The name Nvidia comes from the Latin word for envy, ‘invidia’, and indeed some companies may envy the company’s dominance – along with AMD – in the graphics processing unit market. consumer (GPU).
Founded in 1993 by Jensen Huang (still the company’s CEO), Chris Malachowsky, and Curtis Priem, Nvidia is best known for making hardware that helps run PC and console games.
However, the basic silicon and semiconductor technologies used in these graphics cards can have a wide variety of applications. One of them is promoting computer systems that enable ADAS (advanced driver assistance systems) and autonomous driving, and respectively, Nvidia has made significant investments and advances in this area.
Today, Nvidia’s offerings in the autonomous vehicle and ADAS space can be grouped into four categories. These include software development and testing environments for autonomous vehicles, self-driving hardware and software, as well as a near-turnkey self-driving platform that combines the above products into a complete solution. tuning that automakers can buy to add autonomous driving features to their vehicles.
Test and development environment
Not all companies have access to the resources needed to test autonomous vehicles in a real-world environment, and there may be a variety of regulatory and safety hurdles that could also prevent them. do that.
As a result, many original equipment manufacturers (OEMs) and affiliated companies choose to test their self-driving hardware and ADAS in a virtual environment before hitting the road to make sure the fundamentals work in theory. theory.
Many autonomous systems and ADAS also rely on the development of neural networks, which can recognize various objects on the road, including cars, pedestrians, and animals, and predict the path that the road will take. they will go. However, to ‘train’ these networks to function correctly, they require significant input data sources, including test images and videos.
Nvidia offers two solutions to meet both of the needs described above. Nvidia’s drive infrastructure includes supercomputing hardware, software, and associated workflows to help OEMs and other companies train ADAS and autonomous driving neural networks, and includes Systems like the Nvidia DGX SuperPOD act as a turnkey supercomputer that companies can use to test these systems.
In addition, Nvidia also offers its Drive Sim, which the brand claims provides an accurate physics simulation platform that includes technologies such as the ‘Neural Reconstruction Engine.’
This is intended to bring data from the real thing directly into the simulation, by making it easy to clone recorded drives from a fleet of properly equipped vehicles in the simulation.
Self-driving hardware
In addition to providing OEMs and other developers access to resources for virtual testing of their ADAS and autopilot systems, Nvidia is also developing processing hardware that can be used in cars. to power these systems.
These are called SoCs, or systems on a chip, and integrate the CPU (Central Processing Unit), GPU, RAM, and other components on a single chip.
Nvidia’s Drive Orin is the brand’s most powerful SoC for autonomous driving available and started production in March of this year after it was first announced in December 2019.
The claimant claims the SoC can perform up to 254 trillion operations per second and uses 17 billion transistors seven times more powerful than its previous Xavier SoC for advanced driver assistance systems. Furthermore, the brand claims that the heavy use of Orin SoCs allows OEMs to scale their ADAS and autonomous driving systems from Level 2 to a fully automated Level 5 System.
More recently, Nvidia announced its Drive Thor SoC, which is expected to be available in vehicles manufactured from 2025. The company claims this represents a significant leap in performance. compute compared to the current Drive Orin, with a total performance of up to 2,000 teraflops.
Perhaps just as significantly, Nvidia claims the Thor is more than capable of powering the in-cabin infotainment system and digital instrument cluster, as well as other interior functions that are today distributed between various processors.
Accordingly, the company says that a future OEM could cut costs by allocating a portion of Thor’s computing power to support these interior functions (eliminating the need for a separate chip). and the rest for autopilot systems.
Self-driving software
While it is relatively easy for OEMs to buy powerful computer hardware and include it in their latest models, it is perhaps more difficult to develop software that can effectively leverage these systems to deliver provide customers with reliable, safe and efficient ADAS and automated driving systems.
Besides the hardware, Nvidia also provides the right software to take advantage of the SoCs it has developed, as well as handle input from other sensors such as radar, LiDAR, and cameras.
The foundation for this is the company’s Drive OS, which is a reference operating system that closely interfaces with hardware like the Orin and upcoming Thor SoCs. On top of that, Nvidia also provides software ‘layers’ like DriveWorks, which act as ‘middleware’ and include components like a sensor abstraction layer that can take input from different types of vehicle sensors .
The company has also developed a class of Drive Chauffeur software that combines multiple neural networks to combine cognitive, mapping, and planning functions. These help the vehicle estimate distances, detect and track objects, and control vehicle functions such as acceleration, braking, and lane positioning.
Due to regulatory and safety limitations, some ADAS systems also require the driver to continue to monitor the road ahead in order to function. To support this, Nvidia also offers Drive Concierge software that combines artificial intelligence and other technologies to assist with driver and occupant monitoring using the vehicle’s interior cameras and sensors. other furniture.
Self-driving platform
OEMs and other vendors may purchase just one or several of the components Nvidia has developed above and integrate it into the system from other vendors or those already built in-house. However, Nvidia also offers a largely complete self-driving platform that combines all of these components into a unified solution. This is called Nvidia’s Hyperion Drive.
The company describes Hyperion as an end-to-end, modular and reference architecture development platform for autonomous vehicle design, which also combines the Orin hardware and software described above. In the current version of Hyperion 8, it can support up to 12 exterior cameras, three interior cameras, nine radar sensors, 12 ultrasonic sensors as well as up to two LiDAR sensors.
A variety of automakers have announced that they will adopt Hyperion for their future vehicles. This includes Lucid’s DreamDrive Pro ADAS system (to be included in Lucid Air), some BYD electric vehicles from production in 2023, and Jaguar Land Rover vehicles to be released after 2025. Meanwhile, coming soon. Polestar 3 and the Volvo EX90 SUV will also use components from Nvidia’s Drive lineup.
Read the rest of the Inside Vendor series here:
