Powering Safe, Autonomous Driving Solutions

Enhance Self-Driving Vehicle Performance with Low-Latency Computing.

Autonomous driving demands safety, and that requires a high-performance AI computing solution that processes sensor data with the highest accuracy. The NVIDIA DRIVE  Xavier and Pegasus platforms inference deep neural networks in real time. Deep learning–based perception, localization, and path planning enable the vehicle to understand its surroundings and operate safely.

But the technology doesn’t stop there. As autonomous vehicle solutions evolve, engineers continue to leverage a wide variety of deep learning frameworks to train new deep neural network (DNN) models in the data center. With support for every major framework, NVIDIA DRIVE gets smarter and smarter with over-the-air updates. Even after autonomous vehicles are in production, the platform accommodates new frameworks and models, enabling added capabilities and higher levels of autonomy.

Trailblazing Architecture


The NVIDIA DRIVE architecture for in-vehicle AI computing scales from DRIVE Xavier, a palm-sized, energy-efficient computer for Level 3/4 autonomy, to DRIVE Pegasus, a powerful AI supercomputer for Level 5 robotaxi applications. DRIVE Xavier is based on the world’s highest-performance system-on-chip, while DRIVE Pegasus delivers an unprecedented 320 trillion operations per second (TOPS).

Optimizing With Nvidia TensorRT


With NVIDIA's unified architecture, DNNs can be trained on NVIDIA® DGX Systems in the data center and deployed in autonomous vehicles for real-time inferencing with DRIVE Xavier or DRIVE Pegasus. The NVIDIA TensorRT programmable inference accelerator further optimizes the deep learning models. It rapidly validates and deploys trained DNNs to the automotive platform and accelerates inference for the production deployment of deep learning applications.

More Data, Faster Response


Over 370 companies are using NVIDIA DRIVE in the development of autonomous vehicles. One partner, TuSimple, increased inferencing performance by 30 percent after TensorRT optimization. “NVIDIA is unrivaled in delivering the computing performance needed for autonomous vehicles,” said Xiaodi Hou, Chief Technology Officer of TuSimple. Formed in 2015, TuSimple develops technology for autonomous long-distance freight delivery. The performance gains from TensorRT enable TuSimple to analyze additional camera data and add new AI algorithms to their autonomous trucks, within reasonable response time.


NVIDIA automotive solutions are available to automakers, truck makers, tier 1 suppliers, HD mapping companies, startups, and research institutions working on the future of transportation. Together, we are driving the next generation of innovation.