Nvidia (NVDA) is well-known for its autonomous vehicle efforts, and at the company’s GTC 2021 conference, it’s rolling out three technologies to support its future self-driving capabilities: Nvidia Drive Hyperion 8, Drive Chauffeur, and Drive Concierge.
Taken together, the technologies help Nvidia push deeper into the autonomous car space. What’s more, the technologies provide drivers and passengers with their own personal AI assistant while their car drives them down the street.
Drive Hyperion 8 combines a series of sensors including 12 cameras, nine radars, 12 ultrasonic sensors, and one front-facing lidar. The whole setup is meant to be modular so automakers can take and leave what they want.
Nivida says the system uses two Nvidia Drive Orin systems on a chip to offer redundancy and fail-over safety, as well as the potential for level four self-driving tech. Self driving capabilities are rated on a scale with level 0 being you controlling every aspect of the drive and level 5 meaning the car completely drives itself.
Nvidia’s Drive Chauffeur, for its part, is the company’s AI-assisted driving platform. It uses the sensors from Hyperion 8 and allows vehicles to drive from address to address on their own in both highways and urban environments. In addition to driving itself, Chauffeur can act as a high-end emergency intervention system for those who’d still rather drive on their own.
Then there’s Drive Concierge, Nvidia’s in-car AI assistant. Think of it as a high-powered Siri in your car. It’s designed to recognize each passenger and respond to them individually. So you’ll use your voice to control parts of your car that currently require you to twist knobs or tap touchscreens.
Concierge also acts as a virtual valet with the ability to find parking spaces and park without you inside. So you can jump out at the entrance to a fancy restaurant, like Wendy’s, and your car will take care of itself. When you’re finished with your spicy nugs, you can call your car back and it’ll come pick you up.
That’s not too far out of the realm of possibility now, either. I’ve demoed self-parking cars at the Consumer Electronic Show in past years, and they worked incredibly well. But real-world environments where everyone is racing to get the spot closest to the entrance are a completely different story.
Concierge will also work with Chauffeur to provide drivers with a view of what the AI driver is doing on the road, so you can sit back and watch as it works its magic.
Outside of Hyperion 8, Chauffeur, and Concierge, Nvidia has announced its new Omniverse Replicator for Drive Sim. That’s a lot of high-tech sounding words to say that Nvidia has created a virtual world using its Omniverse technology to help train AI models on potential driving situations that would otherwise require a lot of physical manpower or put humans in dangerous situations. The idea is to help AI models that could eventually be used in self-driving cars better understand how to drive through various scenarios.
To ensure that the synthetic data it gets out of Drive Sim Replicator is accurate, Nvidia uses its RTX path-tracing renderer to add real-world effects to its virtual vehicle sensors including LED flicker, motion blur, rolling shutter, and lidar beam divergence. This ensures that what the AI is seeing in the virtual Omniverse world, is as close to what it would see in the real world as possible.
While Nvidia’s progress in self-driving is certainly impressive, true self-driving cars are still years away from hitting the road. There are massive hurdles to overcome including how AI can safely navigate in climate weather, random construction zones, things like deer, and even reading the road in snow and rainstorms.
Even automakers that offer semi-autonomous features have been criticized for over selling their capabilities. Tesla in particular has been caught in the crossfire over its marketing of its Autopilot and Full Self-Driving modes, both of which require drivers to pay attention to the road and keep their hands on the wheel while on the road.
On Monday The Washington Post reported that a November over-the-air update for Tesla’s self-driving software made cars malfunction, causing them to slam their brakes while driving at highway speeds, forcing a recall.
The software has also been involved in investigations related to crashes while drivers were using self-driving mode.
Still, the advances that Nvidia and companies like it are making are proof that while self-driving isn’t a reality yet, it will be in the not-too-distant future.