Nvidia Corp. tonight bid to expand the reach of artificial intelligence systems into autonomous vehicles and robots with the launch of its new Nvidia DRIVE AGX Orin platform.

Nvidia DRIVE AGX Orin is a software-defined platform that’s powered by a new system-on-a-chip called Orin. The new chip is said to be the fruit of “billions of dollars of R&D investment” over four years, and packs an impressive 17 billion transistors.

It’s integrated with Nvidia’s most advanced graphics processing units and its Arm Hercules central processing units, as well as deep learning and computer vision accelerators. All told, it’s capable of delivering 200 trillion operations per second, making it seven times faster than its previous generation Xavier SoC. The Orin chips can also be programmed for specific computing tasks via open CUDA and TensorRT application programming interfaces.

The DRIVE AGX Orin platform is meant to enable the numerous AI-based applications and deep neural networks necessary to power both autonomous vehicles and robots.

“Creating a safe autonomous vehicle is perhaps society’s greatest computing challenge,” Nvidia founder and Chief Executive Officer Jensen Huang said in a statement. “The amount of investment required to deliver autonomous vehicles has grown exponentially, and the complexity of the task requires a scalable, programmable, software-defined AI platform like Orin.”

Nvidia said the platform is already being used by Didi Chuxing Technology Co., a Chinese mobile transportation company that offers services similar to those of Uber Technologies Inc., to develop its own autonomous driving capabilities.

Didi said it’s using the DRIVE AGX Orin platform to perform inference on Level 4 autonomous driving vehicles. The company also uses Nvidia’s latest GPUs in its data centers to train its machine learning algorithms.

Through those efforts, Nvidia said, it can fuse data together from various in-car sensors, including cameras, lidar and radar, and create deep neural networks that can understand the 360-degree environment surrounding its vehicles and help them to navigate the roads safely.

“Developing safe autonomous vehicles requires end-to-end AI, in the cloud and in the car,” said Rishi Dhall, vice president of autonomous vehicles at Nvidia. “Nvidia AI will enable DiDi to develop safer, more efficient transportation systems and deliver a broad range of cloud services.”

To boost adoption of the new platform further, Nvidia said, it’s giving transportation industry customers free access to its collection of Nvidia DRIVE deep neural networks for autonomous vehicle development. Developers can now access a range of pre-trained AI models and training code, which they’re free to customize to “increase the robustness and capabilities” of self-driving systems.

In related news, Nvidia also announced the availability of new inference software for creating conversational AI applications.

The new software, called Nvidia TensorRT 7, is aimed at eliminating the latency problems that have so far prevented true interactive engagement with conversational AI. Developers will be able to create smarter interactions with voice agents, chatbots and recommendation engines in true real time, Nvidia said.

Nvidia TensorRT 7 will be most useful in the development of recurrent neural networks, which power a new breed of AI models that can make predictions on “time-series, sequence-data scenarios,” the company said. RNNs, as they’re known for short, can help with tasks such as arrival time planning for cars and satellites, the prediction of events in electronic medial records, financial asset forecasting and fraud detection, Nvidia said.

Nvidia said the TensorRT 7 inference software will be made available without charge to members of the Nvidia developer program in “the coming days” via its GitHub repository.

Images: Nvidia