• Baidu recently dispatched engineers to 's Beijing office to work on better integrating its navigation map information with the FSD, according to Reuters.
  • The move aims to improve the FSD V13's understanding of Chinese roads with more accurate and updated map information.
(A Tesla Model Y on display at the June 2024 Shanghai new energy vehicle show. Image credit: CnEVPost)

Tesla (NASDAQ: TSLA) is working with Baidu to improve the performance of its ADAS (Advanced Driver Assistance System) in China, Reuters said in a report today, citing two people familiar with the matter.

Baidu has sent a group of engineers from its mapping team to Tesla's Beijing office in recent weeks to work on better integrating Baidu's navigation map information, such as lane markings and traffic signals, with Tesla's FSD (Full Self-Driving) Version 13 software, the sources said, according to the report.

The sources did not disclose how many engineers were sent to Tesla or if they are still there, the report said.

Join us on or

They said the move is aimed at improving FSD V13's knowledge of Chinese roads with more accurate and updated map information, according to Reuters.

Tesla began pushing out software update version 2024.45.32.12 to eligible vehicles in China on February 25, bringing high-level smart driving features similar to FSD, though avoiding any mention of FSD.

The update adds city navigation to the system, but customers have complained that it fails to fulfill Tesla CEO Elon Musk's promise of a full FSD rollout in China, the Reuters report noted.

The FSD V13 hasn't been trained enough to adapt well to Chinese roads, causing drivers to frequently break traffic rules, such as switching to the wrong lane and running red lights if they aren't paying attention and reacting, Reuters said, citing a source.

In the US, FSD doesn't need navigation maps to be accurate or up-to-date, as local training of the AI helps the technology drive better, the report noted.

But in China, where Tesla can't use data from its 2 million EVs to train the system because of China's data laws, the report said.

After receiving the latest software update, a large number of local auto bloggers shared their experiences with the feature on Chinese social media platforms, with mixed reviews.

Training the algorithm with local data is necessary if it is to provide an improved autonomous driving experience.

Tesla faces a dilemma in this regard -- China doesn't allow data to leave the country, and the US government's chip ban has prevented Tesla from building a large-scale computing center in China.

Local media outlet LatePost said in a February 25 report that Tesla sent some engineers from its US headquarters to China in early February to locally deploy a mature version of FSD and optimize the algorithm.

Theoretically, the move could improve FSD's performance in China as much as possible without moving data outside the country, according to LatePost.

On the same day, Musk responded to a post about FSD's performance in China on X, the social media platform he owns, saying that Tesla used publicly available videos of Chinese roads and signs on the internet to train FSD.

Tesla starts rolling out FSD-like features in China

Follow by email

Choose your topics and subscribe.

Daily Recap:

You'll receive a daily email containing a summary of everything that's in CnEVPost that day.

Topics:

You'll receive an email alert when any new article is published on your chosen topic.

    Get our daily recap:
    Or, choose 1 or more topics: