Smooth Action Chunking Via Movement-Primitive Representations for Imitation Learning
Imitation learning has become a widely adopted approach for vision-based robot policies, owing to its conceptual simplicity, scalability, and ability to leverage large-scale datasets.
Explainer · on the wire
The explainer for this story is still on the wire.
The claim, the mechanism, and the figures land here once generated.
Source · arXiv