MIT wristband drives robotic hand live

This feels less like a flashy demo and more like a real interface primitive: camera-free hand tracking that is expressive enough for teleoperation, prosthetics, and spatial computing.

• –The core advantage is sensing through occlusion and clutter, since ultrasound reads the wrist directly instead of guessing from external video.
• –The paper’s live robotic-hand demos matter because low-latency motion capture only becomes useful when it can close the loop in real time.
• –The biggest product hurdle is generalization; the system still relies on per-user training, so it is not yet plug-and-play for broad deployment.
• –For robotics, the more important output may be data: a wearable that can log dexterous human motion could become a training stream for humanoid hands.
• –If it matures, this could pressure both glove-based trackers and EMG interfaces in XR and assistive control.