Event2Reward: Learning Robotic Rewards from Progress Relations
Learning task progress from robot demonstrations provides a practical way to construct dense shaping rewards for robotic reinforcement learning.
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