Zhiyuan and Partners Launch Collective Experience Learning for 96-Robot Fleet

As the seventh installment of AI WAVE, Zhiyuan joined forces with Tsinghua University and Infinigence AI to release Collective Experience Learning (CEL). The paradigm supports Zhiyuan's operational-intelligence push to put robots on real jobs through a full-stack closed loop from data to intelligence. It connects experience acquisition, experience utilization and large-scale online learning into a single chain.

The core breakthrough is treating robot peers as experience sources for one another. In real operations, every success and failure becomes shared learning material for the collective instead of staying isolated on one machine. This turns fleet activity into a continuously growing data resource.

Using this method, 96 real robots achieved autonomous cluster evolution, demonstrating how expanded robot populations can improve task-execution reliability. The partners say the approach helps embodied intelligence escape a scaling bottleneck in which expert demonstrations cannot increase in sync with deployment.

CEL is positioned as a new route for autonomous growth: rather than relying only on human-provided demonstrations, robots can learn from peers' real-world experience while the fleet scales. The announcement is part of Zhiyuan's broader effort to build a complete loop from data collection to intelligent operation, with collective learning intended to make each deployed robot a potential teacher for others. The company says this can improve reliability as more machines join.