2026年2月11日，星海图完成10亿元B轮融资，本轮投资方包括一线产业资本金鼎资本、北汽产投、碧鸿 投资，顶尖PE基金正心谷资本、前海方舟和国际化基金毅峰资本。五位主要老股东凯辉基金、美团龙珠、 今日资本、襄禾资本和高瓴创投联手超额或满额追加。 此轮融资不仅是市场对星海图技术路线与商业化进展的双重认可，更标志着具身智能正从技术突破迈向规 模落地的价值拐 点 。 股东结构：产业与国际化资本深度协同，构筑高质量增长底盘 本轮融资汇聚了产业资本、国际化基金与老股东的共同支持： 通知 | 关于举办"2026大湾区工业具身智能大会"的通知 产业资本入局： 金鼎资本、北汽产投、碧鸿投资等产业资本的加入，将加速星海图在智能制造、汽车工业 等核心场景的技术融合与战略协同。 顶级PE基金青睐： 正心谷资本、前海方舟的青睐，代表着全周期视角的长期价值投资者对长坡厚雪的赛 道和公司的认可。 国际化资本加码： 毅峰资本的加持，则为公司的全球化布局注入了国际视野与资源，助力技术与产品走向 海外市场。 老股东联手追加： 五位主要老股东联手持续加码，是对星海图超强执行力、团队迭代力与业务成长性的极 佳认可和持续信心。 技术实力：坚持"全 ...

Core Viewpoint - The 2024 World Humanoid Robot Games in Beijing represents a significant shift from laboratory testing to real-world applications, showcasing the potential for commercial opportunities and industry restructuring in the humanoid robotics sector [1]. Event Highlights - The event features diverse project settings that closely resemble real-world applications, including tasks like hotel cleaning, factory material handling, and hospital drug sorting, indicating a move from "showcase" to "practical" robots [3][4]. - A unique "global first humanoid robot half marathon" tests robots' endurance, battery life, and coordination, highlighting technological integration and energy efficiency [3]. Participating Teams - Leading companies such as Xinghaitu, backed by Meituan, are providing exclusive robot solutions for the event, indicating strategic investments in end-service robotics [5]. - Yushu Technology is establishing itself as a foundational player in the industry with its open hardware ecosystem, akin to the Android model in smartphones [5]. - A diverse array of domestic companies, including Tiangong, Jiasu Jinhua, and Fourier, showcases China's multifaceted approach to humanoid robotics [5]. - Academic institutions and international teams from 15 countries are participating, emphasizing the integration of research and global competition in the field [5]. Technological Breakthroughs - Advances in visual recognition and modeling techniques are being demonstrated, such as a high-reflective object detection model developed by a team from Hefei [7]. - Yushu robots are noted for their stability and dynamic control during competitions, showcasing advancements in motion control [8]. - The R1 Pro from Xinghaitu exemplifies the potential of service robots through its ability to understand task processes and execute actions effectively [8]. Industry Trends - The humanoid robotics industry is transitioning from technology validation to commercial application, marking a pivotal moment for market entry [9]. - Capital is increasingly concentrating on a few leading companies, indicating a potential reshaping of the competitive landscape in the next 3-5 years [10]. - Collaboration across the supply chain, including sensors, motors, AI, and operating systems, is essential for success in the humanoid robotics sector [11]. Investment Logic - Attention should be directed towards foundational technology suppliers, such as those providing servo motors and visual recognition chips, which are critical to robotic functionality [12]. - Investment in platform companies like Yushu Technology and UBTECH, which offer robotic platforms and ecosystems, is recommended due to their growth potential [12]. - Companies focused on specific applications, such as Xinghaitu and Fourier, are positioned for clear commercial pathways and profitability [12].

Core Viewpoint - The 2025 World Robot Conference (WRC) showcased advancements in humanoid robots, emphasizing the integration of 5G-A technology and embodied intelligence to enhance operational capabilities and adaptability in complex environments [1][3]. Group 1: Humanoid Robots - The "Kua Fu" humanoid robot by Leju demonstrated significant breakthroughs in communication and decision-making through 5G-A technology, enabling real-time control over distances of 1200 kilometers [1][3]. - The "Kua Fu" robot features a dual-brain system, enhancing human-robot interaction and stability in flexible environments, showcasing its potential in high-risk scenarios like firefighting and chemical industries [3][6]. - The "T800" humanoid robot by Zhongqing is designed for heavy-duty tasks, featuring 41 degrees of freedom and advanced sensor integration for real-time environmental adaptation [7]. Group 2: Industrial Robots - Cyborg Robotics presented the Cyborg-H01 dexterous hand and Cyborg-R01 humanoid robot, focusing on high precision and energy efficiency in industrial applications [4][6]. - The CW series robots by Chuan Robotics combine autonomous mobile platforms with collaborative robotic arms, enhancing flexibility in production environments [30]. Group 3: Perception and Sensing Technologies - Aobi Zhongguang introduced the Pulsar ME450 3D LiDAR and Gemini 345Lg dual 3D camera, enhancing visual perception capabilities for outdoor robots [12][14]. - Pasini launched the third-generation multi-dimensional tactile sensor matrix PX-6AX-GEN3, providing advanced sensing solutions for embodied intelligent systems [10][25]. Group 4: General-Purpose Robots - Zhi Ping Fang showcased the "Ai Bao" robot, capable of performing various tasks in simulated environments, highlighting its adaptability and learning capabilities [17][19]. - The "Wanda" platform by Youliqi demonstrated its application in multiple real-life scenarios, emphasizing its interactive capabilities and service-oriented design [23][24]. Group 5: Collaborative and Ecosystem Development - Xinghai Tu introduced the G-0 VLA model, which enhances human-robot collaboration and opens up new possibilities for developers through a comprehensive toolkit [20][22]. - The collaboration among various companies at WRC indicates a growing ecosystem focused on advancing embodied intelligence and robotics technology across multiple sectors [41].

Group 1 - The 2025 World Robot Conference (WRC) opened in Beijing, showcasing a variety of practical robots, including those for factories, restaurants, and home assistance [1] - The "TianGong 2.0" robot, an upgraded version of the humanoid robot, is designed for sorting and quality inspection tasks in factories, demonstrating advanced capabilities [2][3] - The introduction of a universal multimodal visual perception system allows robots to operate autonomously, enhancing their functionality and reducing reliance on remote control [2][3] Group 2 - JD Group presented over 50 cutting-edge robot products, leveraging its supply chain advantages to connect with emerging tech companies and provide a seamless shopping experience for consumers [3][4] - The JoyInside intelligent platform was launched, enabling interactive capabilities for smart robots, with several leading robot brands integrating this technology [4] - Various assistive devices for elderly individuals, such as stair-climbing equipment, were showcased, highlighting the potential for smart technology in enhancing daily living [4] Group 3 - The exhibition featured interactive projects, including a humanoid robot that can perform household tasks based on voice commands, demonstrating the practical applications of embodied intelligence in home settings [5] - The "guide dog" robot, designed for visually impaired individuals, can recognize traffic signals and obstacles, providing navigation assistance and route planning [6] - Over 200 domestic and international robot companies participated in the event, displaying more than 1,500 exhibits, including the BionicBee, the smallest flying biomimetic device [6]

Core Insights - The core viewpoint is that 2026 will mark the second half of embodied intelligence, focusing on application maturity on both supply and demand sides [1][34]. Group 1: Development and Challenges - The embodied intelligence industry is currently perceived to be in a "technical bottleneck" phase, with a significant need for high-quality data and a correct ontology to drive progress [3][4][10]. - The lack of high-quality data is attributed to the absence of a standard ontology, which is essential for effective data collection and model training [4][11]. - The current focus should be on achieving object and action generalization, as scene and ontology generalization are less critical at this stage [17][27]. Group 2: Product and Model Structure - The ideal product structure for embodied intelligence is a combination of "hardware + pre-trained models + post-training tools," which allows for effective task execution in specific environments [6][9]. - The training process involves a two-phase approach: pre-training to understand basic interactions with the physical world and post-training for specific tasks [21][23]. - The model architecture includes a "slow thinking" component for logical reasoning and a "fast execution" component for real-time actions, which is crucial for operational efficiency [22][19]. Group 3: Market Dynamics and Future Outlook - The market is returning to a rational state, with companies exploring practical applications of embodied intelligence rather than unrealistic expectations of immediate widespread adoption [33][34]. - By 2026, the supply side will see matured robot bodies and initial generalization capabilities, while the demand side will have clearer application scenarios [32][34]. - The growth of the developer community is essential for the prosperity of the embodied intelligence market, as they will create diverse applications that drive value [28][29].