless vehicles. NVIDIA DRIVE PX Pegasus delivers over 320 trillion operations per second -- more than 10x the performance of its predecessor, NVIDIA DRIVE PX 2.

NVIDIA DRIVE PX Pegasus will help make possible a new class of vehicles that can operate without a driver -- fully autonomous vehicles without steering wheels, pedals or mirrors, and interiors that feel like a living room or office. They will arrive on demand to safely whisk passengers to their destinations, bringing mobility to everyone, including the elderly and disabled.



Millions of hours of lost time will be recaptured by drivers as they work, play, eat or sleep on their daily commutes. And countless lives will be saved by vehicles that are never fatigued, impaired or distracted -- increasing road safety, reducing congestion and freeing up valuable land currently used for parking lots.

Of the 225 partners developing on the NVIDIA DRIVE PX platform, more than 25 are developing fully autonomous robotaxis using NVIDIA CUDA GPUs. Today, their trunks resemble small data centers, loaded with racks of computers with server-class NVIDIA GPUs running deep learning, computer vision and parallel computing algorithms. Their size, power demands and cost make them impractical for production vehicles.

The computational requirements of robotaxis are enormous -- perceiving the world through high-resolution, 360-degree surround cameras and lidars, localizing the vehicle within centimeter accuracy, tracking vehicles and people around the car, and planning a safe and comfortable path to the destination. All this processing must be done with multiple levels of redundancy to ensure the highest level of safety. The computing demands of driverless vehicles are easily 50 to 100 times more intensive than the most advanced cars today.

"Creating a fully self-driving car is one of society's most important endeavors -- and one of the most challenging to deliver," said Jensen Huang, NVIDIA founder and CEO. "The breakthrough AI computing performance and efficiency of Pegasus is crucial for the industry to realize this vision.

"Driverless cars will enable new ride- and car-sharing services. New types of cars will be invented, resembling offices, living rooms or hotel rooms on wheels. Travelers will simply order up the type of vehicle they want based on their destination and activities planned along the way. The future of society will be reshaped," he said.

Broad Industry Support
Virtually all carmakers, transportation as a service companies, as well as startups are using NVIDIA AI in the development of Level 5 vehicles.

"Today dozens of companies are racing to develop robotaxis, but they are still gated by the massive computation needs of a truly driverless car," said Luca De Ambroggi, senior principal automotive analyst at IHS Markit. "The new NVIDIA DRIVE PX Pegasus shows the path to production for the automakers, startups and automotive ecosystem working to deliver this amazing vision."

Product Specifications
NVIDIA DRIVE PX Pegasus is powered by four high-performance AI processors. It couples two of NVIDIA's newest Xavier system-on-a-chip processors -- featuring an embedded GPU based on the NVIDIA Volta architecture -- with two next-generation discrete GPUs with hardware created for accelerating deep learning and computer vision algorithms. The system will provide the enormous computational capability for fully autonomous vehicles in a computer the size of a license plate, drastically reducing energy consumption and cost.

Pegasus is designed for ASIL D certification -- the industry's highest safety level -- with automotive inputs/outputs, including CAN (controller area network), Flexray, 16 dedicated high-speed sensor inputs for camera, radar, lidar and ultrasonics, plus multiple 10Gbit Ethernet connectors. Its combined memory bandwidth exceeds 1 terabyte per second.

NVIDIA DRIVE PX Platform
The NVIDIA DRIVE PX platform scales from a single mobile processor configuration delivering Level 2+/Level 3 capabilities to a combination of multiple mobile processors and discrete GPUs for full Level 5. These configurations run on a single, open software architecture. This enables automakers and tier 1 suppliers to move from development into production for a wide range of self-driving solutions -- from AutoCruise on the highway, to AutoChauffeur for point-to-point travel, to Pegasus for a fully autonomous vehicle.

NVIDIA DRIVE PX is part of a broad family of NVIDIA AI computing solutions. Data scientists who train their deep neural networks in the data center on the NVIDIA DGX-1 AI supercomputer can seamlessly run on NVIDIA DRIVE PX inside the vehicle. The unified architecture enables the same NVIDIA DRIVE software algorithms, libraries and tools that run in the data center also perform inferencing in the car.

This cloud-to-car approach enables cars to receive over-the-air updates to add new features and capabilities throughout the life of a vehicle.

Availability
Pegasus will be available to NVIDIA automotive partners in the second half of 2018. NVIDIA DriveWorks software and NVIDIA DRIVE PX 2 configurations are available today for developers working on autonomous vehicles and algorithms.

View at TechPowerUp Main Site

 

[Ferrum Master]

Good, good...

 

[R00kie]

As long as they don't drive me off a cliff, I'm fine with this.

 

[Steevo]

Update to the drivers while doing 80 on the interstate that causes overheating and failure. Nvidia claims no fault. "Obviously this was not made for those roads" claimed the tech giant, and lack of secondary systems in these new cards will be fixed, with Driver SLI, a NEW feature, so they can sell twice as many and have failover redundancy.


Also in the news, is your older Nvidia Pilot driving into curbs.... it might be in need of an upgrade, after the latest Nvidia Pilot optimization for new cars, the oldest ones have been slowed by new software despite functioning fine, and a reminder, any changes to the software will result in your car driving you to the nearest detention facility for tampering with Nvidia DriveWorks.

 

[punani]

Cool stuff but I can already imagine how fragile this system would be..

1. Fly hits sensor lens
2. SYSTEM FATAL ERROR sensor input fault detected
3. Ejector seat activated

 

[R0H1T]

As long as they don't drive me off a cliff, I'm fine with this.

Don't bet on that, also don't ever fall asleep with that bot controlling the wheel, always keep a window breaker with you at all times, so on & so forth.

 

[bug]

Am I the only one that read "robot-axis" (and thought wth)?

 

[wiyosaya]

Don't bet on that, also don't ever fall asleep with that bot controlling the wheel, always keep a window breaker with you at all times, so on & so forth.

What wheel?

Makes me feel really comfortable!

 

[Supercrit]

Oh by the way most crashes with self driving cars are caused by humans at fault.

 

[Steevo]

Oh by the way most crashes with self driving cars are caused by humans at fault.

What happens when the human coded software crashes, or a human made sensor fails?


I install GPS systems that drive big machines, it happens more often than you think.

 

[Divide Overflow]

Cool stuff but I can already imagine how fragile this system would be..

1. Fly hits sensor lens
2. SYSTEM FATAL ERROR sensor input fault detected
3. Ejector seat activated

Or from they way they handled Physx co-processing:
Warning: AMD / ATI device detected in subsystems / cargo. Shutting all systems down.

 

[R0H1T]

What wheel?

Makes me feel really comfortable!

Steering wheel, don't think they're going iRobot or fifth element anytime soon.

AI assisted vehicles probably would be much safer, instead of giving them full control.

 

[Mirkoskji]

At least the AI wasn't awake all night long last turn to work extra hours, with loosen reflexes or had a bad day...
Or takes a minimal wage so it doesn't care of the people who is carrying around.
or took drugs, or was reading that message on whatsapp, or was yelling at that pirate so it didn't see that child on the crosswalk..
Or who cares of that redlight, i hate cyclists.
Or that person on the road seems cristian/gay/islamic/atheist/too brown in complexion, and i hate him...
Or i am 69 years old and still drive even if i see s**t.

I mean, there's a pretty high probability that the next driver you'll encounter on the road is a total asshole..

 

[Steevo]

Steering wheel, don't think they're going iRobot or fifth element anytime soon.

AI assisted vehicles probably would be much safer, instead of giving them full control.

This brings up an interesting question and moral issue.

If 3 AI driven cars were going to wreck in a way that either all 3 passengers would get hurt, or 1 would die, 1 hurt, and 1 fine what outcome would be chosen? Who would get to make that decision? Ultimately it would be the coder, or anyone capable of cracking that system.

 

[wiyosaya]

Steering wheel, don't think they're going iRobot or fifth element anytime soon.

AI assisted vehicles probably would be much safer, instead of giving them full control.

But the article says that they are developing fully autonomous cars that have no steering wheel, brakes, or accelerators...

NVIDIA DRIVE PX Pegasus will help make possible a new class of vehicles that can operate without a driver -- fully autonomous vehicles without steering wheels, pedals or mirrors, and interiors that feel like a living room or office. They will arrive on demand to safely whisk passengers to their destinations, bringing mobility to everyone, including the elderly and disabled.

IMO, this article reads like a marketing blurb.

 

[R0H1T]

But the article says that they are developing fully autonomous cars that have no steering wheel, brakes, or accelerators...



IMO, this article reads like a marketing blurb.

That's the scary part, I wouldn't want my life solely in the hands of God or AI, interchangeable in this case!

This brings up an interesting question and moral issue.

If 3 AI driven cars were going to wreck in a way that either all 3 passengers would get hurt, or 1 would die, 1 hurt, and 1 fine what outcome would be chosen? Who would get to make that decision? Ultimately it would be the coder, or anyone capable of cracking that system.

Also who gets sued in the event of a crash, Apple's AI or Alexa, from Amazon

 

[bug]

Also who gets sued in the event of a crash, Apple's AI or Alexa, from Amazon

A ticket needs to be opened first for investigation. And I'm only half-joking.

 

[LFaWolf]

I don't know why there is so much skepticism over autonomous vehicles. First off, I am a software developer, and I understand creating fail proof software is very, very difficult. It requires tremendous amount of commitment and dedication from the company and developers. Even with that, there are still very small chances of crashes. Actually the goal here is through iterations and testing, to have a lower accident rate than human drivers

That said, today on my way home, a driver from Sears service was tailgating me, then impatiently cut over to the right most lane, then abruptly cut and forced himself over back to the left most lane. Mind you on that 3-lanes freeway, the speed limit was 70 mph and every car in the left most lane was already going 75 mph. I was maintaining a 2-car space behind the vehicle in front of me. The left most lane was already packed with cars. His action caused everyone from the middle, right most, middle again, and then all vehicles in front of me in the left most lane to hit his/her brakes abruptly. All that to gain 5 car spots in front of me. I almost called Sears to complain about his dangerous driving.

I don't drive as much as I did in my previous job, where my daily commute was in excess of 60 miles round trip. However, I saw driver similar to his driving every single day. Yup, every single day on US highway. I will take an autonomous vehicle over his driving any given day. That includes drunk drivers as well.

In addition, this is not the drivers software for your nVidia or AMD GPUs. The commitment to perfection is not necessary when creating the GPU drivers, and I don't care if nVidia or AMD come to me and say "No no, that is not what we commit to". Why? Because if your GPU crashes during the game, so what? No one dies. Just a restart and an annoyed user. Even Garmin/Navteq software does not require the sophistication required for autonomous vehicles.

Humans have been driving cars for more than a century now(?) I say the time has come for autonomous driving.

 

[Divide Overflow]

Humans have been driving cars for more than a century now(?) I say the time has come for autonomous driving.

 

[LFaWolf]

I am pretty sure you have no idea what I wrote, read one line and immediately searched for that image. Not everything in life can be illustrated by an analogy. Is this all TPU can offer these days? Autonomous driving will be a lot safer than you think.

 

[R0H1T]

I am pretty sure you have no idea what I wrote, read one line and immediately searched for that image. Not everything in life can be illustrated by an analogy. Is this all TPU can offer these days? Autonomous driving will be a lot safer than you think.

Not if they remove any human override ~

fully autonomous cars that have no steering wheel, brakes, or accelerators...

Will it be safe ~ yes, but would be more safe with a fail safe option like forced human intervention or override, just my $0.02

 

[Mirkoskji]

I see that many think in "human terms".
"what if there is this extreme situation x where a certain number of vehicles have to make a choice to save someone and let die someone".. ecc ecc
The thing is that with a machine (and even better when there only is a float of automated vehicles on the road), the behaviour of the vehicle will be calibrated to avoid these situations becoming real.
Most of these problems arise because normal humans have a pretty bad predictive capability, and they put themselves into riskful situations.
A machine will simply avoid any riskful situation, because it is programmed to do so, well before a human brain starts to understand that there will potentially be a problem.
A brain is a bad calculator that makes lots of errors.
It is good at speculation and creativity, it is bad at mechanical tasks. Driving is a mechanical task (when you're not racing) and a conscence-free program is the best driver you will find on a road.

 

[bug]

I see that many think in "human terms".
"what if there is this extreme situation x where a certain number of vehicles have to make a choice to save someone and let die someone".. ecc ecc
The thing is that with a machine (and even better when there only is a float of automated vehicles on the road), the behaviour of the vehicle will be calibrated to avoid these situations becoming real.
Most of these problems arise because normal humans have a pretty bad predictive capability, and they put themselves into riskful situations.
A machine will simply avoid any riskful situation, because it is programmed to do so, well before a human brain starts to understand that there will potentially be a problem.
A brain is a bad calculator that makes lots of errors.
It is good at speculation and creativity, it is bad at mechanical tasks. Driving is a mechanical task (when you're not racing) and a conscence-free program is the best driver you will find on a road.

You're right, of course, but AI cannot control everything. It cannot control a landslide suddenly ending your lane 2 meters ahead or a loose rock falling in the middle of the road. At the same time, I'm pretty sure even in these extreme situations it can react better than a human.

But don't think about this as an engineer, think about it as lawyer: an accident happens and despite AI's best effort, a life is lost. A lawyer will have someone look at the decision making software and then conclude that if the condition in a "if" was different, that life could have been saved (never mind that this could in turn mean lives lost that under current circumstances wouldn't happen). How do you argue with that in court?

*insert mandatory 100 lawyer at the bottom of the ocean joke here*

 

[wiyosaya]

Not if they remove any human override ~
Will it be safe ~ yes, but would be more safe with a fail safe option like forced human intervention or override, just my $0.02

That is exactly my concern. I want a human override of some sort. It will take a long time to perfect this whether we like it or not. I agree that there is significant potential for improved safety, however, until all vehicles on the road are autonomous and they communicate with each other, the unpredictability of the human driving the other car is a major concern.

Not to mention that there have been times when GPS data is incorrect. Or how about construction that somehow does not make it into the construction database for example, a bridge out? There are many points beyond the system internal to the vehicle at which humans can induce error, and I am willing to bet that those points will be the most problematic at first.

And one other thing, what happens if the system locks up while moving and there is no human override?

 

[Mirkoskji]

You're right, of course, but AI cannot control everything. It cannot control a landslide suddenly ending your lane 2 meters ahead or a loose rock falling in the middle of the road. At the same time, I'm pretty sure even in these extreme situations it can react better than a human.

But don't think about this as an engineer, think about it as lawyer: an accident happens and despite AI's best effort, a life is lost. A lawyer will have someone look at the decision making software and then conclude that if the condition in a "if" was different, that life could have been saved (never mind that this could in turn mean lives lost that under current circumstances wouldn't happen). How do you argue with that in court?

*insert mandatory 100 lawyer at the bottom of the ocean joke here*

Of course there is the legal question. the thing is, even with humans, you have cases when an accident caused by a human driving a car ends with an absolution because the fact was out of control. an automated car will have sensors and cameras at every angle, so every action will be recorded. when an out of control situation happens the sensors and the cameras will capture it and then a court will decide about that, taking decisions based upon the data. Of course it is an unprecedented case scenario, and our law system will have to adapt. but even in the worst situation that court will have completely objective data to take a decision on. Now tell me, when was the last time a witness, or a guilty or an innocent has been an objective source of informations about facts?
I think the worst period will be when, in the future, there'll be mixed traffic on the road, with many man driven cars, but also a lot of automated cars. it will be the perfect recipe for a total mess, also on the legal side.
I also think there will be a division between courts with a preference towards automation and others that will be more conservative, jurisprudence will be split in half in common law systems. As of civil law systems, like the one in my country, i think they'll remain stuck for tens of years before adapting..