A decade of robot skill learning: from mass imitation to few-shot RL

stepjamUK · x · 2026-09-02

In an Automate Further talk, the author traces a decade of robot skill learning: early imitation learning requiring hundreds of thousands of grasps, through sim-to-real transfer, to RL agents that improve from just a handful of demonstrations. He argues the stack is mature and deployable on real-world problems today.

Original post →

More from Embodied

Embodied channel →