Learning contact-rich whole-body manipulation with example-guided reinforcement learning

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Example-guided reinforcement learning improves whole-body manipulation in humanoid robots, highlighting compliance's role.

The agent effectively manipulates everyday objects like water jugs and boxes using only one example motion.
Training utilized simulation for whole-body skills, showcasing transferability thanks to domain randomization.
The method relies solely on tactile feedback without object pose tracking, enhancing dexterous manipulation capabilities.
Compliance is critical for stability during manipulation, emphasizing its importance in robotics applications.

Barreiros et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark