Figure recently unveiled its new figure 3 humanoid robot. We dig into all the latest updates and announcements to understand how Figure has optimized its newest creation for the home, the warehouse, and beyond. Before we get into the new demos, the obvious question.Figure CEO Brett Adcock has said publicly nothing in this video is teleyoperated. But that isn't quite the same as saying it's fully autonomous either. Autonomy is more like a spectrum with complete human control at one end and pure robotic plann ...

在目前的视触觉传感器领域，VBTS的力估计精度受限于相机接收接触表面三维空间形变信息的缺失。有研究 者使用双层标记点来提升触觉标记点密度，到使用颜色叠加方式改进触觉传感器的传感原理，但双层标记生产 和制造难度大，且视觉传感原理的固有特性——光学形变特征与接触力间的非线性映射易受环境扰动影响，利 用双层标记的方式无疑是降低了接触的稳定性。此外，还有研究利用使用双目相机（双目系统）和结构上的改 进，接收更丰富信息形成来提升力估计精度，但不利于集成。具体而言，现有方案面临双重困境：（1）复杂 光学结构与传感器微型化需求存在根本性冲突；(2)传感原理改进引起的工艺复杂化导致传感器泛化能力不 足。 IROS 2025论文提出了 视触觉结合磁触觉的MagicGel传感器。 通过引入磁传感模态构建视觉-磁场异构数据 融合框架。其核心在于利用磁场信息补偿视觉图像的信息缺失，通过视、磁关联建立更完备的接触力学表征体 系。 MagicGel 的主体结构如 图 1 所示。其结构主要分为：涂层、强磁颗粒标记点、弹性体、灯带、霍尔传感器 和相机。此外还有接收视觉图像和磁场信息的接收模块。 MagicGel 的整体尺寸为 31*31*2 ...

Core Insights - Tesla's CEO Elon Musk claims that the Optimus robot could revolutionize productivity for businesses, referring to it as an "infinite money glitch" due to its potential efficiency [2] - Despite the excitement surrounding Optimus, Tesla's electric vehicle (EV) sales have been sluggish, with a 13% year-over-year decline in the first half of 2025 [9][10] - The company is facing increasing competition in the EV market, particularly from cheaper brands like BYD, leading to a loss of market share [11] Tesla's Future Product Platforms - The Optimus robot is seen as Tesla's most complex manufacturing challenge, requiring in-house production of components due to the lack of an existing supply chain [5] - Musk predicts that humanoid robots could outnumber humans by 2040, with Optimus potentially being five times more productive than a human worker [6] - Long-term revenue from Optimus sales could reach $10 trillion, with plans to scale production from 1 million units annually to as many as 100 million units in the future [7] Current Business Performance - Over 75% of Tesla's revenue still comes from EV sales, which have seen a decline, delivering 720,803 cars in the first half of 2025 [9] - Although there was a 7% increase in deliveries in the recent third quarter, this may have been influenced by consumers purchasing vehicles ahead of the expiration of a tax credit [10] - Tesla's recent launch of a low-cost Model Y aims to revive its passenger EV business, although Musk has previously resisted this strategy in favor of focusing on the Cybercab autonomous robotaxi [13] Competitive Landscape - Tesla's market share in China has decreased from 11.7% to 7.5% in the first half of 2025, highlighting the competitive pressures from other EV manufacturers [11] - The company is testing a ride-hailing service using its EVs with full self-driving software, but current operations require a human supervisor, adding costs [14] Valuation Considerations - Tesla's market capitalization is currently $1.3 trillion, with a price-to-earnings (P/E) ratio of 254, making it significantly more expensive than the Nasdaq-100 index and Nvidia [15][16] - Given the current state of the EV business and the timeline for Optimus production, there are questions about the wisdom of investing at such a high valuation [18]

Core Insights - The rapid evolution of robotic movement capabilities is highlighted, but understanding complex tasks remains challenging for robots [1] - The introduction of the "WoW" embodied world model by a Chinese research team represents a significant advancement in robotic intelligence [1] Group 1: Technological Advancements - The "WoW" embodied world model allows robots to simulate human-like thinking and decision-making, generating future prediction videos that align with physical laws [5] - The model enables robots to connect imagined movements with real-world execution, enhancing their interaction with the environment [5] Group 2: Data Collection and Training - The research team has collected millions of real interaction data points to ensure the world model can operate effectively in diverse real-world scenarios [8] - The model is designed to adapt to various types of robots, including humanoid and robotic arms, and can be applied across multiple settings such as homes, supermarkets, and logistics [10] Group 3: Open Access and Applications - The "WoW" model is open to global researchers and developers, facilitating broader applications and innovations in robotics [10] - It can accurately simulate extreme scenarios, such as water spilling on a computer, providing crucial data for training that is difficult to obtain through real-world testing [10]

Core Insights - The article discusses the evolution and current state of embodied intelligence, focusing on the roles of the brain and cerebellum in robotics, where the brain handles perception and planning, while the cerebellum is responsible for execution [3][10]. Technical Evolution - The development of embodied intelligence has progressed through several stages, starting from grasp pose detection, moving to behavior cloning, and now advancing to diffusion policy and VLA models [7][10]. - The first stage focused on static object grasping with limited decision-making capabilities [7]. - The second stage introduced behavior cloning, allowing robots to learn from expert demonstrations but faced challenges in generalization and error accumulation [8]. - The third stage, marked by the introduction of diffusion policy, improved stability and generalization by modeling action sequences [8]. - The fourth stage, emerging in 2025, explores the integration of VLA models with reinforcement learning and world models to enhance robots' predictive and interactive capabilities [9][10]. Current Trends and Applications - The integration of VLA with reinforcement learning enhances robots' trial-and-error learning and self-improvement abilities, while the combination with world models allows for future prediction and better planning [10]. - The article highlights the growing demand for embodied intelligence applications across various sectors, including industrial, home, restaurant, and medical rehabilitation, leading to increased job opportunities and research interest in the field [10]. Educational Initiatives - The article outlines a structured learning program aimed at equipping individuals with comprehensive knowledge of embodied intelligence algorithms, including practical applications and real-world projects [11][14]. - The course targets individuals with a foundational understanding of embodied intelligence and aims to bridge the gap between theoretical knowledge and practical deployment [18][24].

Core Insights - Shenzhen Leju Robotics has completed nearly 1.5 billion yuan in Pre-IPO financing, marking one of the largest single financings in the humanoid robot sector this year [1] - The company focuses on two product categories: small bipedal robots for the education market and full-sized humanoid robots for industrial and research applications [1] - The industry is witnessing emerging opportunities in collaborative innovation within the supply chain and new technology breakthroughs, such as 5G-A remote control technology and solid-state battery technology [1] Company Overview - The founder of Leju Robotics, Cold Xiaokun, born in 1992, has a strong background in robotics, having won national awards during his school years and later leading a team to establish the company in Shenzhen [2] - The company initially faced challenges in the humanoid robot industry, including high costs of prototype production and reliance on imported core components [2][4] - Leju Robotics adopted a strategy of starting with small robots to generate revenue before advancing to full-sized humanoid robots, which has been crucial for overcoming industry hurdles [2] Product Development - In 2016, Leju Robotics received 10 million yuan in angel investment, leading to the mass production of its first small bipedal robot, "Aelos" [3] - By 2017, the company secured a strategic investment of 50 million yuan from Tencent, which facilitated the upgrade of its educational robots [4] - The full-sized humanoid robot "Kua Fu" was launched in 2023, showcasing advanced capabilities such as jumping and traversing various terrains [5] Market Dynamics - The humanoid robot market is on the verge of explosive growth, with projections indicating that sales in China will exceed 4,000 units in the first half of 2025, reaching 7,300 units for the entire year [6] - Despite the growth potential, the industry faces challenges such as high costs of core components, insufficient scene adaptability, and production capacity issues [6][7] - Leju Robotics stands out by successfully scaling both educational and industrial humanoid robots, with a competitive landscape showing a CR3 of 79.63% by mid-2025 [7] Technological Advancements - The integration of multiple technologies is driving the evolution of humanoid robots, with a focus on combining Model-Based and Reinforcement Learning algorithms for improved adaptability [7] - The "Kua Fu" robot has achieved a terrain adaptation error rate of 3.2% through these technological advancements [7] - Collaborations with companies like CATL aim to enhance battery technology, with plans for the next generation of "Kua Fu" robots to utilize solid-state batteries for improved performance [8] Future Opportunities - Future opportunities in the humanoid robot industry are concentrated in three areas: vertical market penetration, global expansion, and collaborative innovation within the supply chain [9] - The demand for caregiving robots is expected to grow significantly due to the aging population in China, with a projected market size exceeding 20 billion yuan by 2028 [9] - Leju Robotics is conducting research in Southeast Asia, planning to launch a simplified version of its industrial robots by 2026 to meet local market needs [9]

Core Concept - The article presents a novel multi-modal bionic robotic hand that combines magnetic silicone skin with a rigid skeleton, integrating a vacuum suction cup at the fingertips to achieve three gripping modes: wrapping, parallel, and adsorption [2]. Group 1: Design and Mechanism - The bionic fingers utilize a rigid skeletal structure wrapped in soft muscle tissue, providing excellent cushioning and shock absorption for grasping flexible objects [2]. - The internal Hall sensor units detect the deformation of the magnetic silicone film, allowing real-time monitoring of the gripping information [4]. - The silicone film, mixed with magnetic powder, transforms non-magnetic particles into magnetic ones upon unidirectional magnetization, creating a regular magnetic field that changes with deformation during contact with external objects [5]. Group 2: Gripping Phases and Data Collection - The gripping process is divided into four stages: approaching the object, squeezing the object, lifting the object, and releasing the object, with distinct data trends observed in the first two stages [7]. - The first stage shows a lower rate of change in magnetic data compared to the second stage, indicating different interactions during the approach and squeeze phases [7]. - A control system collects magnetic field data every 0.25 seconds to establish a relationship between the change in magnetic field (∆B) and the diameter of the object being grasped [9]. Group 3: Experimental Setup - An experimental platform was established using an RM65 robotic arm and a D455 depth camera to test the robotic hand's grasping capabilities in real-world scenarios, enabling autonomous grasping operations [10].

Core Viewpoint - The article discusses the advancements in humanoid robots, particularly focusing on the latest H2 humanoid robot by Yushu, which features enhanced perception capabilities through a new "robot eye" system that integrates dToF LiDAR technology for improved environmental understanding and interaction [1][15]. Group 1: Humanoid Robot Advancements - Yushu's H2 humanoid robot showcases high dynamic capabilities, including the ability to perform complex movements like dancing while maintaining 360-degree environmental awareness through its advanced head and eye system [1]. - The integration of dToF LiDAR technology allows for precise 3D spatial measurements, overcoming limitations faced by traditional visual sensors, especially in varying lighting conditions [4][6]. Group 2: Importance of Perception in Robotics - The evolution of robotic perception is crucial as humanoid robots transition into unstructured environments, necessitating rapid understanding and interaction with their surroundings [3][12]. - Traditional visual systems struggle with depth estimation and synchronization, leading to challenges in complex task execution, which dToF LiDAR effectively addresses by providing accurate and reliable depth information [4][7]. Group 3: Technological Integration - The article emphasizes the need for hardware-level synchronization among various sensors (dToF LiDAR, RGB cameras, IMUs) to ensure accurate data collection and processing, which is essential for real-time decision-making in dynamic environments [7][9]. - The AC2 sensor system, developed by Sutech, exemplifies this integration, achieving a synchronization precision of 1ms, significantly enhancing the reliability of robotic perception [8][11]. Group 4: Future of Humanoid Robotics - The future of humanoid robots lies in their ability to achieve high-precision, high-consistency perception across various environments, enabling them to perform complex tasks and interact effectively with humans and objects [12][16]. - As humanoid robots evolve from mere tools to autonomous intelligent agents, the development of advanced perception systems will be pivotal in facilitating their integration into everyday life and various industries [15][16].

Core Insights - The incident at the "GEEKCON2025" highlighted significant security vulnerabilities in humanoid robots, as hackers demonstrated the ability to take control of a robot and use it to attack another robot, raising concerns about the safety of such technologies [1][3][4]. Group 1: Security Vulnerabilities - Two white hat hackers successfully hacked into a humanoid robot from Yushutech, demonstrating a critical security flaw that allowed them to execute attack codes remotely [1][4]. - The hackers exploited a logic flaw in the robot's built-in model, enabling them to gain arbitrary code execution permissions on the targeted robot [5]. - A previous report from security researchers indicated a high-risk vulnerability in Yushutech's low-power Bluetooth and Wi-Fi configuration interface, which could allow controlled robots to attack others within Bluetooth range [5]. Group 2: Industry Response and Awareness - Yushutech acknowledged the vulnerabilities and stated that they had already begun addressing these issues, with most repairs completed by the time of the event [5]. - The lack of focus on security in the robotics industry was noted, with many manufacturers prioritizing functionality over security measures, leading to widespread vulnerabilities [7]. - The chairman of the GEEKCON organizing committee emphasized the dual challenges of talent loss in cybersecurity and budget constraints faced by companies, which exacerbate security threats in emerging technologies [7]. Group 3: Recommendations for Manufacturers - Security experts recommend that manufacturers incorporate security checks during the development process to identify and eliminate superficial issues [8]. - Companies should consider building their own security systems or collaborating with external security teams to conduct offensive and defensive testing, which is essential for assessing the reliability of robotic products [8]. - The importance of security in robotics was underscored, with a call for the industry to recognize its value to build trust with users and establish themselves as reliable market leaders [7][8].

Core Insights - The article discusses the integration of algorithms and bionic interaction technology in the development of intelligent robots, enabling them to replicate subtle expressions like smiling and winking, as well as perform dynamic actions such as flips, running, and long jumps [2] Group 1 - Intelligent robots are being designed to serve various scenarios including home services, outdoor companionship, and educational research [2] - The fusion of advanced technology is aimed at accelerating the "awakening of the soul" in robots, enhancing their interaction capabilities [2]