On April 20, the first fleet of L4-level Robotaxi vehicles from Pony.ai officially landed in Shenzhen. The fleet consists of 30 vehicles, which will be put into operation in Pony.ai’s Robotaxi service in Shenzhen.
The fleet adopted Pony.ai’s latest L4-level autonomous driving solution, DeepRoute-Drive 2.0, which is designed for mass production. The hardware configuration includes:
- 2-5 LiDARs;
- Surround-view vision system composed of 8 cameras;
- Orin chips (two, based on last month’s announcement).
Pony.ai claims that DeepRoute-Drive 2.0 is currently the lowest-cost L4-level autonomous driving solution in the industry. Currently, the system has been successfully retrofitted on vehicles such as the FEV MARVEL R, Lincoln MKZ, and the Geely Geometry A.
The accurate perception distance of DeepRoute-Drive 2.0 exceeds 200 meters. To ensure safety, Pony.ai also designed a redundant visual perception system to take over the vehicle and maintain autonomous driving when some of the sensors fail.
Pony.ai CEO James Peng said: “To develop safe and reliable technology and achieve mass production of autonomous driving, we still need to solve many problems. One of the key issues is how to reduce costs and increase efficiency. The cost of DeepRoute-Drive 2.0 is currently less than $10,000. Based on ensuring the reliability of the system, after large-scale cooperation with automakers, the cost of the entire autonomous driving system can be further reduced by 70%.”
Last month, Pony.ai officially announced its cooperation with NVIDIA to adopt the Orin system-level chip and develop a vehicle-level solution for mass production of L4-level autonomous driving.
At the beginning of 2020, Pony.ai adopted the NVIDIA Xavier to create a low-power computing platform, reducing the cost of the autonomous driving computing platform to half that of traditional solutions, and significantly reducing the volume relative to traditional solutions.
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.