再 Call 灵巧手：本轮 T 链灵巧手的重要性再探讨！ 20251117 摘要 人形机器人灵巧手是实现精细操作的关键，其性能直接影响产品发布， 因此是当前最紧迫的技术瓶颈。产品延迟发布的核心原因之一正是由于 灵巧手的性能和寿命不达标，新供应商的选择值得关注。 灵巧手技术结构复杂，单手及前臂需约 25 个执行器，主要由电机、微 型丝杠、腱绳等组成，并可能搭配电子皮肤。新一轮更新大概率沿用微 型丝杠加腱绳方案，并可能引入电子皮肤技术。 大部分公司已在 2025 年 11 月中上旬向海外客户送样，预计 11 月底收 到反馈，有望年底前拿到定点订单。2026 年二三月份发布前夕，灵巧 手核心零部件供应链非常值得重视。 零部件方面，电机、微型丝杠、电子皮肤环节可能出现增量、直径变化 或扭矩调整等变化，投资者应密切关注这些技术参数的调整。 微型丝杠环节，除浙江荣泰外，重点关注恒立液压的进展。电子皮肤方 面，看好汉威科技，该公司有望逐步将电子皮肤搭配至灵巧手上。 腱绳和电机端，可关注南山智尚、明志电器以及伟创电气，这些公司有 望成为新的供应链成员，具备较大的投资潜力。 方案，并可能引入一些电子皮肤相关技术。 当前阶段对灵 ...

点击下方 卡片 ，关注" 具身智能 之心 "公众号 作者丨 Nhat Chung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。 针对机器人在非马尔可夫场景下缺乏对象级记忆的问题，阿肯色大学联合卡内基梅隆大学等研究团队提出LIBERO-Mem基准套件与Embodied-SlotSSM模型，通过 结构化对象记忆与时间序列建模，实现长时程、部分可观测环境下的稳健操作决策。 核心贡献 LIBERO-Mem基准：非马尔可夫机器人操作评估 设计目标 聚焦对象级部分可观测-非马尔可夫场景，通过引入对象身份、位置、关系历史的模糊性，强制模型依赖时间推理而非仅当前视觉信息。 核心特征 任务类型：包含四类任务（figure 1），覆盖不同记忆维度 现实机器人操作场景中，任务成功依赖对象交互历史（如"是否已操作过某个物体""物体之前的位置"），而非仅当前观测。 现有视觉-语言-动作模型多遵循马尔可夫假设，仅依赖即时感官输入，缺乏对象级记忆机制，在重复操作、视觉相似 ...

详情请戳视频 以上就是本期全部内容，如果您想了解更多 最新 企业 动态， 欢迎登录 睿兽分析 查看，解锁 赛道 图谱 以及 行业报告 。 硅谷人形机器人公司 K-Scale Labs, 烧光今年上半年刚完成的 400 万美元种子轮融资后，创始人宣 布无法完成预计的 A 轮融资，正式宣告关停。 大家好，我是大邦，今天一起聊一聊 K-Scale Labs 。 成立仅 18 个月！ K-Scale Labs 宣布倒闭！ ...

Group 1 - The core product showcased at the event is a lightweight exoskeleton device weighing less than 2 kilograms, designed to assist individuals with mobility challenges in activities such as climbing stairs and hiking [2] - Over 5,000 new products and achievements were released, with more than 90% of the physical exhibits featuring "high, precision, and cutting-edge" technologies and products, and over 20% being debut displays [2] Group 2 - The event attracted over 450,000 professional visitors [3] - A total of 1,023 supply-demand matching and investment financing projects were signed, with the intended transaction and financing amount exceeding 170 billion yuan [3]

AMD's Robotics Solutions - AMD provides embedded computing portfolio including X86 embedded processors and adaptive compute solutions for real-time control and AI [1] - AMD's Ryzen Embedded processors deliver performance per watt for factories and autonomous industrial robots [1] - AMD's adaptive computing products, such as Zynq and Versal, enable vision AI, motor control, real-time networking, and sensor fusion [2] - AMD supports open-source software, facilitating ROS 2-based systems using Kria SOM platforms and the Kria robotics stack [3] AI in Robotics - AI-powered digital twins can enhance efficiency, dexterity, and autonomy in factories, predicting system failures [4] - Physical AI promises to transform robotics in areas like industrial process optimization, healthcare, and humanoid personal assistance [4] - Embodied AI allows robots to learn through physical interactions, improving safety and efficiency [5] - Real-time sensor processing from AMD can accelerate ROS functions, enabling robots to adapt to new tasks [5]

Core Insights - Knightscope, Inc. is launching the K7 Autonomous Security Robot, which aims to address a $2.6 trillion problem in public safety by providing 24/7 coverage at an estimated cost of $15 per hour [7] - The company has a strong financial position with $24.2 million in cash as of August 2025, a clean balance sheet, and no toxic debt [7] - Knightscope has established multiple contracts exceeding $1 million across various sectors, including healthcare, education, gaming, and local government [7] - A strategic partnership with Palantir Technologies enhances Knightscope's capabilities by integrating advanced AI and defense infrastructure [7] Company Overview - Knightscope was founded on April 4, 2013, in response to significant public safety incidents, aiming to transform public safety through robotics and AI technologies [3][4] - The company's long-term goal is to make America the safest country in the world [4] Leadership - William Santana Li has been the Chairman and CEO since April 2013, bringing over 30 years of experience from the global automotive sector and various startups [5][6]

Core Insights - Xpeng, a Chinese EV startup, is focusing on creating a humanoid robot named Iron that is designed to be "highly human-like" to compete with Tesla's Optimus robot [1][2] - The company aims to sell 1 million units of Iron annually by 2030, with mass production expected to start by the end of 2026 [3][4] - Xpeng plans to utilize Iron in retail settings and as tour guides, indicating a strategic approach to market entry [4] Company Developments - Xpeng has showcased Iron with customizable body types, including male and female variants, to enhance user appeal [2] - The company is set to launch seven new vehicles next year, including three robotaxi models, and plans to begin mass production of a $280,000 flying car in 2026 [6] - Xpeng has achieved significant sales growth this year and is nearing profitability, positioning itself as a strong competitor in the EV market [5] Market Outlook - The CEO of Xpeng predicts that the market for humanoid robots could surpass that of the automotive market in the long run [4] - Xpeng's strategy to incorporate humanoid robots into its stores next year reflects a proactive approach to leveraging technology for customer engagement [4]

LATEST: ⚡ Tether is reportedly considering a $1.15B investment in German robotics startup Neura, a move that would value Neura between $9.3B and $11.6B, according to a recent Financial Times report. https://t.co/QbfrRGvcJR ...

Group 1 - The event showcased the integration of advanced robotics in sports, highlighting the participation of humanoid robots and robotic dogs at the 15th National Games [1][2] - The robotic dogs were utilized for transporting equipment for three throwing events, marking the first time such technology was employed in this capacity at the National Games [2][5] - The robotic dogs have a load capacity of 40 to 60 kilograms, a battery life exceeding 3 hours, and can move at speeds up to 6 meters per second, demonstrating significant improvements over previous models [5] Group 2 - The humanoid robot Walker C, which participated in the event, is designed for high flexibility and can run at a speed of 6 kilometers per hour, showcasing advanced movement capabilities [7][8] - Custom developments for the Walker C included venue mapping and performance optimization specifically for the National Games, enhancing its operational efficiency [8] - The technology behind these robots is supported by over a decade of research and development, focusing on servo motor technology and motion control, with plans for future enhancements [8]

Core Insights - Digua Robot has announced a deep collaboration with Alibaba's Damo Academy's embodied intelligence laboratory to develop a new generation of domestic embodied intelligence core platform [1] Group 1: Partnership Details - The collaboration will focus on the Damo Academy's open-source VLA model and the RCP (Robotics Context Protocol) specifically designed for the robotics field [1] - Digua Robot's new domestic high-performance computing platform, RDK S600, will undergo deep software and hardware adaptation in this partnership [1] Group 2: Goals and Objectives - The partnership aims to empower the efficient research and development of embodied robots and facilitate large-scale implementation [1] - Both parties will jointly build an open industrial ecosystem and an educational research ecosystem to accelerate the exploration and innovative evolution of embodied intelligence through industry-academia-research collaboration [1]