China's Self-Driving Landscape in the Year 2024
In the bustling landscape of China's automotive industry, a new race is unfolding - the race for autonomous driving. With giants like Huawei, Nio, Xpeng Motors, Li Auto, and SAIC throwing their hats into the ring, the competition is fierce.
Huawei, a subsidiary of the telecommunications powerhouse, announced the full-scale deployment of advanced autonomous driving features to users, covering 99% of China's roads. Following suit, Nio became the second Chinese automaker to roll out these advanced features. However, both companies require users to pass assessments and tests before enabling autonomous driving in urban areas.
The autonomous driving arena in China saw a surge in activity in 2023, with these companies competing for market dominance. Huawei's ADS 3.0 introduced a large model that prioritizes perception, and Xpeng introduced a neural network-based large model called XPlanner for planning and control.
Achieving nationwide capabilities does not guarantee usability and functionality. Industry experts suggest that end-to-end solutions can exhibit more human-like behaviors, especially in navigating complex urban intersections. However, challenges persist in machine perception, particularly at large intersections.
Nvidia's upcoming next-generation autonomous driving chip platform, Thor, is anticipated to boost AI computing power and potentially transform vehicles into more dynamic entities. Companies are gearing up to prioritize the development of planning and control modules, with the eventual aim of achieving end-to-end autonomy.
Domestically developed solutions that are supposedly end-to-end still need to cycle through the perception, prediction, and planning stages. End-to-end solutions in autonomous driving treat sensor data as input and output vehicle control commands, processing them directly. Domestic end-to-end solutions are unlikely to adopt Tesla's radical approach and will instead opt for a more moderate strategy.
Tesla's Full Self-Driving (FSD) package has entered the Chinese market, adding a new dimension to the competition. Tesla's FSD V12 is powered by an end-to-end neural network, representing the next-generation solution for autonomous driving. However, the technical implementation of Tesla's end-to-end solution remains undisclosed, posing a challenge for the industry to assess its capabilities.
Chinese companies with the largest computing power for autonomous driving training so far are primarily BYD and SAIC. BYD uses Nvidia Orin-X chips (300 TOPS) and runs a multimodal AI architecture called Xuanji AI, supported by a large automotive cloud database developed by around 5000 engineers. SAIC, in collaboration with its AI Lab and tech company Momenta, runs robotaxis on SAE Level 4 with computing power delivering 600 trillion operations per second. They also share data from millions of test kilometers between their robotaxis and passenger vehicles.
Users need to accumulate a certain amount of autonomous driving mileage on highways before they are allowed to utilize it in city centers. This cautious approach is reflective of the actual deployments of autonomous driving features by automakers, which tend to be relatively cautious.
Li Auto's autonomous driving team has been working intensely for the past half-year. Li Xiang, CEO of Li Auto, has been conducting midnight tests of autonomous features and leading bug-fixing sessions. Working seven days a week, often into the early hours, is the new normal during Li Auto's closed development phase. Li Auto is transitioning toward becoming an artificial intelligence technology company.
Xpeng has consistently positioned autonomous driving as a strategic advantage. Skepticism is rife among industry professionals about the definitions of nationwide capabilities and the intensity of testing. Ren Shaoqing, head of autonomous driving at Nio, expects it to take at least two years to achieve nationwide autonomous driving capabilities.
Despite the challenges, the race for autonomous driving in China is heating up. With each company pushing the boundaries of technology and testing, it's uncertain who will emerge as the leader in this exciting and rapidly evolving field.
Read also:
- IM Motors reveals extended-range powertrain akin to installing an internal combustion engine in a Tesla Model Y
- Ford Embraces Silicon Valley Approach, Introducing Affordable Mid-Sized Truck and Shared Platform
- Future Outlook for Tesla in 2024: Modest Expansion in Electric Vehicle Sales, Anticipated Surge in Self-Driving Stock
- Vegetable oils are similarly utilized in the process of road cleaning.
