Core Viewpoint - The domestic humanoid robot industry is experiencing a historic development opportunity driven by policy support and technological breakthroughs, with Yushu Technology, Zhiyuan Robotics, and Leju Robotics being the core competitors in this round of ranking [1][16]. Group 1: Company Strategies - Yushu Technology adopts a technology-focused strategy, emphasizing hardware development and cost control, with a strong brand influence in dynamic balance and high explosive motion [4]. - Zhiyuan Robotics follows a scenario-focused strategy, aiming for productization of technology and leveraging an open ecosystem to maximize its leverage effect, combining technical depth with commercial acumen [6]. - Leju Robotics employs a full industry chain integration strategy, controlling upstream core components through self-research and equity investment, while ensuring midstream production capacity through joint ventures and ecosystem cooperation [8][9]. Group 2: Manufacturing and Integration - In midstream manufacturing, Yushu relies on its proven "self-built capacity + supply chain optimization" model to enhance cost efficiency [8]. - Zhiyuan focuses on rapid prototyping and seeks manufacturing partners, indicating a reliance on large OEMs for future mass production [8]. - Leju positions itself as an "organizer" of the industry chain, creating deep adaptations between products and scenarios through strategic partnerships with major manufacturers [9]. Group 3: Market Deployment and Commercialization - Yushu maintains a steady commercialization pace, leveraging its existing market foundation to gradually penetrate complex scenarios [11]. - Zhiyuan aims for high-profile demonstrations to establish its long-term goal of a "universal robot," although its initial focus is on high-value tasks [11]. - Leju prioritizes high-paying scenarios to reduce commercialization risks, successfully deploying its humanoid robot "Kua Fu" in various fields, including industrial manufacturing and educational services [13][14]. Group 4: Industry Positioning and Future Outlook - Leju is likened to "BYD" in the robot field for its vertical integration and ecosystem construction, while Yushu is compared to "DJI" for its technical excellence and cost control [16]. - The recent 1.5 billion RMB financing for Leju signifies its advancement towards an IPO, with a dual strategy of resource locking and overcoming production bottlenecks [16]. - The comprehensive layout of Leju across the industry chain enhances its resource integration capabilities and risk resilience, although it faces challenges in management complexity and maintaining core technology focus [17].

然而，这类融合了光学、机械、电子与计算等多领域技术的高集成度智能微系统，仍面临诸多挑战。目前， 受制于传统运动部件的制造方法与驱动方式，视场调节机构的设计仍主要依赖 于 全刚性结构与微型驱动电 机。 此类结构通常体积较大（数十立方厘米）、重量较重（数百克），严重限制了机器人头部的有效负载能 力，并阻碍了其在对于尺寸、重量 更 为敏感平台（如紧凑型仿生机器人）上的应用。 因此，开发能够在光学成像系统、视场调节机构与目标识别算法之间实现协同优化与多维性能深度集成的仿生 视觉系统，具有重要的科学研究意义与工程应用价值。该系统应兼具轻量化紧凑型结构、高分辨率成像能力、 大视场多自由度动态调节特性以及强大的环境适应性。 ▍ 开发 BES 视觉系统：实现具有人眼解剖学等效特征的仿生眼球系统 针对仿生视觉系统所面临的突出问题， 国防科技大学智能科学学院吴学忠教授与肖定邦教授团队受人类眼球 生理结构启发，研制出一种基于三自由度折纸机构的新型仿生眼球系统（ BES ），成功突破了传统机电式仿 生眼球在物理尺寸与功能集成度之间难以权衡的技术瓶颈。 BES 在视场调节能力方面达到人眼水平，单眼视场范围超过 151.6° × 151 ...

Core Insights - Embodied intelligence has become a global buzzword, with countries competing in technology advancements, particularly in the U.S. which is forming a complete support system from basic research to industrial application [1][5] - China is a key player in the embodied intelligence sector, with significant growth in startups and investment, totaling over 27.1 billion yuan in the first seven months of the year [3][5] - The rise in the embodied intelligence industry is driven by supportive policies in both the U.S. and China, recognizing it as a critical future industry [5][6] Investment and Market Dynamics - In the first seven months of 2025, China saw 108 investment events in the embodied intelligence field, surpassing the total for 2024 [3] - Major companies like Alibaba, JD, and Tencent, along with automotive giants and manufacturing firms, are entering the embodied intelligence market, pushing for commercialization [3][5] Technological Developments - The industry is shifting from a focus on humanoid robots to practical designs based on specific scene requirements, leading to the emergence of two paths: remote operation and autonomy [6][10] - The combination of wheeled and bipedal robots is becoming essential, with wheeled robots gaining traction due to their efficiency and adaptability in various industrial applications [8][10] Operational Efficiency - Wheeled embodied intelligence robots are emerging as core engines of the industry due to their lower complexity and cost, with BOM costs being one-tenth of bipedal robots [10][11] - These robots are capable of performing tasks such as vehicle inspection and material sorting, demonstrating their commercial viability [11][13] Challenges and Future Outlook - Wheeled robots face challenges in transitioning from structured to semi-structured environments, while bipedal robots excel in high-risk scenarios [13][19] - The future of embodied intelligence will likely see parallel development of both wheeled and bipedal robots, addressing diverse needs across various sectors [13][19] Technological Architecture - The development of "big brain" and "small brain" capabilities is crucial for enhancing operational efficiency, with a focus on intelligent decision-making and precise execution [15][19] - The dual-mode architecture combining autonomous operation and remote control is becoming a practical choice for balancing technology maturity and commercial needs [17][19] Industry Competitions and Collaborations - The upcoming 2025 Zhongguancun Embodied Intelligence Robot Application Competition aims to evaluate and promote the capabilities of various companies in the sector [21][22] - The competition will cover multiple core scenarios and invite top experts to ensure a comprehensive assessment of the industry's leading players [22][24]

Core Insights - The article discusses the launch of the world's first waterproof dexterous hand, the WA series, by the Chinese startup Heiman Technology, which addresses a significant limitation in the dexterous hand market and expands the operational capabilities of robots in various environments [1][10]. Product Overview - The WA series includes two versions: the standard version for everyday tasks like washing dishes and vegetables, and the engineering version designed for underwater operations, capable of withstanding pressures at depths of up to 10,000 meters [3][5]. - The engineering version is tailored for applications such as marine farming, deep-sea scientific research, underwater engineering, archaeology, and rescue operations [3][12]. Technical Challenges - Key technical challenges in developing the waterproof dexterous hand include ensuring reliable sealing, adaptive grasping control, and resistance to corrosion and electromagnetic interference [3][5]. - The team utilized advanced sealing technologies and high-performance drive solutions to achieve the required functionality and reliability in extreme underwater conditions [5][6]. Team Expertise - The core team at Heiman Technology has extensive experience in robotics, with the founder and team members having backgrounds in military and industrial robotics, contributing to the development of various robotic systems [6][7]. - The team has a strong academic foundation, with members from top universities and significant industry experience, enhancing their capability to innovate in the field of dexterous hands [6][7]. Market Potential - The standard version of the WA series targets the rapidly growing humanoid robot market, which is projected to exceed $100 billion by 2030, with service robots expected to reach $40 billion by 2025 [10][12]. - The engineering version addresses specific needs in underwater operations, which are becoming increasingly important as marine exploration and resource extraction activities expand [12][14]. Business Strategy - Heiman Technology's business strategy focuses on solving specific engineering problems in vertical markets while aiming for long-term integration with the humanoid robot industry to enhance product stability and cost-effectiveness [15][10]. - The company has established partnerships with several universities and organizations to facilitate product testing and market entry, demonstrating a clear and pragmatic commercial logic [14][15].

Group 1: Humanoid Robots - The new humanoid robot "Xiao Bumi" by Songyan Power achieved impressive sales, with over 200 units sold within 3 hours of launch, priced at 9998 yuan, making it the first high-performance humanoid robot under 10,000 yuan [2] - The launch of "Xiao Bumi" aims to bring emotional value to users before humanoid robots fully enter the home service market, indicating strong market interest and potential for future household applications [2] Group 2: Medical Robotics - Xiamen University Affiliated Cardiovascular Hospital successfully completed the world's first cross-border remote robotic-assisted cardiac intervention surgery, demonstrating real-time collaboration between Chinese and French experts over a distance of over 10,000 kilometers [6][7] - The surgery utilized a robot system developed by Xiamen Hospital, showcasing its independent intellectual property and receiving high praise from Bordeaux University Hospital experts for its milestone significance in global medical cooperation [7] Group 3: Robotics Infrastructure - The humanoid robot pilot base and application promotion center in Hangzhou officially opened, with a total investment of nearly 200 million yuan and an area of over 15,000 square meters, aimed at addressing mass production challenges for humanoid robots [11] - The base integrates various functions, including indoor precision testing and outdoor scenario validation, and is part of the core of Hangzhou's Zhijiang Robot Industry Service Port, enhancing the city's innovation system in the robotics sector [11] Group 4: Advanced Robotics Technology - A research team from the University of North Carolina developed "DNA flower" micro-robots, which can rapidly fold and unfold in response to environmental changes, showing potential for targeted tumor treatment and environmental cleanup [13] - The technology is still in early testing stages but holds promise for future applications, including digital storage capabilities [13] Group 5: Vision Restoration Technology - The PRIMA artificial vision technology has enabled a 70-year-old woman, previously blind due to age-related macular degeneration, to regain functional central vision, with 84% of patients in clinical trials achieving similar results [16] - PRIMA consists of an implant and special glasses, representing a significant advancement in vision restoration, with ongoing regulatory processes in Europe and the FDA [16]

近日，辛辛那提大学的研究团队搞出了一个有意思的东西—— 仿蛾 扑翼无人 机 。 它最大 的 亮 点是用了 极值搜索 （Extremum-Seeking）反馈控制，简单说就是 不需要AI、不需要复杂模型，靠着简单的反馈机制就能实现稳定飞行。 硬件配置上，它有四个金属丝和织物做的翅膀，通过高速独立扑动来控制飞行姿态——快到肉眼只能看到一片模糊，跟蜂鸟似的。 功能方面很有意 思： 能悬停、能倒着飞，遇到风或障碍物会自动调整，还能跟着移动的光源走。 飞行时会有点小晃 悠，但这不是缺陷而是特意设计的——系统需 要这 种扰 动来不断优化飞行状态。 此外，它还能模拟飞蛾、蜻蜓、蜂鸟等各种飞行生物的姿态。而且整个系统可以做得非常小，特别适合一些特殊应用场景。 ▍ 悬停飞行：曾被认为 "不可能"的奇迹 昆虫和蜂鸟的悬停飞行一直是个谜。 它们如何在空中保持稳定？如何实时调整翅膀来对抗扰动？这些问题困扰了科学家几十年。 不久前，辛辛那提大学的研究团队还提出了一个 全新的解释框架 ——将昆虫悬停飞行描述为一种天然的极值搜索（Extremum Seeking，ES）反馈系统。 | XIV > math > arXiv:2402.049 ...

Core Viewpoint - The company, Obi Zhongguang, reported significant growth in its third-quarter earnings, with a revenue increase of 103.5% year-on-year, indicating a strong recovery and operational improvement in its core business segments [1][2]. Financial Performance - For the first three quarters of 2025, the company achieved a revenue of 714 million yuan, with a net profit of 108 million yuan and a non-recurring net profit of 68.57 million yuan, marking a turnaround from previous losses [1][4]. - In the third quarter alone, revenue reached 279 million yuan, a year-on-year increase of 102.49%, and a net profit of approximately 47.83 million yuan, reflecting a 33.32% increase from the previous quarter [1][2]. Business Growth Drivers - The company has seen rapid growth in its three-dimensional scanning, payment verification, and various robotics sectors, driven by improvements in the upstream 3D visual perception supply chain and the expansion of downstream application scenarios [2][4]. - The biological recognition and AIoT sectors have emerged as core growth engines for the company, with significant collaboration with ecosystem partners to develop smart terminal devices for various applications [5][6]. R&D and Manufacturing Strategy - The company has focused on enhancing its R&D efficiency and reducing ineffective cost inputs, with an average R&D investment of nearly 70% of annual revenue from 2022 to 2024 [4][9]. - A recent fundraising plan aims to raise up to 1.918 billion yuan to strengthen its R&D platform for robotics AI vision and spatial perception technology, as well as to establish a manufacturing base for AI visual sensors and smart hardware [9][10]. Market Position and Future Outlook - The company has achieved over 70% market share in the domestic service robot and commercial/industrial mobile robot 3D visual fields, indicating strong competitive positioning [10]. - Future plans include increasing annual production capacity for 3D visual sensors and consumer-grade application devices to 5 million and 2 million units, respectively, targeting high-demand automation scenarios [12].

Core Insights - Jingxiang Mo has left K-Scale Labs to establish a new company, Gradient Robotics, focusing on critical issues in robotics and physical AI in the U.S. [1][6] - During his tenure at K-Scale Labs, he led the development of K-Bot and Z-Bot, achieving significant milestones in under eight months with a small team [2][10]. Group 1: K-Scale Labs Achievements - K-Scale Labs successfully created K-Bot, an open-source humanoid robot platform, which was the first of its kind in the U.S. for consumer use, featuring a reinforcement learning-based motion system [10][11]. - The K-Bot project saw rapid success, with 150 units sold out quickly, generating over $2 million in sales, and attracting clients like Google DeepMind and OpenAI [11][13]. - The Z-Bot project aims for mass production with a price point below $1,000, and has already garnered significant interest, with over 20,000 people on the waiting list [13]. Group 2: Industry Context and Future Directions - The robotics and physical AI sectors are becoming competitive focal points, with companies like Physical Intelligence (PI) and FieldAI emerging as key players, both valued over $2 billion [18][29]. - PI focuses on developing general AI models for robots, while FieldAI aims to create AI models that can operate in real-world environments, showcasing the industry's demand for technological breakthroughs [24][27]. - Gradient Robotics is positioned to leverage its experience in open-source humanoid robotics to address the challenges in general robotics technology, drawing parallels to advancements in autonomous driving [4][30].

Core Insights - The research team from EPFL has developed an innovative implantable closed-loop spinal neuroprosthesis system that successfully helps paraplegic patients regain walking ability by integrating spinal cord electrical stimulation with rehabilitation robots [1][3]. Rehabilitation Technology - Rehabilitation robots have made significant advancements over the past decades, but they face a fundamental challenge: paraplegic patients cannot generate sufficient muscle activity [5]. - The core mechanism of gait rehabilitation is to promote activity-dependent reorganization of the nervous system, which requires strong, coordinated, and repetitive activation of the neuromuscular system [5]. - Traditional solutions like Functional Electrical Stimulation (FES) have limitations, including cumbersome electrode placement and muscle fatigue, and they do not accurately simulate natural neural control patterns [5][6]. Innovative Approach - The Swiss team's approach involves activating spinal motor neurons through epidural electrical stimulation (EES) rather than directly stimulating muscles, leading to more natural and coordinated muscle activity patterns [6]. - The developed system can adjust stimulation parameters in real-time based on different rehabilitation robots and exercise modes, achieving "activity-dependent biomimetic electrical stimulation" [6]. Closed-Loop Control System - A flexible and precise closed-loop control system is essential for the integration of spinal cord stimulation and rehabilitation robots [8]. - Key components include an implantable pulse generator and an electrode array with 16 electrodes for precise targeting of spinal segments controlling lower limb muscles [8]. - A real-time control platform allows for adjustments based on external sensor signals, with a latency of only 134±26 milliseconds [8]. Sensor Technology - The system employs various sensor modalities, including: - Inertial Measurement Units (IMUs) for detecting limb movement and key gait events [9]. - Ergonomic buttons for patients to actively trigger stimulation [9]. - Force-sensing pedals for bicycle training to measure applied force and adjust stimulation intensity accordingly [9]. Clinical Testing and Results - The system was tested on five spinal cord injury patients, covering a range of rehabilitation activities from bed rest to outdoor activities [11]. - In tests with the Lokomat exoskeleton, patients showed significant muscle activity improvement and better timing of muscle activation, indicating increased active participation in movement [12]. - The system allowed patients to walk independently on various terrains and even participate in a charity run, demonstrating its effectiveness in real-world scenarios [12][14]. Long-Term Rehabilitation Effects - Long-term follow-up of four chronic spinal cord injury patients revealed improvements in lower limb motor scores, even after electrical stimulation was turned off, indicating true rehabilitation effects [14][15]. - Patients gradually reduced their reliance on robotic assistance, showcasing the combined effect of robotic support and electrical stimulation [17]. - The technology does not require modifications to existing rehabilitation equipment, making it easier for therapists to implement [18]. Future Directions - The current neurostimulator is adapted from deep brain stimulation devices and needs a dedicated hardware platform for spinal stimulation [18]. - Larger clinical trials are necessary to validate the findings, but the research opens new avenues for rehabilitation of spinal cord injury patients [18].

当宇树 最新 H2 人形机器人踩着节奏跳出 跳舞 ，当智元机器人 G2 的灵巧手稳稳 拉弓射箭 ， 这些看似流畅 的 " 拟人动作 " 背后，藏着一个容易被忽视的核心支撑 —— 动捕技术。 在 202 5 年 IROS 国际机器人与自动化展览会上，青瞳视觉展出的光惯混合动捕设备成为焦点： 它仅用 2 ~ 3 台相机就能精准捕捉手指关节的细微动作 ，还能在玻璃、金属等强反光环境下稳 定工作，一举解决了传统 动捕 " 遮挡丢数据 "、" 惯性飘移 " 的行业痛点。 张海威解释 ， 运动智能的核心是让机器人掌握平衡与协调， 比如走路时重心如何转移、转弯时关节如何配 合，这些数据都需要通过高精度动捕采集。而比会跑会跳更重要的，是机器人的作业智能 ， 也就是 " 会干活 " 的能力。这也是智元、优必选等企业研发灵巧手的核心诉求。 据悉， 青瞳视觉早在去年就推出了光学手指动捕设备，能捕捉人类手指屈伸、抓握的细微动作， 国内 例如灵 巧智能等 多数灵巧手企业都在使用这套设备训练机器人。 " 比如抓杯子，人类手指会根据杯子的形状调整力度，指尖关节的弯曲角度误差不能超过 1 度，传统动捕根本 做不到这么精准。 " 张海威举例， ...