If it moves forward smoothly, Nio will be the fastest among Chinese companies to mass-produce end-to-end features, LatePost said, citing a person close to the company.
(Image credit: Nio)
Several new Chinese car-making forces are following in the direction of Tesla’s FSD Beta V12 and are actively investing in the development of autonomous driving architecture models based on end-to-end neural networks, local media outlet LatePost said in a report today.
Nio (NYSE: NIO) will launch active safety features based on end-to-end technology in the first half of this year, the report said.
“If it moves forward smoothly, Nio will be the fastest among Chinese companies to mass-produce end-to-end features. To that end, a team of nearly 100 people at Nio has been working on it for nearly half a year,” LatePost quoted a person close to the electric vehicle (EV) maker as saying.
Li Auto (NASDAQ: LI), which will launch its new model in the first half of this year, is hiring heavily for smart driving talent, according to the report.
On January 30, Xpeng’s (NYSE: XPEV) chairman and CEO He Xiaopeng said the company’s end-to-end model will be used across the board in vehicles.
Replacing legacy technologies with end-to-end architectures is becoming a new race in China’s smart driving space this year, the report noted.
For smart driving systems, modules such as perception, prediction, decision-making, and control all need to be handled by engineers in different fields.
Some companies choose to rely on programmers to write rule-based code to solve massive problems, such as Huawei, which has over 1,000 engineers in its planning and control team alone, the LatePost report noted.
An end-to-end smart driving system could alleviate this problem by taking sensor data as input and using it directly in the vehicle’s control commands, with intermediate processes relying on neural network models, the report said.
However, the problems that arise in decision-making are difficult to be pinpointed, becoming a challenge that prevents the end-to-end system from taking hold, the report noted.
Despite the aggressive plans of the leading new car makers, it is still extremely difficult for an end-to-end system to be used for vehicles.
“The cost of switching from common architecture to end-to-end technology is very high and the chain is very long. By the end of this year, maybe 10 percent of the functionality could be switched to an end-to-end system,” LatePost quoted a source in the smart driving space as saying.
Competition is likely to become more intense after end-to-end technology is used on a wide scale.
“The end-to-end technology is currently immature and there is no significant gap between companies,” the report quoted another industry source as saying, adding that once the model matures, car sales, and the amount of data that corresponds to it, will become new barriers between different brands.
On January 30, Nio’s Harry Wong, head of smart driving products and experience, said on Weibo that “end-to-end” will be the hottest smart driving marketing term this year, following “HD map-free” last year.
