东吴证券：从《秧BOT》到《武BOT》，人形机器人从哪些方面完成“超进化”？

Core Insights - The significant advancement in core capabilities of robots is crucial for their transition from laboratories to factories, marking a key development phase in the robotics industry from 2021 to 2025, with large-scale production expected to begin in 2026 by Tesla and leading domestic companies [1] Investment Recommendations - Recommended investment targets include: - Assembly: Sanhua Intelligent Control, Top Group [1] - Screw Rod: Hengli Hydraulic, Zhejiang Rongtai, New Coordinates, Wuzhou New Spring [1] - Harmonic Reducer: Green Harmonic, Siling Intelligent Drive [1] - Yushu Chain: Shoucheng Holdings, Changsheng Bearing, Meihu Co., Ltd. [1] - Galaxy General Chain: Tianqi Co., Ltd. [1] - Electronic Skin & 3D Vision: Hanwei Technology, Fule New Materials, Obsidian Light [1] Robotics Performance at Spring Festival Gala - Four robotics companies showcased their capabilities at the Spring Festival Gala, including: - Yushu Technology performed a dance titled "Yang BOT" and a martial arts display "Wu BOT" featuring advanced stunts [2] - Magic Atom collaborated with artists in "Creating the Future" and showcased a quadruped robot in a panda design [2] - Songyan Power participated in a skit and demonstrated a humanoid robot's capabilities [2] - Galaxy General demonstrated dexterity in a microfilm with various tasks [2] Technological Advancements - Key advancements in robotics include: - Hardware rigidity and stability have improved significantly, enabling complex movements and stunts [3] - Lightweight materials such as carbon fiber and magnesium-aluminum alloys have reduced robot weight, exemplified by Yushu's reduction from 47 kg to 35 kg [3] - High-explosive joints and high-performance motors have enhanced torque density and reduced motor load through physical energy storage [3] Precision Control - Precision control is essential for robotic performance, requiring real-time calculations of rotation speed, angle adjustments, and landing buffers [4] - Integration of visual sensors, six-dimensional force sensors, and inertial measurement units allows for rapid adjustments to maintain balance [4] Learning and Simulation - Robots utilize reinforcement learning and virtual simulation for complex choreography and human-robot interaction [5] - Pre-training in virtual environments has been crucial for mastering difficult maneuvers, with extensive training conducted prior to performances [5] Dexterity Improvements - Significant progress in dexterous manipulation has been observed, with robots successfully performing tasks requiring fine motor skills [5] - The integration of tactile sensors allows for real-time adjustments based on changes in weight and balance during tasks [5]