Active Contact Sensing for Robust Robot-to-Human Object Handover
Robots can now tell a firm grasp from accidental touch with 97.5% accuracy.
Deep Dive
Active Contact Sensing for robust robot-to-human handover uses information-gathering motions and sensing of human-applied forces to distinguish a firm grasp from incidental touch. Tested with 12 participants and 30 rigid objects, the method achieved a 97.5% success rate—over 30% higher than two common baselines. (Authors: Linfeng Li, Lin Shao, David Hsu)
Key Points
- Uses a Bayesian linear model to map robot motions to human-applied forces, distinguishing firm grasp from accidental touch
- Achieved 97.5% success rate with 12 participants and 30 diverse rigid objects
- Over 30% improvement compared to common passive-sensing baselines
Why It Matters
Reliable object handover is critical for home and surgical robots; this method brings them closer to human-level dexterity.