Nvidia: Next-Gen 'Pegasus' AI Will Enable Level 5 Self-Driving Robotaxis

At GTC Europe, Nvidia announced world’s first computer for Level 5-capable “robotaxis.” The new system, called “Pegasus,” is an extension of Nvidia’s Drive PX system for autonomous cars, but with 10x the performance of its predecessor, Drive PX 2.

Level 5 Autonomy

Cars that can change lanes and decelerate on their own are Level 2. An example of this would be Tesla’s Autopilot. Level 3 would be allowing the car to take over during certain driving scenarios such as being in traffic. All levels up until Level 4 will still require the driver to be attentive in case the car can’t handle a certain driving scenario.

Ford, Volvo, and Waymo have already said that they'll skip Level 3 because that’s a level where drivers may put too much trust into the cars’ systems, which may put their lives in danger. Level 4 is when a car can do a full trip on normal road conditions, but it may not properly handle itself in certain scenarios such as driving on dirt roads.

Level 5 is when a car should be able to drive as well or better than a human in any driving environment. Nvidia believes its Pegasus computer for self-driving cars will enable Level 5 for “robotaxi” vehicles, upon which you should be able to call to drive you to your destination, all without any human driver.

Nvidia's New AI Computer For Robotaxis

According to Nvidia, the new Pegasus system is as small as a taxi license plate, but can do 320 trillion operations for second. The company said it can achieve this level of performance due to a combination of factors including the already announced Xavier SoC, which will include a powerful GPU based on Nvidia’s next-generation graphics technology (the generation following Volta).

Nvidia hasn’t revealed any information yet about its new GPU technology, but we can also expect it to bring significant improvements to machine learning hardware such as the Tensor Cores, too. It’s also possible the Pegasus AI computer will be built on a 7nm process, which may explain a large portion of the 10x jump in performance that Pegasus offers over Drive PX 2.

Nvidia believes that this class of performance will be able to achieve “Level 5 autonomy” for self-driving cars, which means vehicles that will not have a steering wheel, pedals, or mirrors. Because of this, the interior could also look like a living room or an office, and not like your typical car interior.

The robotaxis that will use Nvidia’s Pegasus system will stream data from up to 16 cameras, ultrasonic sensors, radar, and LIDAR, all enabled by more than 1 TB/s bandwidth. The car’s self-driving systems will be trained via driven miles but also through simulation and re-simulation. In a re-simulation, an autonomous car simulates driving the same roads it has driven before but under different simulated conditions (at night, in rain, while snowing, etc).

Nvidia noted that Pegasus comes with multiple fail safes and redundancies and is ASIL D certified (highest safety level). The company added that the Pegasus system will be available to customers in the second half of 2018.

“Creating a fully self-driving car is one of society’s most important endeavors — and one of the most challenging to deliver,” said Jensen Huang, Nvidia founder and CEO. “The breakthrough AI computing performance and efficiency of Pegasus is crucial for the industry to realize this vision. Driverless cars will enable new ride- and car-sharing services. New types of cars will be invented, resembling offices, living rooms or hotel rooms on wheels. Travelers will simply order up the type of vehicle they want, based on their destination and activities planned along the way. The future of society will be reshaped,” he said.