Xpeng Autonomous Drive Results Fail to Impress

Paul Myles

TU-Editor

1st April 2021

Chinese automaker Xpeng has finally released data a full year after a ‘real-world’ autonomous driving expedition by a fleet of its production passenger vehicles.

It claims that a total of 1,820 miles of highway driving in full autonomous mode by its fleet of Xpeng P7 cars saw a human intervention rate of 0.71 per 62 miles. While the automaker hails the 2,200 mile excursion from Guangzhou to Beijing by around a dozen of its vehicles as a success, the intervention rate appears dangerously high compared to Waymo’s claimed human intervention rate of once every 5,996 miles during US testing – a figure itself roundly condemned as being too high to allow for mass adoption of the technology.

Nonetheless, the automaker trumpeted the abilities of its autonomous technology and the fact the XPeng P7 is China’s only mass production car with a 360-degree dual camera and radar fusion perception system. During the eight-day trip a year ago, the average success rate for autonomous lane changing and overtaking was 94.41%. The average highway ramp entering and exiting success rate was 92.76% and average tunnel pass-through success rate was 94.95% during the same period.

Invited auto journalists, EV enthusiasts and industry experts participated in the expedition, driving a fleet of 12 to 15 P7s, visiting 10 cities in six provinces along China’s eastern costal corridor. The route was said to be chosen to cover some of the most complex and representative road conditions and driving scenarios in China. It included rainstorm conditions, passing through tunnels, on highways in mountainous regions and including areas with no GPS signal. Mr He Xiaopeng, chairman and CEO of XPeng, said at a press briefing in Beijing: “We strive to become the world’s top autonomous driving hardware and software provider, and our strategy and R&D capabilities place us well in achieving this goal.”

— Paul Myles is a seasoned automotive journalist based in Europe. Follow him on Twitter @Paulmyles_

Components