机器之心原创 作者：张倩 当机器人成为各大科技展会最受瞩目的焦点，当具身智能论坛场场爆满、一票难求，我们不难发现：这个领域正在经历前所未有的关注热潮。 然而，热潮之下，仍有诸多关键议题悬而未决：面对 数据 稀缺，有人寄希望于合成数据的突破，有人坚持真机数据才是根本；在 技术路线 之争 中，有人押注端 到端的整体范式，有人则认为分层架构更符合演进规律；至于 模型 形态，有人视 VLA 为智能的最终归宿，也有人认为世界模型才是真正的未来。 现阶段出现这种分歧非常正常，因为整个行业的发展路径尚未收敛。有些问题甚至还没有来得及系统讨论，比如量产之后会出现哪些新的卡点，谁来解决？ 正是因为存在这些问题，业界迫切需要一个开放的对话平台。在 今年 云 栖大会的 具身智能论坛 上，我们见证了这样一场深度交锋：不同派系的代表坐到同一张 桌子前，将技术分歧、商业思考和基础设施需求一并摊开讨论，试图在碰撞中寻找新的共识。 论坛过后，我们也和这场论坛的发起者 —— 阿里云 聊了聊。这家云计算巨头选择在此时深度介入具身智能领域，本身就值得关注。 聊完之后，我们发现，他们真正的入局其实是在四五年前，如今更是在提前为具身智能行业即将到来的 ...

点击下方 卡片 ，关注" 具身智能 之心 "公众号 编辑丨具身智能之心，作者：张峻川 本文只做学术分享，如有侵权，联系删文 >> 点击进入→ 具身智能之心 技术交流群 更多干货，欢迎加入国内首个具身智能全栈学习社区 ： 具身智能之心知识星球 (戳我) ， 这里包含所有你想要的。 一、写在前面&背景 神经渲染（NERF/3DGS）引发了三维重建技术的革命，目前已经在辅助驾驶/具身智能领域得到大量应用。NERF和3DGS使用神经网络表达空间，其在新视角合成 方面的优越表现直击辅助驾驶/具身智能仿真的一大痛点：传感器仿真。如果这一类深度学习技术能够推广应用，就能够很大程度上解决传统计算机图形学渲染出的 图像缺乏真实性的问题，可以广泛应用在算法的闭环测试和训练中。 目前已经有一些研究项目在围绕NERF和3DGS技术打造全新的面向闭环测试的仿真框架。然而完全新开发一个仿真框架，使其具有现行场景仿真软件类似的功能将 会有巨大的工作量。因此另一个应用神经渲染新技术的思路是：将NERF和3DGS训练出的模型嵌入到现有仿真软件的框架中去，在保证实时渲染的前提下，同时能 够应用仿真软件已有的3D数字资产和算法接口等工具链。 在N ...

Core Insights - Nvidia has officially launched the Thor chip, referred to as the "new brain" for robots, priced at $3,499, aimed at enabling real-time intelligent interaction between embodied intelligent robots and the physical world [1] - The Thor chip significantly enhances computational power, offering up to 2070 TFLOPS, a 7.5 times increase over the previous Orin chip, addressing the computational limitations faced by robots [1][3] - The chip's performance improvements allow robots to process large amounts of sensor data and operate AI models at the edge, reducing reliance on cloud computing [3] Group 1: Product Launch and Features - The Thor chip is designed to support embodied intelligent robots with real-time processing capabilities, essential for autonomous operation in various environments [1] - It features a CPU performance increase of 3.1 times, 128GB of memory (a 2 times increase), and a 3.5 times improvement in energy efficiency [1][3] Group 2: Industry Adoption and Ecosystem - Notable companies such as Boston Dynamics and Figure AI, along with domestic firms like UBTECH and Galaxy Universal, have already begun deploying the Thor chip [3] - Nvidia has built a robust developer ecosystem in the robotics field, with over 2 million developers engaged across various industries since 2014 [4] Group 3: Financial Performance - Despite the advancements in robotics, the segment currently contributes a minimal portion to Nvidia's overall revenue, accounting for approximately 1.29% with a total income of $567 million, although it has seen a significant year-on-year growth of 72% [5]

他特别强调中国在机器人领域的三大优势：人才储备、制造能力和丰富应用场景。目前辉达正与宇树科 技、北京银河通用机器人等中国企业合作。虽然大模型并非中国机器人传统强项，但在具身智能研发仍 处于起步阶段。 获利回吐风险仍然存在：继本次大会后，8 月14 到17 日将举行世界人形机器人嘉年华，8 月21 日智元 机器人将召开首届合作伙伴大会并预告发布「神秘新品」。接下来10 月Optimus V3 发布可能成为新催 化剂。在中间的空窗期，机器人板块有获利回吐风险，但我们仍然看好长期技术进步、场景拓展和政策 支持带来的投资机会。首选：双环传动（002472 CH，优于大市）（T 链、分拆上市、变速箱业务）、 恒立液压（601100 CH，优于大市）（行业复苏、精密部件业务）、优必选（9880 HK，优于大市） （机器人链主）。 主要风险：获利回吐风险、中期业绩不及预期、经济增速放缓。 超500 个机器人应用场景：本届大会最大亮点是今年是人形机器人的「量产元年」，其背后有着技术进 步与应用场景的支撑。各厂商相继推出视觉- 语言- 动作大模型（ 如阿尔特机器人的GOVLA、银河之眼 的G-0、思灵机器人的iLoabot-M ...

Group 1 - The core judgment presented by Rev Lebaredian emphasizes that the IT industry, valued at approximately $5 trillion, is a small part of the global economy exceeding $100 trillion, with significant value lying in the physical world sectors such as transportation, manufacturing, logistics, and healthcare [1][2] - The emergence of artificial intelligence enables machines to possess "physical intelligence," allowing for a true connection between the physical and information worlds, with robotics serving as a bridge for this transition [1][2] Group 2 - China is positioned uniquely to excel in the robotics and AI field, with nearly half of the global AI researchers and developers based in the country, alongside unmatched electronic manufacturing capabilities and a vast manufacturing base for large-scale deployment and testing [2] - NVIDIA's mission is to create computers specifically designed for the "toughest problems," necessitating the development of three types of computers: embedded computers in robots, AI factory computers for data processing and model training, and simulation computers for data generation and testing [2] Group 3 - Wang Xingxing views humanoid robots as crucial carriers for general-purpose robotics, suggesting that as general AI matures, the complexity of hardware requirements will decrease, making it easier for individuals to assemble humanoid robots similar to building a computer [3] - UTree Technology launched a humanoid robot priced at approximately 99,000 RMB last year, with a new version this year priced at around 39,000 RMB, supporting customization and expected to reach mass production by the end of the year [3] Group 4 - Wang He emphasizes that general-purpose robots will be revolutionary products in a market potentially worth trillions, with the core elements being the robot itself, the embodied intelligence model driving it, and the data supporting the model [3][4] - The next-generation humanoid robot project announced by Galaxy General and NVIDIA will utilize the Isaac platform for data collection and remote control, capable of training and deploying various task abilities in both simulated and real environments [3] Group 5 - Wang He predicts that the market for humanoid robots will grow exponentially, estimating that production will increase tenfold every three years, potentially surpassing the total output of industrial robotic arms [4] - The future of robotics will require a combination of top-tier computing power, simulation capabilities, cost-effective hardware engineering, and a large-scale training system driven by synthetic data to achieve widespread deployment [4]

Core Insights - The article introduces VLN-PE, a physically realistic platform for Vision-Language Navigation (VLN), addressing the gap between simulated models and real-world deployment challenges [3][10][15] - The study highlights the significant performance drop (34%) when transferring existing VLN models from simulation to physical environments, emphasizing the need for improved adaptability [15][30] - The research identifies the impact of various factors such as robot type, environmental conditions, and the use of physical controllers on model performance [15][32][38] Background - VLN has emerged as a critical task in embodied AI, requiring agents to navigate complex environments based on natural language instructions [6][8] - Previous models relied on idealized simulations, which do not account for the physical constraints and challenges faced by real robots [9][10] VLN-PE Platform - VLN-PE is built on GRUTopia, supporting various robot types and integrating high-quality synthetic and 3D rendered environments for comprehensive evaluation [10][13] - The platform allows for seamless integration of new scenes, enhancing the scope of VLN research and assessment [10][14] Experimental Findings - The experiments reveal that existing models show a 34% decrease in success rates when transitioning from simulated to physical environments, indicating a significant gap in performance [15][30] - The study emphasizes the importance of multi-modal robustness, with RGB-D models performing better under low-light conditions compared to RGB-only models [15][38] - The findings suggest that training on diverse datasets can improve the generalization capabilities of VLN models across different environments [29][39] Methodologies - The article evaluates various methodologies, including single-step discrete action classification models and multi-step continuous prediction methods, highlighting the potential of diffusion strategies in VLN [20][21] - The research also explores the effectiveness of map-based zero-shot large language models (LLMs) for navigation tasks, demonstrating their potential in VLN applications [24][25] Performance Metrics - The study employs standard VLN evaluation metrics, including trajectory length, navigation error, success rate, and others, to assess model performance [18][19] - Additional metrics are introduced to account for physical realism, such as fall rate and stuck rate, which are critical for evaluating robot performance in real-world scenarios [18][19] Cross-Embodiment Training - The research indicates that cross-embodiment training can enhance model performance, allowing a unified model to generalize across different robot types [36][39] - The findings suggest that using data from multiple robot types during training leads to improved adaptability and performance in various environments [36][39]

点击下方 卡片 ，关注" 具身智能 之心 "公众号 作者丨 Xiaoxiao Long等 编辑丨具身智能之心 本文只做学术分享，如有侵权，联系删文 >> 点击进入→ 具身智能之心 技术交流群 更多干货，欢迎加入国内首个具身智能全栈学习社区 ： 具身智能之心知识星球 (戳我) ， 这里包含所有你想要 的。 出发点与工作背景 本综述聚焦具身智能在机器人研究中的前沿进展，指出实现强大具身智能的关键在于物理模拟器与世界模 型的整合。物理模拟器提供可控高保真环境用于训练评估机器人智能体，世界模型则赋予机器人环境内部 表征能力以支持预测规划与决策。 文中系统回顾了相关最新进展，分析了两者在增强机器人自主性、适应性和泛化能力上的互补作用，探讨 了外部模拟与内部建模的相互作用以弥合模拟训练与现实部署的差距。此外，还提及维护了一个包含最新 文献和开源项目的资源库，网址为https://github.com/NJU3DV-LoongGroup/Embodied-World-Models-Survey， 旨在为具身 AI 系统的发展提供全面视角并明确未来挑战。 一些介绍 随着人工智能与机器人技术的发展，智能体与物理世界的交互成为研 ...

Core Viewpoint - Nvidia is increasingly emphasizing the concept of AI factories, which are designed to leverage AI for value creation, contrasting with traditional data centers that focus on general computing [1][2]. Group 1: Nvidia's AI Factory Concept - Nvidia's CEO Jensen Huang announced collaborations to build AI factories in Taiwan and Germany, featuring supercomputers equipped with 10,000 Blackwell GPUs [1]. - The AI factory concept includes a computational center and a platform to upgrade factories into AI factories, with a focus on simulation and digital twin technologies [4]. - The Omniverse platform is integral to Nvidia's strategy, allowing manufacturers to utilize AI for simulation and digital twin applications [2][3]. Group 2: Industry Applications and Collaborations - Various manufacturers are integrating Nvidia's AI technology through software from companies like Siemens and Ansys, enhancing applications in autonomous vehicle simulations and digital factory planning [3]. - Companies like Schaeffler and BMW are utilizing Nvidia's technology for real-time collaboration and optimization in manufacturing systems [3]. Group 3: AI Model Utilization - The industrial sector has been using small models for AI applications prior to the emergence of large models, focusing on data intelligence and visual intelligence [6][10]. - Small models are expected to continue to dominate industrial AI spending, with estimates suggesting they will account for 60-70% of the market [10][11]. Group 4: Cloud and Computational Needs - Nvidia's approach to building large-scale AI clouds is one option, but many companies prefer private cloud solutions due to data security concerns [13][14]. - The demand for computational power is expected to grow as AI applications become more prevalent, although current infrastructure may not be a bottleneck [15].

Core Viewpoint - The article discusses NVIDIA's significant advancements in the robotics sector, highlighting the launch of the GR00T N1 humanoid robot model and the overall growth potential of the robotics market, which is projected to reach a value of $10 trillion by 2030 [1][8]. Group 1: NVIDIA's Robotics Business Layout - NVIDIA has expanded its business into robotics, launching the Jetson series in 2014 and the Isaac platform in 2018, which includes hardware and software solutions for autonomous robots [2][3]. - The GR00T N1 model, introduced in 2025, features a dual-system architecture that allows for rapid and slow processing, enabling the robot to perform complex tasks and adapt to various environments [3][4]. - NVIDIA's robotics ecosystem encompasses hardware, software, and partnerships, with the Jetson series providing essential computing power and the Isaac platform offering simulation tools for developers [5][6]. Group 2: Robotics Market Size - The global robotics market is expected to grow from $100.59 billion in 2025 to $178.63 billion by 2030, with a compound annual growth rate (CAGR) of 12.2% [8]. - The humanoid robot market is projected to increase from $300 million in 2024 to $37.8 billion by 2035, driven by advancements in AI and robotics technology [8]. Group 3: Robotics Investment Dynamics - The robotics sector has seen a surge in investment activity, with significant funding rounds indicating investor optimism. For instance, companies like Zhijid动力 and Fourier have raised substantial amounts in recent financing rounds [12]. - The increase in funding reflects a growing interest from entrepreneurs in the robotics field, which is expected to drive technological innovation and broaden application scenarios [12].