Genie Master: A Framework for Embodied Agents to Evolve in Real-World Scenarios
机器人实现“边上岗边进化” 智元发布具身Agent框架Genie Master
Intelligence Yuan has released the Genie Master framework, which allows robots to continuously self-improve in real-world settings without human intervention. The framework enables robots to identify failures, adjust strategies, and accumulate experience, supporting tasks like grocery restocking and food packaging in real stores.
The release of Genie Master represents a significant step toward real-world deployment of embodied agents, as it enables robots to self-improve through continuous interaction with complex environments without human intervention.
Source: Robotics — Chinese web discovery · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
据悉，这是行业首次跳出仿真环境、走出实验室，首次在真实业务场景中验证 Physical RSI（具身递归自进化）。
Open source S8
搭载该框架的精灵 G2，在真实便利店上岗期间，不依赖人工示范与调试，自己定位失败原因、调整策略、沉淀经验，做到了在多个复杂精细任务上持续改进，实现“ 无人在环 ”的持续自我进化。
Open source S7
于是，Agent 调整执行引导：左手先靠近目标，用腕部相机补充近距离视角；确认挂钩位置后，再由右手完成挂货。
Open source S17
《中国经营报》记者了解到，有了这套协议，Agent 结合现场观测下达执行引导，机器人执行后，成功与否、卡在哪一步等信息实时返回；一旦任务中断，系统保留已完成的进展，Agent 从断点处重新规划，而不是从头再来。
Open source S14
在上述挂货任务中，Agent 回溯失败记录后发现，头部视角分辨率不足，挂钩又被标签牌遮挡，导致机器人无法识别挂钩位置。
Open source S16
在Agent协助执行、接管和恢复的过程中，系统会产生包含问题、调整和结果的数据。
Open source S22
Source：Robotics — Chinese web discovery · news.qq.com