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 >>自动驾驶前沿信息获取 → 自动驾驶之心知识星球 今年整个自动驾驶行业融资已经超过300亿！其中L4相关公司的融资比较密集，柱哥周末盘了下部分公司的融资情况，分享给大家。据我 们了解，这其中很多公司也在抓紧落地端到端、VLA等核心技术， 明年将会是L4大爆发的一年。 新石器 新石器今年完成了C+轮和D轮两起融资，2025年2下旬的10亿C+轮融资，和10月23日的D轮超6亿美元融资。D轮融资由阿联酋磊石资本领 投，高成投资、信宸资本、鼎晖VGC等机构联合领投。 以金额计，该融资是迄今为止中国自动驾驶领域最大的一笔私募融资。 公司简介： 成立时间：2018 年 2 月，创始人余恩源 定位：全球领先的 L4 级无人城配 (RoboVan) 解决方案提供商，专注城市物流 "最后一公里" 技术：全栈自研 L4 级无图自动驾驶技术，已交付超 1 万台无人车，累计行驶里程超 5000 万公里 核心产品：RoboVan 系列无人配送车，覆盖快递、即时物流、零售等多场景 里程碑：2025 年累计融资超 62 亿元，产品已拓 ...

以下文章来源于自动驾驶数据挖掘 ，作者黑客与作家 自动驾驶数据挖掘 作者 | 黑客与作家 来源 | 自动驾驶数据挖掘 原文链接： 【E2E训练】首创！同济孙剑团队三层框架解析端到端自动驾驶训练生态！ 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 >>自动驾驶前沿信息获取 → 自动驾驶之心知识星球 本文只做学术分享，如有侵权，联系删文 导读 破解"训练碎片化"痛点！现有端到端自动驾驶（E2E-AD）训练存在数据堆砌、策略孤立、平台割裂三大问题，导致模型泛化弱、部署难。同济大学 +UNC联合发布 Data-Strategy-Platform三层训练生态框架 ，实现三重核心突破： ① 数据层从 "规模扩张"转向"价值密度" ，聚焦 高风险/长尾场景 ； ② 策略层从 "任务级拼接"升级为"能力型基础" ，涵盖IL/RL到生成式范式（扩散/LLM/世界模型）； ③ 平台层从 "静态离线"进化为"持续闭环" ，整合分布式训练、云边协同与灰度发布。 该综述整合280+论文、6大车企工业实践，明确未来三大趋势，为E2E-AD从科研走向量产提供系统级"训练导航图"。 图 ...

Core Viewpoint - The article discusses the breakthrough technology CoherentGS developed by Peking University, which enables high-quality 3D scene reconstruction from a limited number of blurry images, addressing the challenges of sparse views and motion blur [5][33]. Group 1: CoherentGS Technology Overview - CoherentGS utilizes a "dual prior guidance" strategy, allowing the reconstruction of high-definition, coherent 3D scenes from just 3 to 9 blurry images [5][7]. - The system effectively addresses both "deblurring" and "geometric completion" issues through a collaborative optimization process [7][12]. - The core framework integrates deblurring and geometric completion into the entire 3D Gaussian optimization process, ensuring clear and coherent reconstruction results [10][12]. Group 2: Key Technologies of CoherentGS - The deblurring prior restores clear details and provides photometric guidance, essential for extracting reliable details from blurry images [13]. - The diffusion prior completes geometric gaps, ensuring global coherence by filling in unobserved areas with structured images [18]. - Consistency-guided camera exploration intelligently selects valuable viewpoints, enhancing optimization efficiency without blindly increasing perspectives [19][21]. - Joint optimization incorporates geometric regularization to avoid distortion, ensuring the reliability of reconstructed geometry [24][26]. Group 3: Performance Validation - CoherentGS outperforms existing methods, achieving a PSNR improvement of up to 2.78 dB and reducing LPIPS by over 40% on the Deblur-NeRF and DL3DV-BLUR datasets with 3 to 9 sparse blurry inputs [26]. - The qualitative results demonstrate that CoherentGS can recover texture details and maintain structural coherence, unlike other methods that produce either blurriness or fragmented structures [29]. - Frequency analysis shows that CoherentGS retains natural high-frequency details, confirming that the restored details are genuine and not artificially generated [32]. Group 4: Future Implications - CoherentGS represents a significant advancement in 3D reconstruction, breaking the dependency on dense, clear inputs, and has the potential to extend to various real-world scenarios involving defocus blur and exposure anomalies [33].

Core Insights - The article emphasizes the growing importance of embodied intelligence as a key technology trend globally, with numerous companies emerging in this field, both domestically and internationally [1][5][9]. Group 1: Industry Trends - There is a significant talent gap in the embodied intelligence sector, with many master's graduates being pre-booked for positions [1]. - High-profile executives from various industries, including autonomous driving, are transitioning to startups in the embodied intelligence space [1]. - A community platform, "Embodied Intelligence Heart Knowledge Circle," has been established to facilitate knowledge sharing and networking among industry professionals [1][9]. Group 2: Community and Learning Resources - The community offers continuous live sharing sessions, including roundtable discussions and webinars, to keep members informed about developments and challenges in the embodied intelligence industry [3]. - A comprehensive technical roadmap has been created for beginners, outlining essential technologies and learning paths in the field [5]. - Valuable industry frameworks and project proposals are provided for those already engaged in related research [7]. Group 3: Job Opportunities and Networking - The community has established a job referral mechanism with various embodied intelligence companies, facilitating connections between job seekers and employers [9]. - Members have access to exclusive learning materials and can interact with industry leaders for guidance on career and research choices [13][9]. Group 4: Research and Development - The community compiles a wide range of open-source projects, datasets, and simulation platforms relevant to embodied intelligence, aiding both newcomers and experienced professionals [10][25]. - A collection of research reports and academic resources related to embodied intelligence and robotics is available for members to stay updated on industry advancements [18][17].

Core Viewpoint - The article discusses the distinction between world models and end-to-end models in autonomous driving, emphasizing that world models are a means to achieve end-to-end autonomous driving rather than a specific technology [2]. Group 1: World Model Overview - The article highlights the recent surge in publications related to world models, particularly in the context of closed-loop simulation, which is becoming a trend in the industry due to the high costs associated with corner cases [2]. - It introduces a new course focused on world models, covering various algorithms such as general world models, video generation, and OCC generation, with applications in Tesla's world model and the Marble project by Fei-Fei Li's team [2][5]. Group 2: Course Structure - The course consists of six chapters, starting with an introduction to world models and their relationship with end-to-end autonomous driving, followed by a discussion on the historical development and current applications of world models [5][6]. - The second chapter covers foundational knowledge related to world models, including scene representation and technologies like Transformer and BEV perception, which are crucial for understanding subsequent chapters [5][6]. Group 3: Advanced Topics - The third chapter focuses on general world models, discussing notable models such as Marble, Genie 3 from DeepMind, and the latest developments from Meta, including the VLA+ world model algorithm [6][7]. - The fourth chapter delves into video generation-based world models, presenting classic works and recent advancements in the field, including projects like GAIA-1 & GAIA-2 and OpenDWM [7][8]. - The fifth chapter addresses OCC generation methods, explaining their potential for trajectory planning and end-to-end implementation [8]. Group 4: Industry Application and Career Preparation - The sixth chapter provides insights into the practical applications of world models in the industry, discussing pain points and how to prepare for job interviews in this field [9]. - The course aims to equip participants with the skills to understand and implement world model technologies, preparing them for roles as world model algorithm engineers [10][13].

Core Viewpoint - The article discusses the rapid growth and market dynamics of urban NOA (Navigation on Autopilot) suppliers, highlighting the strategic investments and partnerships that are shaping the industry landscape [4][5]. Group 1: Investment and Market Position - Yuanrong has secured strategic investments from leading Tier 1 suppliers and luxury car manufacturers, indicating strong industry interest in high-quality urban NOA suppliers [4]. - Major players like Huawei, Yuanrong, and Momenta each hold over one million urban NOA project orders, suggesting a competitive market structure [5]. Group 2: Growth and Market Trends - Yuanrong has delivered 200,000 vehicles equipped with urban NOA, achieving a nearly 40% market share in the third-party supplier market by October 2025 [4]. - The urban NOA market is expected to experience significant growth, surpassing highway NOA as the mainstream solution due to the increasing adoption and technological advancements [4][6]. Group 3: Future Projections and Challenges - By 2026, urban NOA is projected to see a major surge in volume, driven by reduced hardware costs and the integration of intelligent driving in traditional fuel vehicles, potentially adding millions of units to the market [6]. - Achieving a production scale of over one million units will be a critical milestone for leading intelligent driving companies, as it will help establish data barriers and competitive advantages [6][7]. Group 4: Technological Evolution - The article emphasizes the importance of technological iteration, particularly the transition from VLA (Vehicle Level Automation) from initial production to significant performance improvements in 2026 [7]. - Companies must balance the need for cost-effective urban NOA solutions with advancements in cutting-edge technologies to remain competitive in the evolving market [8].

Group 1 - The core issue for autonomous driving company A is internal management problems leading to its operational halt, rather than just technical shortcomings [4][5] - Company A's decline was evident as early as April last year, triggered by a whistleblower report regarding a high-salaried employee's resume fraud, which uncovered significant financial issues [4] - Following the loss of trust from the parent group B, all operational permissions of company A were revoked, leading to its eventual takeover by group B [4] Group 2 - New energy vehicle company C's supply chain head was dismissed due to failure to stockpile memory chips during a price surge, which angered the CEO [6] - This incident was not isolated, as company C had previously faced similar issues with core component shortages, indicating a pattern of mismanagement [6] Group 3 - The departure of the autonomous driving head from new energy vehicle company D was triggered by plans to eliminate the existing mapping team, which led to internal conflicts and ultimately his resignation [7] - Despite D's significant investment in high-end technology and a large team, the challenges of developing a mapped route have hindered progress, leaving the company under pressure to deliver results [7] Group 4 - Logistics company E is investing 150 million in developing an L4 autonomous driving demo, but its internal team structure is causing inefficiencies due to merging teams with fundamentally different architectures [8] - The success of this demo is critical for attracting investment, but failure could lead to significant layoffs within the company [8] Group 5 - Autonomous vehicle company F's plan to split and seek independent financing failed due to difficulties in securing investment, despite its low valuation of approximately 1 billion RMB [9] - The company previously operated a fleet of nearly 1,000 autonomous vehicles but has since faced significant team instability and leadership changes, leading to a decline in operational effectiveness [9] Group 6 - A well-known automotive manufacturer G has adopted a "performance theater" culture, leading to ineffective innovation practices and minimal output from its large engineering team [10] - The company has only managed to successfully run two demo routes, despite having a substantial number of engineers, indicating a disconnect between innovation goals and actual productivity [10] Group 7 - Company H's management style is characterized by a lack of accountability and engagement among executives, leading to a disorganized and ineffective workforce [12] - The company has seen a decline in morale and productivity, with many core talents feeling undervalued and overworked [12] Group 8 - In a new energy vehicle company I, internal conflicts led by executive A have resulted in significant inefficiencies and a failure to adapt to industry trends, particularly in autonomous driving technology [13][14] - The CEO's decision to ignore advice to follow industry leaders in end-to-end technology has caused the company to fall behind competitors [14] Group 9 - Autonomous trucking company J is facing financial losses due to its L2 driving assistance model, which has not effectively reduced operational costs and has led to inventory issues [15] - The company's strategy of incentivizing usage through subsidies initially worked but has since resulted in customer dissatisfaction and vehicle returns due to supply chain issues [15]

Core Insights - The article discusses various advanced directions in autonomous driving research, emphasizing the importance of deep learning and traditional methods for different academic backgrounds [2][3]. Group 1: Research Directions - Key areas of focus include VLA, end-to-end learning, reinforcement learning, 3DGS, and world models, which are recommended for students in computer science and automation [2]. - For mechanical and vehicle engineering students, traditional methods like PnC and 3DGS are suggested due to their lower computational requirements and ease of entry [2]. Group 2: Paper Guidance Services - The article announces the launch of a paper guidance service that covers various topics such as end-to-end learning, multi-sensor fusion, and trajectory prediction [3][6]. - The service includes support for topic selection, full process guidance, and experimental assistance [6]. Group 3: Publication Success - The guidance service has a high acceptance rate for papers submitted to top conferences and journals, including CVPR, AAAI, and ICLR [7]. - The article highlights the range of publication venues, including CCF-A, CCF-B, and various SCI categories [10].

Core Insights - The article discusses the introduction of SpaceDrive, a new framework for autonomous driving that enhances spatial awareness in Vision-Language Models (VLMs) by integrating 3D positional encoding, addressing existing limitations in spatial reasoning and trajectory planning [3][4][31]. Group 1: Framework Overview - SpaceDrive replaces traditional VLM methods that treat coordinate values as text tokens with a unified 3D positional encoding, improving the system's spatial reasoning and trajectory planning capabilities [4][5]. - The framework demonstrates state-of-the-art (SOTA) performance in open-loop evaluations on the nuScenes dataset and ranks second in closed-loop evaluations on the Bench2Drive benchmark, achieving a driving score of 78.02 [3][21]. Group 2: Methodology - SpaceDrive employs a unified spatial interface that integrates visual tokens with 3D positional encoding, allowing for explicit spatial representation and improved accuracy in trajectory planning [5][6]. - The framework utilizes a regression decoder instead of a classification head for predicting trajectory coordinates, addressing the inherent limitations of language models in numerical processing [4][13]. Group 3: Experimental Results - In open-loop planning, SpaceDrive+ outperformed existing VLM-based methods, achieving an average L2 error of 0.32m and a collision rate of 0.23% [17][18]. - In closed-loop planning, SpaceDrive+ achieved a driving score of 78.02 and a success rate of 55.11%, ranking second among VLM-based methods [20][21]. Group 4: Contributions to the Field - SpaceDrive represents a paradigm shift from "language modeling geometry" to "explicit geometric encoding," effectively linking visual spatial perception with physical planning [31][33]. - The framework's introduction of a unified 3D positional encoding across perception, reasoning, and planning modules signifies a major architectural innovation, enhancing the generalizability of spatial intelligence [33].

Core Insights - GAIA-3 represents a significant advancement in the evaluation of autonomous driving systems, transitioning world modeling from a visual synthesis tool to a foundational element for safety assessment [4][20] - The model combines the realism of real-world data with the controllability of simulations, enabling the generation of structured and purposeful driving scenarios for safety validation [6][20] Group 1: GAIA-3 Features - GAIA-3 is a powerful testing tool that can modify vehicle trajectories, weather conditions, and adapt to different sensor configurations [3] - It is built on a latent diffusion model with 15 billion parameters, doubling the video tokenizer size compared to its predecessor GAIA-2 [3][19] - The model allows for the generation of controlled variants of real-world driving sequences, maintaining consistency in the environment while altering vehicle behavior [6][8] Group 2: Safety and Evaluation - GAIA-3 addresses the limitations of traditional testing methods by generating systematic variations of critical safety scenarios, such as collisions, using real-world data metrics [7][8] - The model enables offline evaluation of autonomous systems by recreating unexpected events, allowing for quantitative testing of recovery capabilities in edge cases [9][20] - It emphasizes consistency in generated scenarios, ensuring that changes in vehicle behavior do not disrupt the physical and visual coherence of the environment [8][11] Group 3: Data Enrichment and Robustness - GAIA-3 enhances data coverage by generating structured variants from rare failure modes, facilitating targeted testing and retraining [12][13] - The model supports controlled visual diversity, allowing for measurable changes in appearance while keeping the underlying structure consistent, thus improving robustness assessments [11] - It can transfer scenarios across different sensor configurations, enabling data reuse across various vehicle projects without the need for paired collection [10] Group 4: Technical Advancements - The advancements in GAIA-3 are driven by increased scale, with training compute five times that of GAIA-2 and a dataset covering eight countries across three continents [16][19] - The model captures critical spatial and temporal structures, enhancing the fidelity of generated scenarios and improving the understanding of causal relationships in driving behavior [19][18] - GAIA-3's capabilities provide a reliable framework for structured, repeatable testing, marking a significant step towards scalable evaluation of end-to-end driving systems [20]