1、 宇树科技将发布四足机器人新品 近日，宇树科技官方发布新品预告，宣布将推出一款全新的四足机器狗产品。宇树科技在四足机器人领域 深耕多年，此前已凭借多款产品积累下良好口碑与技术基础。此次新品备受关注，据官方透露，其动力性 能表现十分亮眼，约等于现有产品Go2的两倍。这意味着新品在运动能力、负载能力等方面或许会有显著 提升，能更好地适应复杂多变的场景需求，无论是工业巡检、应急救援，还是科研教育、娱乐互动等领 域，都可能展现出更强大的应用潜力。目前，关于这款新品四足机器狗的其他详细参数与特性尚未完全公 布，行业内外都在翘首以盼其正式发布。 2、 富士康投用人形机器人助力英伟达AI服务器生产 近日，全球最大电子产品制造商、英伟达主要AI服务器制造商富士康称，将在休斯顿工厂部署人形机器 人。该工厂承担着为英伟达生产AI服务器的任务。据悉，富士康与英伟达是长期战略合作伙伴，在AI服务 器制造、自动驾驶、机器人等多领域合作紧密。此次休斯顿工厂将成为首批部署由英伟达Isaac GR00T N 模型驱动人形机器人的工厂之一，双方旨在打造世界领先AI智慧工厂。Isaac GR00T N是英伟达推出的开 源基础模型，可加速人形 ...

斯坦福大学助理教授、 Physical Intelligence （ PI ）联合创始人 Chelsea Finn 近日在社交平台 X 上连续 发文，为其斯坦福课题组最新研究 " 点赞 " ： " 生成看起来不错的视频很容易，难的是构建一个真正对机器 人有用的通用模型 —— 它需要紧密跟随动作，还要足够准确以避免频繁幻觉。我们在这两条战线上都取得了 长足进步。 " 这项进步，正是她与清华大学陈建宇团队联合提出的可控生成世界模型 "Ctrl-World"—— 一个 能让机器人在 " 想象空间 " 中完成任务预演、策略评估与自我迭代的突破性方案，其相关论文《 CTRL- WORLD: A CONTROLLABLE GENERATIVE WORLD MODEL FOR ROBOT MANIPULATION 》已发布 于 arxiv 平台，核心数据显示：该模型使用 " 零真机数据 " ，大幅提升策略在某些在下游任务的指令跟随能 力，成功率从 38.7% 提升至 83.4% ，平均改进幅度达 44.7% 。 Ctrl-World 专为通用机器人策略的策略在环轨迹推演而设计。它生成联合多视角预测（包括腕部视角），通过帧级 ...

Core Viewpoint - The article discusses the rise of "embodied intelligence" as a transformative force in the next industrial revolution, emphasizing its potential to reshape global competition and economic landscapes, particularly in the Yangtze River Delta region of China [1][3]. Group 1: Regional Development and Strategy - The Yangtze River Delta, as a key economic engine in China, is leveraging its strong manufacturing base, dense innovation resources, and open market environment to seize the opportunities presented by embodied intelligence [1]. - The region is guided by policies such as the "Yangtze River Delta Regional Integration Development Plan," focusing on "integration" and "high quality" to promote industrial transformation and upgrading [1][3]. Group 2: Huzhou's Strategic Positioning - Huzhou is positioning itself as a leader in the "embodied intelligent robotics" sector, aiming to create a regional innovation community that influences both national and global markets [3][5]. - The city is rooted in a strong manufacturing background and is accelerating the integration of technological and industrial innovation to foster new productive forces [5]. Group 3: Modern Industrial System - Huzhou is developing a "1366" modern industrial system, with artificial intelligence as the leading industry, and is focusing on three key sectors: modern textiles, special alloys, and green home furnishings [5][6]. - The city is also actively investing in six emerging industries, including new energy vehicles and intelligent equipment, as well as six future industries such as new energy aircraft [5][6]. Group 4: "One Core, Two Wings" Development Model - Huzhou is implementing a "One Core, Two Wings" industrial development model, with the "Xisai Science Valley" as the core engine for innovation and two newly expanded embodied intelligence industrial parks as the supporting wings [8][10]. - This model aims to create a closed-loop ecosystem that supports research, rapid commercialization, and large-scale production of innovative technologies [10]. Group 5: Upcoming Investment Conference - Huzhou will host an "Artificial Intelligence (Embodied Intelligent Robotics) Industry Investment Matching Conference" on November 8, 2025, to attract top domestic capital, technology, and talent [10][12]. - The conference aims to showcase Huzhou's industrial ecosystem and facilitate discussions on industry trends, resource matching, and project collaborations [12][13].

Core Insights - The article discusses the development and optimization of "vine robots," inspired by the growth and flexibility of vine plants, which can navigate through challenging environments and perform tasks in areas inaccessible to traditional robots [1][3]. Group 1: Key Features and Applications - Vine robots can explore life signs in rubble, search for leaks in narrow pipes, and access unknown environments, successfully completing tasks in urban rescue training sites, archaeological sites, and salamander cave habitats [3]. - The flexibility of vine robots allows them to operate in complex environments, but their performance is limited by factors such as top load, design parameters, and environmental adaptability [5][6]. Group 2: Manipulability Challenges - The manipulability of vine robots is influenced by three main factors: the impact of top load, the ambiguity of design and control parameters, and poor environmental adaptability [6]. - A research team from the University of Notre Dame and MIT Lincoln Laboratory focused on optimizing the manipulability of vine robots by analyzing the effects of top load, chamber pressure, length, diameter, and actuator design through systematic experiments [8]. Group 3: Experimental Findings - Experiments revealed that increasing top load significantly reduces the robot's bending ability, especially beyond 100 grams, which limits its operational range [13]. - Chamber pressure experiments showed that the feature length initially increases with pressure, peaking at 5.52 kPa, before decreasing due to excessive rigidity [14]. - Length experiments indicated that longer bodies enhance horizontal movement but reduce vertical movement, necessitating a balance between flexibility and structural stability [16]. - Diameter experiments demonstrated that while diameter affects collapse resistance, it has limited impact on manipulability once structural integrity is ensured [17]. Group 4: Design and Control Guidelines - The research team established design and control guidelines to optimize vine robot performance, emphasizing the need to minimize top load and balance length for flexibility and stability [28]. - Recommendations include using lightweight sensors and modular designs to enhance maneuverability and selecting actuator designs based on required pressure ratios for specific tasks [28][29]. Group 5: Future Directions - Future research will address issues related to non-reset phenomena and explore low-latency materials and proprioceptive sensing technologies to improve precision [33]. - The development of higher pressure-resistant actuator designs aims to synchronize rapid growth and high curvature turning, expanding the application range of vine robots in urban rescue, archaeological exploration, and industrial inspection [33].

Core Insights - The article discusses the limitations of Transformer models in handling long-term memory tasks and introduces Memo, a new architecture designed to enhance memory efficiency in long-sequence reinforcement learning tasks [1][3][18] Group 1: Memo Framework - Memo mimics human note-taking behavior by allowing the model to autonomously generate and store summaries of past experiences, enabling efficient retrieval of long-term memory with minimal memory overhead [3][5] - The framework processes long input sequences in segments and generates a fixed number of optimized summary tokens at the end of each segment [4][5] Group 2: Technical Implementation - Memo employs a special attention masking mechanism to ensure the model accesses past information only through summary tokens, creating a conscious information bottleneck [6] - It utilizes flexible positional encoding to help the model understand the temporal position of observations and summaries, which is crucial for causal relationships [6] - The introduction of segment length randomization during training enhances the model's adaptability to varying task rhythms [6] Group 3: Experimental Validation - Memo was tested in two embodied intelligence scenarios: the ExtObjNav task and the Dark-Key-To-Door task, comparing its performance against baseline models like Full Context Transformer (FCT) and Recurrent Memory Transformer (RMT) [7][11] - In the ExtObjNav task, Memo demonstrated superior performance, reducing the number of context tokens used by 8 times while maintaining strong reasoning capabilities beyond the training sequence length [9] - In the Dark-Key-To-Door task, Memo consistently remembered the locations of the key and door, while FCT showed significant performance decline after a certain number of steps, highlighting the challenges faced by full-context models [11] Group 4: Key Findings from Ablation Studies - The cumulative memory mechanism of Memo outperforms fixed memory models, akin to human wisdom accumulation rather than relying solely on recent experiences [14] - Long-range gradient propagation is essential for effective memory utilization, as limiting gradients to short-term memory significantly degrades performance [17] - An optimal summary length of 32 tokens strikes a balance between information compression and retention, as excessive summary tokens can introduce redundancy and noise [17] Group 5: Conclusion and Future Directions - Memo represents a significant advancement towards more efficient and intelligent long-term reasoning in AI, allowing models to autonomously manage their attention and memory [18] - The memory mechanism has broad applications, including autonomous navigation robots and personalized systems that understand long-term user preferences [18] - Future research will focus on enhancing the adaptability and interpretability of memory mechanisms, as well as balancing memory stability and flexibility [18]

Core Insights - The article discusses the development of a flexible spiking hair sensillum (FISH) inspired by spider hair receptors, which enables non-contact perception with ultra-low power consumption [2][3]. Group 1: FISH Development and Characteristics - FISH can convert airflow signals into electrical pulses in real-time, with a power density of less than 100 nW/cm² and energy consumption of approximately 660 pJ per sensing event, which is two orders of magnitude lower than traditional non-contact sensors [2][11]. - The sensor consists of a hair-like sensor made from polyimide (PI) and a flexible TS memristor, which together facilitate the conversion of airflow information into pulse sequences [4][6]. Group 2: Sensor Performance - The hair-like sensor can detect airflow speeds as low as 0.4 m/s, with a minimum detection limit of 0.04 m/s achievable by adjusting the PI substrate thickness. It has a response time of about 40 milliseconds and a recovery time of 26 milliseconds at 7.0 m/s airflow speed [6]. - The flexible TS memristor exhibits synaptic behavior, switching between high and low resistance states based on applied voltage, allowing it to generate self-oscillating voltage spikes with adjustable frequencies [8][10]. Group 3: NCTP System Integration - The research team integrated the FISH matrix with a spiking neural network (SNN) to create a complete non-contact tactile perception (NCTP) system, mimicking biological sensory processing mechanisms [13]. - The NCTP system demonstrated over 92% accuracy in recognizing non-contact targets by analyzing airflow patterns and directions through collective encoding from multiple FISH sensors [15]. Group 4: Practical Applications - The NCTP system was tested in a spider robot, which successfully responded to visual and non-contact tactile stimuli in various experimental scenarios, showcasing its ability to enhance environmental perception beyond visual limitations [19][20].

Group 1 - 1X Technologies has launched the home humanoid robot NEO, priced at approximately 140,000 RMB, with a subscription option available [3] - NEO is designed for household use, featuring a height of about 167 cm, weight of 30 kg, and 22 degrees of freedom, powered by NVIDIA Jetson Thor chip [3] - The robot can perform household chores and interact naturally with family members, with a battery life of 4 hours and enhanced hardware reliability [3] Group 2 - Qianlong Intelligent has deployed eight humanoid service robots at the Shangri-La Hotel in Shanghai, marking the first hotel to utilize humanoid robots for service [6] - The robots include various models with specific functions, such as door opening, luggage handling, and food delivery, showcasing a collaborative model of general and specialized robots [6] - This deployment signifies a trend towards the integration of robotic technology in high-end luxury hotels [6] Group 3 - JD Logistics has introduced new members to its "Wolf Pack" robot army, including the Smart Wolf Expansion version and the Tian Wolf omnidirectional vehicle [9] - These robots enhance storage efficiency and reduce labor requirements, with plans for large-scale deployment across national warehouses by November 11 [9] - JD Logistics aims to significantly increase its investment in robots, unmanned vehicles, and drones over the next five years [9] Group 4 - Geek+ has unveiled a new unmanned picking workstation and the industry's first full-process unmanned picking robot solution at the CeMAT Asia logistics exhibition [12] - This innovation addresses the challenges of precise picking with robotic arms and supports multiple picking modes across various industries [12] - The launch is a key milestone in advancing the "general warehousing robot" strategy [12] Group 5 - A thermal moxibustion robot developed by Professor Chen Rixin's team at Jiangxi University of Traditional Chinese Medicine has received approval for market release as a Class II medical device [16] - The robot integrates original thermal moxibustion technology with modern AI, offering 28 effective moxibustion programs and precise dual-point treatment capabilities [16] - This technology enhances treatment efficiency and effectiveness while maintaining a clean operation [16]

机器人大讲堂获悉，智能激光除草机器人研发商「灵动智能」近日成功完成 A轮融资。 本轮融资由科力投资 独家投资完成，其具体融资金额并未对外进行披露。 据机器人大讲堂了解， 灵动智能 （全称：哈尔滨灵动智能装备有限公司） 成立于 2024年9月，是一家专注 于特种机器人及农机装备软硬件部组件研发的企业。 公司依托哈尔滨工业大学机器人技术与系统全国重点实 验室付宜利教授团队的科研成果孵化而来 ，具备扎实的技术背景与创新能力。 在核心团队方面 ， 灵动智能由哈尔滨工业大学机器人研究所副所长付宜利教授领衔。其核心成员来自机器人 技术与系统全国重点实验室，包含 2 名教授与 6 名副教授。团队长期深耕液压 / 电驱仿生机器人系统研制、 机器人液压驱动与动力技术，以及机器人高动态运动控制三大领域。在项目与资金支持上，团队已获得黑龙江 省龙江天使投资基金 500 万元注资，并成功获批 2025 年黑龙江省重点研发计划重大项目 ——"四足轮式机 器人关键技术在无人值守场景下的应用"。 在产品研发与业务拓展方面 ， 灵动智能聚焦农机具智能化赛道，并依托团队在 "AI+机器人"领域的长期技术 积累，形成了"核心产品+技术服务"双驱 ...

10月24日， "2025 加速进化生态大会" 在北京国家速滑馆举办 。限时售价仅 2.99万元的全新Booster K1人 形机器人首次亮相现场。该机器人定位具身开发入门级平台。即便是入门版也搭载了48 TOPS算力芯片，并 且支持二次开发。 加速进化创始人程昊 表示： " 个人计算机作为信息时代的通用硬件母体，已经历从极客玩具到专业工具、再 到生活伴侣的演进 。 而人形机器人正是 AI 时代的个人计算机，未来将成为承载智能应用的物理平台与生态 基石。 因此加速进化的公司愿景是 让人形机器人如个人计算机一样，简单、可靠、实用。 " ▍ 人形机器人价格没有内卷 同样交付端也没有竞争！ 无独有偶，加速进化 Booster K1 人形机器人发布前1天，友商已提前推出定位情感陪伴、高科技玩具的万元 以内人形机器人本体产品，这引发了关于人形机器人是否提前进入价格内卷的讨论。 2024 年底，5 万元以内的人形机器人本体产品数量尚为个位数。截至 2025 年 10 月底，这一价格区间的产 品已超过 10 款，一线本体厂商更将入门级产品价格下探至 4 万元以内。该价格区间内，多数本体厂商难以 通过硬件盈利，行业价值更多向 ...

在产品研发与业务拓展方面 ， 松延动力目前已形成双足人形机器人与仿生人形机器人两大产品系列。其中， 双足人形机器人进一步划分为专注下肢运动能力、追求极限动态性能的 N系列，以及上下肢协同作业的Dora 系列。 公司成立之初即完成了人形机器人 N1初代机的结构设计，并在随后两个月内实现双足稳定行走，成功通过草 地、雪地、柏油路、斜坡与台阶等多种复杂地形的行走测试。 成立仅三个月时，松延动力推出国内首款高自由度、可沉浸式交互的仿生机器人头部。该产品搭载自研多模态 大模型，实现超低延迟人机对话，面部采用超仿生机械结构，支持多自由度表情表达，在标准人头尺寸内实现 技术突破。 2025年初，凭借"人形机器人后空翻"视频广泛传播，松延动力引发公众高度关注。视频中应用的人形机器人 N2被定位为"运动健将型人形机器人"，整机采用铝合金结构与塑料外壳，重量仅30kg，高度1.2米，相当于人 类儿童体型。N2全身共18个自由度（单腿5、单臂4），头部与脚踝roll向未设自由度，最大关节扭矩≥150 机器人大讲堂获悉，人形机器人研发商「松延动力」近日对外官宣成功完成近 3亿元Pre-B轮融资。 本轮融 资由方广资本领投，祥峰投资、 ...