点击下方 卡片 ，关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近30个 方向 学习 路线 智驾的技术焦虑，正在中游厂商快速传播。 周末有机会和一位深耕主机厂L2量产交付的负责人线下交流，其认为 明年才是端到端等前沿技术大规模量产的起点。 智驾前沿的技术发展放缓，业内量产方案趋同，L2整体在走下沉路线。二十万以上的乘用车销量在700万左右，但头部新势力的销量不及1/3，更不用说端到端量产 占比的车型。从落地趋势上来看，端到端技术的成熟反而才是更大规模量产的开端。随着明年L3法规的进一步推进， 中游厂商的技术升级也是迫在眉睫。 所以这 两个月很多公司算法负责人联系自动驾驶之心，迫切的想要了解前沿的技术：端到端、世界模型、VLA、3DGS等等。 端到端不仅仅是一个算法，需要完善的云端&车端基建，数据闭环、工程部署、闭环测试、模型优化、平台开发等等，可以预见，中阶智能驾驶的岗位需求会更旺 盛。而在昨天的2025地平线技术生态大会上，地平线CEO也表示将挺进10万级市场，高阶智驾正在迅速下沉至更多的国民车型。明年，智能驾驶的故事将更精彩。 以上。 基本上可以判断端到端、VLA的招聘需求会更旺盛。最近几个月， ...

Core Viewpoint - He Xiaopeng, CEO of XPeng Motors, has placed pressure on the autonomous driving team by betting that the second-generation VLA will match Tesla's FSD V14.2 capabilities by August 30, 2026, or face a humorous consequence for the team leader [2][3][21]. Group 1: Autonomous Driving Development - XPeng Motors plans to officially release the second-generation VLA in the first quarter of 2026, with a full rollout to Ultra models [5][23]. - He Xiaopeng expressed confidence that the second-generation VLA could potentially reach L4 capabilities, and possibly L5 with additional time [6][24]. - The VLA concept integrates visual, language, and action capabilities, aiming for a seamless execution of tasks without breaking them into steps [8][26]. Group 2: Technological Advancements - The second-generation VLA has been trained using nearly 100 million video clips, equating to the driving experience of a human driver over 65,000 years [8][26]. - XPeng's self-developed Turing AI chip has a computing power of 750 TOPS per chip, totaling 2250 TOPS for the vehicle, significantly surpassing the industry standard [11][29]. - The model utilizes a cloud computing cluster with 30,000 cards and plans to expand to 50,000 cards, ensuring ample computational resources for development [8][26]. Group 3: Competitive Landscape - Tesla's FSD has a significant advantage with over 600,000 test vehicles generating 1.6 billion frames of image data daily, accumulating over 9.6 billion kilometers of driving distance [14][32]. - In practical tests, Tesla's FSD V13.2.9 required 5 interventions over a 20-kilometer complex route, while XPeng's second-generation VLA only needed 1 [16][33]. - The latest FSD V14.2 has improved system performance and addressed over 95% of hesitation and abnormal braking issues from the previous version, enhancing the driving experience [17][34].

Core Viewpoint - He Xiaopeng, CEO of XPeng Motors, has placed pressure on the autonomous driving team by betting that the second-generation VLA will match Tesla's FSD V14.2 capabilities by August 30, 2026, or face a humorous consequence for the team leader [2][4]. Group 1: Autonomous Driving Development - XPeng Motors plans to officially launch the second-generation VLA in the first quarter of 2026, with a full rollout to Ultra models [4]. - He Xiaopeng expressed confidence that the second-generation VLA could potentially reach L4 capabilities, and possibly L5 with additional time [6]. - The VLA concept integrates visual, language, and action capabilities, aiming for a seamless execution of tasks without breaking them into multiple steps [8]. Group 2: Technological Advancements - The second-generation VLA will eliminate the language translation step, allowing for direct generation of action commands from visual signals [9]. - The model has been trained using nearly 100 million video clips, equating to the driving experience of a human driver over 65,000 years [9]. - XPeng's self-developed Turing AI chip boasts a computing power of 750 TOPS per chip, with a total of 2250 TOPS across three chips, significantly surpassing the industry standard [12]. Group 3: Competitive Landscape - Tesla's FSD has a significant advantage due to its extensive global testing network, with over 600,000 vehicles generating 1.6 billion frames of data daily and accumulating over 9.6 billion kilometers of driving experience [13]. - The latest FSD V14.2 version has improved system performance and resolved over 95% of hesitation and abnormal braking issues from previous versions [13]. - XPeng's second-generation VLA is still in development, and its actual performance remains to be seen [14].

Core Viewpoint - The article discusses the technological anxiety in intelligent driving, particularly among mid-tier manufacturers, and highlights the anticipated growth in demand for end-to-end (E2E) and VLA (Vision-Language-Action) technologies in the coming year [2]. Group 1: Industry Trends - The mass production of cutting-edge technologies like end-to-end systems is expected to begin next year, with L2 technology becoming more standardized and moving towards lower-tier markets [2]. - The total sales of passenger vehicles priced above 200,000 are around 7 million, but leading new forces account for less than one-third of this, indicating a slow adoption of end-to-end mass production models [2]. - The maturity of end-to-end technology is seen as a precursor to larger-scale production, with the advancement of L3 regulations prompting urgent upgrades among mid-tier manufacturers [2]. Group 2: Recruitment and Training - There is a growing demand for positions related to end-to-end and VLA technologies, as many professionals are seeking to quickly learn these advanced skills [3]. - The article mentions the launch of specialized courses aimed at practical applications of end-to-end and VLA technologies, designed for individuals already working in the field [3][6]. - The courses will cover various modules, including navigation information application, reinforcement learning optimization, and production experiences related to diffusion and autoregressive models [3][6]. Group 3: Course Details - The end-to-end production course will focus on practical implementation, including seven major practical applications, making it suitable for those looking to advance their careers [3][6]. - The VLA course will cover foundational algorithms and theories, including BEV perception and large language models, with practical projects based on diffusion models and VLA algorithms [6][11]. - The instructors for these courses are experienced professionals from top-tier companies and academic institutions, ensuring a high-quality learning experience [5][8][13].

Core Viewpoint - The article discusses the challenges and advancements in the VLA (Variable Learning Algorithm) field, emphasizing the importance of real machine data and practical experience in achieving effective results in embodied intelligence applications. Group 1: Data Collection - Data collection methods for VLA primarily include imitation learning and reinforcement learning, with remote operation, VR, and full-body motion capture being key techniques [8][9] - The quality of data collected is crucial, and methods like real2sim2real are highlighted as important for effective data acquisition [8] Group 2: VLA Training - Before deploying models in real machines, simulation debugging is essential, especially when real machine data is insufficient [10] - Training techniques are critical, with challenges in fine-tuning models and achieving good results with small data sets [10] - Some algorithms, like ACT, are easier to train, while others, such as π0 and π0.5, require more intricate techniques and experience [10] Group 3: VLA Deployment - After training, models often need to be "slimmed down" due to their large parameter sizes, which poses challenges for deployment on edge chips [12] - Techniques like quantization and distillation are necessary to minimize parameter size while maintaining performance [12] Group 4: Educational Initiatives - The article introduces a practical course aimed at helping individuals learn VLA effectively, covering various aspects such as hardware, data collection, algorithms, and deployment [13][16] - The course is designed for a wide audience, including students and professionals looking to transition into the embodied intelligence field [27]

Core Viewpoint - The article discusses the challenges and advancements in the field of VLA (Vision-Language Alignment) models, emphasizing the importance of real machine data and practical applications in robotics and embodied intelligence. Group 1: Challenges in VLA Implementation - Many students struggle with the transition from theoretical knowledge to practical application, often finding it difficult to achieve satisfactory results without hands-on experience [2][6] - The reliance on real machine data for effective training and deployment of VLA models is highlighted, with a focus on the limitations of simulation data [2][8] Group 2: Data Collection and Training - Data collection methods for VLA include imitation learning and reinforcement learning, with a particular emphasis on remote operation and VR techniques [8] - The training of VLA models requires careful tuning and optimization, with specific challenges noted for models like π0 and π0.5, which demand a high level of expertise [10][12] Group 3: Deployment and Optimization - Post-training, VLA models often require optimization techniques such as quantization and distillation to reduce parameter size while maintaining performance [12] - The deployment of VLA models on edge devices presents significant challenges due to their typically large parameter sizes [12] Group 4: Educational Initiatives - The article introduces a practical course aimed at helping individuals learn about VLA, covering various aspects such as hardware, data collection, algorithm implementation, and real-world applications [14][30] - The course is designed for a diverse audience, including students and professionals looking to transition into the field of embodied intelligence [27][30]

Core Viewpoint - The article discusses the challenges and advancements in the VLA (Variable Learning Algorithm) field, emphasizing the importance of real machine data collection and the complexities involved in training and deploying VLA models. Group 1: Data Collection - Real machine data collection is crucial for VLA models, with methods including remote operation, VR, and full-body motion capture [2][8] - The effectiveness of data collection methods and ensuring high-quality data are significant challenges, particularly in the context of real-to-sim-to-real transitions [8] Group 2: VLA Training - Training VLA models typically requires simulation debugging before real machine deployment, especially when real machine data is insufficient [10] - Techniques for fine-tuning models and achieving good results with small data sets are critical, as many students struggle with training models effectively [10] Group 3: VLA Model Deployment - After training, VLA models often require "slimming" due to their large parameter sizes, which poses challenges for deployment on edge chips [12] - Lightweight operations such as quantization and distillation are essential to minimize parameter size while maintaining performance [12] Group 4: Educational Initiatives - The article introduces a practical course aimed at helping students effectively learn about VLA, covering hardware, data collection, algorithms, and deployment [14][16] - The course is designed for various audiences, including those seeking jobs in the field, beginners looking to advance, and researchers in embodied intelligence [27]

Core Viewpoint - The article discusses the challenges and advancements in the VLA (Variable Learning Algorithm) field, emphasizing the importance of real machine data collection and the complexities involved in model training and deployment. Group 1: Data Collection - Real machine data collection is crucial for VLA models, with methods including remote operation, VR, and full-body motion capture being highlighted as effective approaches [2][8]. - The article stresses the need for high-quality data and the significance of the real2sim2real process in ensuring effective data collection [8]. Group 2: Model Training - Training VLA models typically requires simulation debugging before real machine deployment, especially when real machine data is insufficient [10]. - The article notes that many beginners struggle with model training, particularly with advanced models like π0 and π0.5, which require specific techniques and experience to achieve good results [6][10]. Group 3: Model Deployment - After training, VLA models often need to undergo a "slimming" process due to their large parameter sizes, which poses challenges for deployment on edge devices [12]. - Techniques such as quantization and distillation are essential to minimize parameter size while maintaining performance [12]. Group 4: Educational Initiatives - The article introduces a practical course aimed at helping individuals learn about VLA, covering various aspects such as hardware, data collection, algorithms, and deployment [14][16]. - The course is designed for a wide audience, including students and professionals looking to transition into the VLA field, and includes hands-on experience with hardware [27][30].

Investment Rating - The report maintains a "Buy" rating for the company, specifically an "Increase" rating, indicating a projected investment return exceeding the market benchmark by 5% to 15% over the next 6-12 months [4]. Core Views - The company's performance in Q3 2025 was under pressure, with total revenue declining by 36.2% year-on-year and 9.5% quarter-on-quarter to 27.36 billion yuan. The gross margin also decreased by 5.2 percentage points year-on-year to 16.3%. The Non-GAAP net loss attributable to shareholders was 360 million yuan, marking the first quarterly Non-GAAP loss in 2023 [1][2]. - The automotive business revenue fell by 37.4% year-on-year, with sales volume down by 39.0% year-on-year to 93,000 units. The average selling price (ASP) increased by 2.6% year-on-year to 278,000 yuan. The gross margin for the automotive business was 15.5% [2]. - Management indicated that the i6 battery supply will adopt a dual-supplier model starting in November, with production capacity expected to reach 20,000 units by early 2026. The company is also focusing on improving product capabilities and operational efficiency through internal adjustments [3]. Summary by Sections Q3 2025 Performance - Total revenue for Q3 2025 was 27.36 billion yuan, down 36.2% year-on-year and 9.5% quarter-on-quarter. Gross margin decreased to 16.3%, with a Non-GAAP net loss of 360 million yuan [1]. Automotive Business - Revenue from the automotive segment was 25.87 billion yuan, a decline of 37.4% year-on-year. Sales volume dropped to 93,000 units, with an ASP of 278,000 yuan. The gross margin for this segment was 15.5% [2]. Future Outlook - The company expects continued pressure on fundamentals in Q4 2025 and Q1 2026 due to policy fluctuations and intensified competition. However, management's shift back to a startup management model and advancements in self-developed technologies are anticipated to enhance product capabilities and operational efficiency [3][4].

Core Insights - The article discusses the challenges in recruiting talent in the autonomous driving sector, highlighting a shortage of experienced professionals in advanced roles [2] - It emphasizes the importance of education and training in cutting-edge technologies related to end-to-end and VLA (Vision-Language-Action) autonomous driving [2] Course Offerings - A course titled "End-to-End and VLA Autonomous Driving" is being offered, focusing on the latest technologies in the field, including BEV perception, VLM, diffusion models, and reinforcement learning [2][12] - The course is designed for individuals with a foundational knowledge of autonomous driving and related technologies, and it includes practical assignments to build VLA models and datasets [12][16] Instructor Profiles - The course features a team of instructors with strong academic backgrounds and practical experience in autonomous driving and large models, including researchers from top universities [8][11][14] - Instructors have published numerous papers in prestigious conferences and have experience in developing and implementing advanced algorithms in the industry [8][11][14] Target Audience - The course is aimed at individuals who have a basic understanding of autonomous driving modules and are familiar with concepts such as transformer models, reinforcement learning, and BEV perception [16] - Participants are required to have access to a GPU with recommended specifications of 4090 or higher [15][16]