NVIDIA unveiled the world's first artificial intelligence computer designed for the development of fully autonomous robot taxis at 'GTC Europe,' which is taking place in Munich from the 10th to the 12th (local time).

Codenamed Pegasus, this new system is an extended version of the NVIDIA DRIVE PX AI computing platform for processing Level 5 autonomous vehicles, capable of 320 trillion operations per second, providing more than 10 times better performance than the previous version, NVIDIA DRIVE PX 2.

Accordingly, the NVIDIA Drive PX Pegasus does not require a driver to be on board. There is no steering wheel, pedals, or mirrors for a driver. Additionally, since it arrives at a destination according to the passenger's needs, it can be used by everyone, including the elderly and the disabled.

Drivers can work, rest, eat, and sleep even while commuting, allowing them to make new use of the time they previously wasted. In addition, it can prevent traffic accidents caused by driving fatigue, injury, or distraction, enhance road safety, alleviate congestion, and enable the utilization of land currently used as parking lots.

Among the 225 partner companies currently developing based on the NVIDIA Drive PX platform, more than 25 are developing fully autonomous robotaxis using NVIDIA CUDA GPUs. The trunks of the vehicles currently under development resemble small data centers, housing racks of multiple computers equipped with server-grade NVIDIA GPUs, where deep learning, computer vision, and parallel computing algorithms are executed. However, due to their size, power demands, and costs, it is still difficult to apply them to mass-produced vehicles.

Robot taxis require massive computational processes. They must perceive their surroundings using high-resolution 360-degree omnidirectional cameras and lidar, determine their own position with centimeter-level accuracy, and track the movements of surrounding vehicles and people to plan a safe and comfortable route to their destination. Furthermore, all of these processes must be repeated across multiple stages to ensure the highest level of safety. Autonomous vehicles require at least 50 to 100 times more intensive computing power compared to today's state-of-the-art vehicles.

Jensen Huang, founder and CEO of NVIDIA, stated, “The development of fully autonomous vehicles is one of the most socially significant endeavors and one of the most difficult challenges to implement,” adding that “Pegasus’s groundbreaking AI computing performance and efficiency are essential for the industry to realize this vision.”

He continued, “Driverless vehicles will enable the emergence of new ride-sharing and car-sharing services. New types of vehicles will appear that feel like an office, a living room, or a hotel room. Travelers will simply need to call for the type of vehicle they want, depending on their destination and the activities they intend to do along the journey. The future of our society will be completely different.”

Luca De Ambroggi, principal automotive analyst at IHS Markit, stated, “Today, dozens of companies are competing to develop robotaxis, but the massive computational demands required to realize truly autonomous vehicles are a stumbling block for them,” adding, “The new NVIDIA Drive PX Pegasus is paving the way for mass production for automakers, startups, and the automotive industry ecosystem striving to realize the fantastic vision of autonomous vehicles.”