MIT researchers have created an ultrasound wristband that captures detailed hand motion data to train robots for dexterous tasks. - The wristband uses high-frequency sound waves to see through skin and relay muscle and tendon movements.
- An AI algorithm decodes the images into 22 degrees of freedom, matching the human hand's full range of motion. - In tests with eight volunteers, the device mirrored hand gestures, including all 26 ASL letters, within 120 milliseconds.
- The wireless system can operate remotely, and the team aims to build large datasets for robots to learn without human guidance. This innovation could revolutionize how humanoid robots perform tasks like housework and surgery.