TextOp通用小脑
Search documents
让机器人「秒懂人话」!中国电信TeleAI发布首个实时文本驱动人形机器人控制框架TextOp
机器之心· 2026-02-12 03:00
Core Viewpoint - The article discusses the revolutionary advancements in humanoid robotics, particularly focusing on the TextOp system developed by China Telecom's TeleAI team, which allows for real-time control of robots through natural language commands, moving away from pre-programmed actions to a more interactive and fluid user experience [2][10][36]. Group 1: Technological Innovations - TextOp introduces a streaming text command system that enables users to issue real-time instructions, allowing robots to seamlessly switch actions without pre-programming or remote control [2][10]. - The system employs a dual-layer architecture that separates high-level intention planning from low-level motion execution, ensuring that robots can maintain balance and coherence even when commands are modified mid-action [15][17]. - TextOp utilizes a unique motion representation tailored for robotic structures, ensuring that generated movements are both human-like and physically feasible [18][21]. Group 2: Performance Metrics - The research team conducted extensive real-world testing on the Unitree G1 humanoid robot, demonstrating its ability to perform a variety of skills, including dance and martial arts, with impressive fluidity and responsiveness [22][25]. - Quantitative evaluations showed that TextOp achieved a high success rate in action execution and trajectory quality, with user interaction latency measured at just 0.73 seconds, establishing a strong foundation for real-time applications [34]. Group 3: Application Prospects - TextOp has the potential to redefine human-robot interaction across various scenarios, such as live performance adjustments, sports training demonstrations, and intelligent remote operation in hazardous environments [36][37]. - The system addresses the long-standing challenge of mapping natural language to physical actions, paving the way for fully autonomous and versatile embodied intelligent robots [37][38].