At GTC 2022, NVIDIA announced its new innovations in omniverse, automotive, healthcare and robotics, ranging from bots to autonomous vehicles. In the automotive industry, NVIDIA strives to accelerate innovation in autonomous vehicles, infotainment systems, and graphics simulation in the omniverse and introduced new systems and technologies in these areas.
Read: Nvidia GTC 2022 takes place. Here’s what to expect
“People never stop learning. This is exactly how we need to envision an AI car.” The pipeline to create an autonomous vehicle from perception to mapping to localization continues to be an endless process. NVIDIA’s Safety First policy has rated them with ISO 26262 security, where the development of sensors and algorithms in their automotive products is safe and secured by 15,000 engineers in 5 million lines of code.
“Advancements in accelerated computing and AI are progressing at the speed of light,” said Jensen Huang, NVIDIA founder and CEO.
NVIDIA’s DRIVE is an end-to-end platform for the development and deployment of autonomous vehicles. Founded in 2015, DRIVE provides solutions for the transportation industry using deep learning networks.
NVIDIA DRIVE introduced Thor, a centralized in-car computing system that includes infotainment, automated driving, and a cost-saving system. In 2018, the Thor chip Parker had a performance of one teraflop. Now, the Thor superchip achieves 2,000 teraflops of performance with FP8 precision, which enables the transition from 32-bit FP data to 8-bit integer format and packs all the smarts into a single architecture for great efficiency.
With 77 billion transistors, the architecture leverages Hopper GPU, Next-Gen GPU, and Grace GPU, allowing simultaneous processors to run without interruption. All automotive companies will be able to implement a single architecture for all their AI and technology installations in their vehicles by 2025, starting with Geely-owned automaker ZEEKR for its next-generation electric vehicles.
Thor is the first AV platform to include the Inference Transformer Engine, a component of Tensor Cores in NVIDIA GPUs that will accelerate the inference performance of deep neural networks by 9x, essential for supporting self-driving workloads is of utmost importance.
DRIVE Thor replaces DRIVE Atlan and is a successor to DRIVE Orin currently in production in most 254 powered vehicles. The superchip also uses NVLink C2C to enable automakers to calculate headroom and flexibility and help build software-based vehicles.
The in-vehicle infotainment (IVI) has been upgraded with the NVIDIA DRIVE concierge platform and offers features beyond music and directions. The concierge platform offers AI-enabled security, comfort and entertainment features for each passenger of the vehicles with personalized cluster capabilities.
With the cross-domain platform DRIVE Thor, Concierge can host multiple virtual machines with a single chip for simultaneous and streamlined development. To make the console user-friendly, the platform supports Android Automotive operating system for customization.
In addition, the digital cockpit also enables premium functionality throughout the vehicle with video conferencing functions, digital assistants and even games. In addition to the entertainment features, Concierge is also integrated with the DRIVE Chauffeur platform, which offers 4D visualization and 360-degree view for passenger safety and comfort.
NVIDIA is also taking the next step by using the software capabilities to integrate in-vehicle gaming systems and stream them to any screen. The company uses its GeForce NOW cloud gaming service to stream 1,400 titles with no download and unlimited cloud storage. With the DRIVE AGX compute platform, the DRIVE IX software stack, and the Omniverse Avatar Cloud Engine (ACE), NVIDIA makes in-vehicle times more enjoyable.
When it comes to safety, Concierge uses cabin sensors and deep neural networks for driver monitoring to ensure attention when humans are in control. In addition, Omniverse ACE develops easily customizable, interactive AI assistants using speech AI, computer vision, and natural language understanding. These created avatars can also make recommendations, provide alerts, and access vehicle controls.
OVX Computing Systems for the automotive industry
NVIDIA announced the second generation of OVX, which is based on the next-generation GPU architecture and offers real-time graphics, AI, and simulation capabilities for digital twins. The OVX systems use 3D applications to create 3D virtual worlds to create digital twins in NVIDIA Omniverse Enterprise.
Integrating OVX systems into the automotive industry would enable photorealistic creation of omniverse twins and develop simulations leading to inexpensive and simpler design and production processes. XPENG’s G9 also includes NVIDIA’s Orin software for on-highway and off-road driving and parking functions.
NVIDIA OVX with DRIVE Sim, which integrates a neural engine for Earth-scale scene reconstruction, enables developers to simulate a wide range of real-world driving scenarios to efficiently and safely test autonomous vehicles. This also allows vehicle features and functions to be replicated for demonstration to customers.
BMW and Jaguar Land Rover will initially equip the OVX systems. “Using this technology to create large volumes of lifelike, physically accurate scenarios in a scalable and cost-effective manner will accelerate our progress toward our goal of a future with no accidents and less congestion,” said Alex Heslop, director of electrical, electronics and software development. Jaguar Land Rover. WPP is also building a suite of services for automotive customers using the omniverse cloud, saving a ton of costs.
The OVX system is powered by an L40 GPU and 4th generation Tensor Cores, which together with ConnectX-7 SmartNIC, which improves network and storage performance, enables high-performance RTX workloads.