← The Vault
Everyday AI

Teaching robots to mirror human tendons

Robotic hands have long struggled to replicate human dexterity because they lack the complex internal feedback of our own anatomy. By using ultrasound sensors to track muscle activity beneath the skin, researchers are bridging this gap, allowing machines to move with a new level of precision. This development moves robotics away from bulky, external sensors and toward a more biological approach to control, offering a glimpse into how machines might soon handle delicate tasks with human-like grace.

Edition № 089Room: Everyday AI23 June 20261 min readSources: 1
Article

Human hands rely on a complex internal system of 34 muscles and over 100 tendons to perform delicate tasks. Most robotic hands fail to match this dexterity because software designers struggle to map that complex, internal mechanical coordination into digital commands.

Researchers at MIT are now using ultrasound imaging to track the actual movement of muscles underneath the skin of a human forearm. By feeding this real-time data into a robot, the machine can mirror the wearer's hand gestures and fine motor skills without requiring external sensors on the fingers themselves.

Moving past external sensors

The system works by training a machine learning model to associate specific ultrasound images of moving muscle fibers with the resulting hand shapes. Think of it like a translator; instead of reading the final motion of your fingertip, the system monitors the internal "cables"—your tendons and muscles—that pull to make the motion happen. The robot processes these internal signals to calculate the corresponding position, effectively bypassing the need for physical sensors on the hand.

This approach shifts the burden of dexterity from external hardware to better data interpretation. It is particularly useful for prosthetic limb design, where comfort and natural movement are primary constraints for the user. We are moving toward a future where robots coordinate with us by reading our internal biological signals rather than waiting for us to press buttons or tap screens.

Sources
← PreviousThe physical limits of AI computeNext →Marketing shifts to a one-to-one automated strategy
Tomorrow's edition · free

Liked this one? The next lands at breakfast.

Every story in tomorrow's AI news, rebuilt in plain English — five minutes, sources linked, free forever.

By joining you agree to receive Article's daily newsletter — unsubscribe in one click. Privacy

← Back to the Vault