Tsinghua, Infinigence AI and AgiBot propose Collective Experience Learning for robots
Overview
Researchers from Tsinghua University, Infinigence AI and AgiBot propose Collective Experience Learning (CEL), a training method in which robots learn from other robots' successful experiences rather than mainly from human demonstrations.
QbitAI reports that in tests with 96 real robots, the success rate on a phone-packaging task rose from 79.2% to 98.3%, and rose from 86.7% to 98.3% as robot count grew from 16 to 96.
Pandaily reports a 97.5% success rate across five tasks on the 96-robot cluster, above the average of six human operators. Pandaily says the results come from the research team and have not been independently replicated.
Written by AI from the articles below · updated Oct 30, 3:43 AM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- PandailyNewsTsinghua, Infinigence AI and AgiBot train 96 real robots to learn from each other's experience
AIResearchers from Tsinghua University, Infinigence AI and AgiBot propose Collective Experience Learning, in which robots learn from other robots' successful experiences. On a cluster of 96 real robots, the approach reached a 97.5 percent success rate across five tasks, higher than the average of six human operators. The results are reported by the research team and have not been independently replicated.
- QbitAINewsTsinghua, Infinite Synergy and AgiBot propose collective experience learning for embodied AI
AITsinghua University, Infinite Synergy and AgiBot propose Collective Experience Learning (CEL), in which robots learn from successful experiences of other robots instead of relying mainly on human demonstrations. In tests with 96 real robots, the robots' success rate on a phone-packaging task rose from 79.2% to 98.3%, and the success rate rose from 86.7% to 98.3% as robot count grew from 16 to 96.
Heat Trend
Not enough continuous observations to show a trend yet.