Nvidia unveils new self-driving taxi hardware

Called Pegasus, the AI platform is touted as the world's first artificial intelligence computer created to control fully autonomous "robotaxis".

The self-driving computer is built on Nvidia's CUDA GPUs, and also increase the practicality of the computing hardware requirements for actually fielding self-driving cars on real roads.

Pegasus is set to be available to Nvidia partners during the second half of 2018, while companies can get started now with Drive PX 2 if they want to chart a path to upgrade later on. Level 5 autonomous driving is technically possible today - albeit it's not dimensionally ready for commercial operation. Under the SAE International Standard J3016, Level 5 autonomy is described as "the full-time performance by an automated driving system of all aspects of the dynamic driving task under all roadway and environmental conditions that can be managed by a human driver".

NVIDIA co-founder and CEO, Jen-Hsun Huang said during GTC Europe 2017: "Creating a fully self-driving vehicle is one of society's most important endeavors - and one of the most challenging to deliver". The company claims that the Pegasus hardware will aid them in their work: "The breakthrough AI computing performance and efficiency of Pegasus is crucial for the industry to realise this vision", said Nvidia's founder and CEO, Jensen Huang.

"NuTonomy is building for Level 5 and Pegasus is the kind of platform that will be required to support these types of systems."

The new Drive DX Pegasus system is powered by two of Nvidia's newest Xavier SoC processors (with an embedded GPU based on the Nvidia Volta architecture) alongside two next-gen discrete GPUs. The system will provide the enormous computational capability for fully autonomous vehicles in a computer the size of a license plate, drastically reducing energy consumption and cost. This means it will cope with a terabyte of bandwidth each second - the chip bundles four AI processors working together, and there are connections for 16 high-speed sensors like cameras, radar, lidar, or ultrasonic units.

Vanessa Coleman