AI & softwareResearch findings
Achieving Safety and Performance Balance in Dexterous Hand Manipulation
Original published 2026.10.28 02:58 KSTAdded to Blackdeck 2026.11.02 12:50 KST
What has changed?
Diverse goals and resets with adaptive friction curriculum enable precise in-hand manipulation policy training
Key points
- Policies trained in simulation achieve high precision in in-hand pose reaching
- Safety filter reduces object drops by 75% while maintaining 91.5% goal success
- Successfully reaches goals across diverse orientations and wrist positions
Connect it to your work
Achieves 95.2% success rate in 4,800 unseen goals across 4 objects with 75% drop reduction via safety filter
What remains unproven
The abstract reports simulation results; real-world performance is not established by this input.
Originals used for this note
Links open the original publisher in a new tab.