Robotics & manufacturingResearch findings
Dexterous Manipulation via Supervised Play
Original published 2026.10.16 02:55 KSTAdded to Blackdeck 2026.10.19 00:16 KST
What has changed?
A robot learns a play policy from human demonstrations and autonomously collects interaction data to train a world model.
Key points
- A robot learns dexterous manipulation policies from human play demonstrations.
- It autonomously collects interaction data to train a world model without task-specific demonstrations.
- The policy guides model-predictive control for real-world manipulation tasks.
Connect it to your work
The learned world model enables model-based planning through policy-guided MPC and self-play distillation.
What remains unproven
This research note is based on an abstract; check the original for test conditions and application scope.
Originals used for this note
Links open the original publisher in a new tab.