MIT researchers channel AI to turn hand gestures into robot training data
MIT researchers channel AI to turn hand gestures into robot training data
Publish Date: 2026-06-12 08:59:00
Source Domain: www.wcax.com
- Researchers at MIT have developed an ultrasound wristband to capture detailed data of human hand movements.
- The tool uses high-frequency sound waves to relay images of muscles and tendons to a computer, which in turn trains robots to mimic these gestures.
- The wristband aims to help humanoid robots achieve the dexterity required to perform tasks like grasping and handling, which have been notoriously difficult for machines.
- The captured data will also assist robots in performing delicate tasks such as surgery, requiring precision and flexibility in hand movements.
- The wristband can operate wirelessly, indicating that it can control robots even when not in direct proximity to them.
- The system has demonstrated precise replication of hand gestures, including all 26 letters of American Sign Language, within 120 milliseconds.
- The researchers envision using large datasets from this wristband to eventually train robots in dexterous tasks without constant human intervention.
