The Rise of Autonomous Vehicles

From Data Centers to Roadways

Autonomous vehicles are fundamentally nurtured in data centers, where their Deep Neural Networks (DNNs) learn to identify objects and interpret their environment. This is also where self-driving software undergoes rigorous testing and validation through extensive virtual simulations.

Once the DNNs are adequately trained and validated, the vehicles require sophisticated computing hardware and software to achieve true intelligence. Hardware-in-the-loop and cloud-based systems facilitate efficient, safe, and cost-effective testing across various scenarios — from everyday driving to extreme and potentially hazardous conditions — ensuring that autonomous solutions can be safely introduced to public roads.

A Comprehensive Platform for AV Development

The NVIDIA DRIVE platform provides a holistic solution for highly automated driving, transitioning from supervised to fully autonomous operations. It encompasses active safety features, automated driving, parking, and AI cockpit functionalities, spanning Levels 2+ to 5.

Key components of the NVIDIA DRIVE platforms include:

• An in-vehicle computer (DRIVE AGX).
• A comprehensive reference architecture (DRIVE Hyperion).
• An end-to-end simulation environment (DRIVE Sim).
• DNN training platforms (DGX).

These platforms are equipped with robust software development kits (SDKs) aimed at expediting the development of autonomous vehicles.

The NVIDIA DRIVE Level 2+ solution harnesses dual NVIDIA Orin systems-on-a-chip, integrating active safety, automated driving, parking features, and AI cockpit functionalities. Additionally, it incorporates the NVIDIA DRIVE Hyperion sensor suite, which equips developers with the tools to validate their AV platforms.

The DRIVE Hyperion sensor array features:

• Eight cameras
• Five radars
• Twelve ultrasonic sensors for comprehensive 360-degree scene interpretation
• Three interior sensing cameras for monitoring drivers and passengers

The NVIDIA DRIVE Software stack serves as a complete framework for crafting and deploying state-of-the-art AV applications, including:

• Perception
• Localization
• Mapping
• Planning and control
• Driver monitoring
• Natural language processing

The platform also includes the NVIDIA DRIVE OS, a secure operating system designed for accelerated computing, alongside the NVIDIA DriveWorks SDK, a versatile middleware solution.

The NVIDIA DRIVE AV and DRIVE IX stacks integrate DNNs and sophisticated algorithmic modules for perception, mapping, planning, and intelligent cockpit functionalities. The Level 2+ solution is trained and validated using the NVIDIA DRIVE Infrastructure, a precise end-to-end development framework built on a unified computing architecture.

This process begins with NVIDIA DGX systems, which streamline large-scale DNN training and optimization. Developers can utilize the power of GPUs and AI to train DNNs for various aspects of autonomous vehicle functionality, including perception, planning, and control.

Both NVIDIA DRIVE Constellation and DRIVE Sim provide virtual environments for testing and validating DNNs under diverse driving conditions, all while maintaining consistency with the hardware used in vehicles.

When paired with the DRIVE AV solution, the DRIVE Infrastructure establishes a continuous improvement cycle for development. This software-defined vehicle platform also allows automakers to introduce ongoing enhancements for users through over-the-air updates, prolonging customer satisfaction and paving the way for innovative business models.

In the previous year, NVIDIA unveiled the DRIVE AGX Orin, an advanced software-defined platform powered by the new Orin system-on-a-chip (SoC). This platform achieves 200 TOPS, nearly seven times the performance of its predecessor, the Xavier SoC, and can simultaneously manage a vast number of DNNs and applications operating within autonomous vehicles and robotics.

NVIDIA DRIVE Developer Day — A Conference on Autonomous Vehicles

For those eager to delve deeper into the technology and development surrounding autonomous vehicles, NVIDIA will host a significant virtual conference from April 20–22, 2021. This event will feature in-depth sessions focused on the creation of safe and reliable autonomous vehicles.

The NVIDIA DRIVE Developer Day promises to be an engaging experience for automotive developers, offering interactive sessions led by NVIDIA’s core automotive architects and engineers, covering every facet of the AV computational pipeline.

Current NVIDIA DRIVE customers will have exclusive access to specialized sessions about their own autonomous vehicle development processes. Attendees will be among the first to witness the latest advancements in AI-driven vehicles and discover the tools and platforms that contribute to safer, more efficient transportation solutions.

During the DRIVE Developer Day, participants will learn to utilize advanced learning tools to enhance DNN training within data centers. They will also explore the application of transfer learning and federated learning to streamline training while safeguarding proprietary data.

Moreover, attendees will gain insights into the DRIVE AV perception and DRIVE IX intelligent cockpit AI models available through the NGC, optimized for the DRIVE AGX platform. They will also learn to validate hardware and software in vehicles using the NVIDIA DRIVE Hyperion sensor developer platform, a sensor suite that can be retrofitted to vehicles and operates with DRIVE AGX hardware.

Each session will be facilitated by NVIDIA’s engineering teams, providing networking opportunities for one-on-one interactions with experts, giving participants an insider's perspective on AI-driven automotive solutions.

The NVIDIA DRIVE Developer Days will feature eight detailed sessions on the development of safe and robust autonomous vehicles, including:

• DRIVE OS & DriveWorks middleware
• DRIVE AV autonomous vehicle stack
• DRIVE Orin, Pegasus & Hyperion
• DRIVE Sim and Constellation for AV testing and validation
• NVIDIA DGX and EGX for DNN training

These sessions are complimentary, scheduled for live Q&A, and will also be available on-demand. For more information, check the list of sessions:

Additional Resources

If you are keen to learn more about autonomous driving technologies and opportunities, consider exploring the following articles:

• Analyzing The Autonomous Vehicle Governance Ecosystem: A Guide for Decision-Makers
• Measuring the impacts of Shared and Electric Autonomous Vehicles (SAEV) on urban mobility
• Autonomous Vehicles: a $50 billion opportunity for disruption
• Volvo Group and Aurora partnership on Autonomous transportation
• The impact of Autonomous Trucks on the Transportation industry
• Companies leading the autonomous truck race
• The impending arrival of Autonomous Vehicles
• The deployment of 5,000 autonomous trucks in China in 2021

References

• Introduction to Neural Networks and Deep Learning
• NVIDIA DRIVE AV Software
• NVIDIA DRIVE Infrastructure
• NVIDIA DRIVE Constellation
• NVIDIA DRIVE AGX
• NVIDIA DRIVE EGX
• NVIDIA DRIVE Hyperion
• NVIDIA DRIVE Sim
• NVIDIA Orin
• NVIDIA DRIVE Software stack for Autonomous Vehicles
• NVIDIA DRIVE OS
• NVIDIA DriveWorks SDK
• NVIDIA Driver IX
• NVIDIA DRIVE Level 2+

Stay updated on the latest developments in Autonomous Vehicles and Artificial Intelligence by following me on Medium.