Xpeng Motors Releases China’s First Mass-Produced End-to-End LM On-Board

One year ago, the L4 level autonomous driving track entered a cold winter; today, with rumors of Tesla’s FSD entering China, and XPeng Motors taking the lead in entering the AI smart driving era, advanced autonomous driving is showing signs of heating up again.

On May 20th, XPeng Motors released the first domestically produced end-to-end large model: neural network XNet + rule control large model XPlanner + large language model XBrain. It is reported that it will double the intelligent driving capability, double the perception distance, and recognize over 50 target objects; it can identify waiting areas for turns, tidal lanes, read road signs text, infer traffic participant intentions; in complex scenarios such as jerks, stalls, safety takeovers are all reduced by nearly 50%. According to the plan, in the third quarter of 2024, XPeng Motors will achieve nationwide driving on every road with full realization without maps. By 2025 in China achieving a class L4 level intelligent driving experience.

The transformation of intelligent vehicles by large-scale models is disruptive, not proportional changes, not tens of percent, but tens of times the change.” XPeng Motors Chairman He Xiaopeng bluntly stated that he used to think that fully autonomous driving in China would take many years to achieve. “After getting on board with the end-to-end large model, I saw the data and really felt there was a chance. Fully autonomous driving will come within a few years.”

He Xiaopeng’s optimistic expectations also represent a new perspective on the future prospects of Level 4 autonomous driving in the industry.

Recently, discussions about Tesla’s entry into China with FSD and the launch of Robotaxi in August this year have sparked continued market attention. The industry is looking forward to another “catfish effect” triggered by Tesla. Previously, Tesla CEO Musk posted on social media that Tesla plans to launch Robotaxi on August 8, 2024. Subsequently, there were market reports that Robotaxi surpassed the new entry-level model Model 2 and became a priority project for development within Tesla. Against this backdrop, Tesla has started recruiting technical talents such as automatic assisted driving software test engineers for its Shanghai data center.

For the Robotaxi that Tesla will launch, Baidu, Inc. will consider possible cooperation opportunities based on specific application models and the pace of entering the Chinese market from Tesla.” Xu Baoqiang, general manager of Baidu‘s autonomous driving technology department, said.

In addition to Tesla and Baidu, in recent times many car companies and autonomous driving enterprises have publicly expressed optimism towards commercial projects for advanced autonomous driving. Renault is collaborating with WeRide to plan a large-scale commercial deployment of L4-level autonomous driving models; GAC AION jointly established Guangzhou Andi Technology Co., Ltd. with DiDi for registration completion. The company plans to launch its first commercial L4-level autonomous driving model by 2025; FAW announced multi-level long-term cooperation with Huawei on an interactive platform. Autonomous driving products will be applied in L4-level low-speed scenarios, expected to be mass-produced and applied by the end of 2025.

SEE ALSO: DiDi Autonomous Driving to Create Joint Venture with GAC AION to Mass Produce Electric Robotaxi

The most core feature of Level 4 autonomous driving is that the vehicle can independently complete all driving operations on specific roads and under certain environmental conditions without human driver intervention. The industry generally believes that this level of autonomous driving technology is a key step towards full automation. Despite remarkable technological breakthroughs and progress, there are still many uncertainties in commercial prospects, such as laws and regulations, infrastructure, and policy support.” Analysis by industry insiders said.

The organization predicts that Robotaxi will achieve large-scale commercialization by 2026; by 2030, the cost of domestic Robotaxi will drop to around 1 yuan per kilometer, lower than the cost of human-driven ride-hailing services. The domestic Robotaxi market size is expected to reach nearly 500 billion yuan.