Tsinghua University, Wuwen Xinqiong, and Agibot Propose Collective Experience Learning Paradigm for Embodied Intelligence
机器人互相“取经学习”！产学研联合提出具身智能规模化新方案
Tsinghua University, in collaboration with Wuwen Xinqiong and Agibot, has proposed a new paradigm called Collective Experience Learning (CEL) for embodied intelligence. Validated on a group of 96 humanoid robots, CEL allows machines to share experiences rather than relying solely on human teleoperation. In tests involving five tasks such as flower arranging and phone packaging, the robot group achieved an overall success rate of 97.5%, exceeding the average performance of six human operators.
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.
Reported numbers
Reported success rate
97.5%
Reported trials
Reported success rate
98.3%
Reported trials
- paper: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
近日，清华大学联合无问芯穹、智元推出群体经验学习（CEL）新范式，并且在96台真实人形机器人群体上完成验证。
Open source S2
经过群体学习训练，机器人群体整体任务成功率达到97.5%，五项任务的表现全部超过6名人类操作员的平均水平。
Open source S9
其中流程复杂的手机包装任务，成功率从79.2%提升至98.3%。
Open source S10
系统层面搭建容错异步训练系统FATS，即便出现网络波动、部分机器人硬件故障，近百台机器仍可持续协同训练，硬件故障带来的时间损耗仅占总运行时长1.3%。
Open source S13
实验还证实规模化效应：参与训练机器人由16台扩大到96台，整体能力随之进一步提升。
Open source S11
Source：Robotics — Chinese web discovery · msn.cn