Robots Learn from Each Other: Collective Experience Learning for Scalable Embodied Intelligence
当机器人开始向同伴学习：96台真机探索具身智能的规模化成长
A research paper proposes Collective Experience Learning (CEL), a framework where robots improve their task execution by learning from each other's experiences. Using 96 robots, the study demonstrates that shared data significantly enhances task success rates, with failures reduced by 92% in some cases. The method employs failure-guided experience retrieval (FERT) and a fault-tolerant asynchronous training system (FATS) to enable scalable, real-world robot learning.
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
98.3%
View original evidence
CEL 论文封面 在手机包装为代表的多阶段、长流程工业场景中，训练后的策略在两台未参与训练的同型号机器人上完成 120 次测试，成功率从 79.2% 提升至 98.3%。
Open source S9Reported success rate
86.7%
View original evidence
更值得关注的是规模带来的收益：在相同物理训练时间内，参与学习的机器人从 16 台增加至 96 台，测试成功率由 86.7% 提高至 98.3%。
Open source S11Reported success rate
98.3%
View original evidence
更值得关注的是规模带来的收益：在相同物理训练时间内，参与学习的机器人从 16 台增加至 96 台，测试成功率由 86.7% 提高至 98.3%。
Open source S11
- paper: not_reported
- code: 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.
CEL 论文封面 在手机包装为代表的多阶段、长流程工业场景中，训练后的策略在两台未参与训练的同型号机器人上完成 120 次测试，成功率从 79.2% 提升至 98.3%。
Open source S9
换成更直观的数字：同样 120 次测试，失败从 25 次降至 2 次，减少了 92%。
Open source S10
更值得关注的是规模带来的收益：在相同物理训练时间内，参与学习的机器人从 16 台增加至 96 台，测试成功率由 86.7% 提高至 98.3%。
Open source S11
Source：Robotics — Chinese web discovery · sohu.com