Tesla's latest announcement regarding its self-driving taxi (Robotaxi) has attracted worldwide interest, indicating a significant advancement in its Full Self-Driving (FSD) technology. However, hurdles exist for the electric vehicle vendor pushing autonomous driving in not only the US but also China.
The US National Highway Traffic Safety Administration (NHTSA) is actively investigating the specifics of Tesla's fully autonomous driving (FSD) technology. Chinese officials have once again raised national security concerns, which may result in a higher entrance criterion for foreign-owned self-driving cars, according to industry sources.
As AI gains prominence in 2024, Tesla's end-to-end learning approach has drawn renewed global interest. This methodology is also being pursued by other companies, including the United Kingdom's Wayve and Canada's Waabi.
End-to-end autonomous driving: Progress but hurdles
The end-to-end large model is regarded as a watershed point in advanced self-driving technology. The market's confidence was affected in the first half of this year as a result of an event involving Cruise in late 2023. However, the capital markets increasingly see "end-to-end" as a game-changing approach to overcome barriers in autonomous vehicles.
The end-to-end theory was first proposed by Carnegie Mellon University in 1988, suggesting that neural networks could process data from cameras and laser rangefinders to generate driving directions.
This technology has seen rapid adoption in China, where many companies shifted away from high-definition maps in late 2022 in favor of lighter mapping solutions. Major players include automakers NIO, Xpeng, Zeekr, and Li Auto, alongside tier-1 suppliers Huawei, Haomo.AI, and QCraft, each advancing the technology through different approaches.
Industry insiders have observed that the end-to-end model prioritizes "universality." Tesla and the aforementioned Chinese automakers are the primary representatives of this technology, which seeks to operate anywhere by autonomously collecting vast amounts of data and generating driving instructions.
Industry insiders point out that the end-to-end model prioritizes "universality." It aspires to function anywhere by autonomously collecting enormous quantities of data and creating driving instructions, with Tesla and the aforementioned Chinese automakers leading the way.
Tesla is capable of deploying up to 1.8 million vehicles annually, as it is able to deploy updated FSD versions on Model 3, Y, S, and other models without interference from government policies.
Nevertheless, this model is not supported by all advanced self-driving operators. Recently, Aurora, Intel's Mobileye, and Google's Waymo have all pledged to employ the most advanced AI models to develop more sophisticated strategies that integrate safety protocols and verification methods. These conventional advanced self-driving companies prioritize stable routes, particularly in the aftermath of the Cruise incident, underscoring the necessity of supplementary safety assurances and validation procedures for autonomous vehicles.
The end-to-end model, which absorbs vast quantities of data and provides driving commands without intermediate coding protection or insight creation procedures, is perceived by certain industry participants as being too risky. Mobileye acknowledges that this technology is fraught with substantial risk and does not anticipate the model's advancement optimistically. Aurora, which is scheduled to introduce unmanned commercial vehicles by the conclusion of this year, has also rejected the end-to-end model.