A force sensor on a prosthetic or robotic hand, encoded into vibration on the user's upper arm. Closing the loop is the frontier in prosthetics, and the University of Utah's NeuroRobotics work is the local reference.
A myoelectricMyoelectric prosthesis An artificial limb controlled by EMG from the remaining muscles. Related: Electromyography (EMG)Electromyography (EMG) Recording the electrical activity of muscles, hundreds of microvolts to millivolts, much larger than EEG.. Glossary entry hand opens and closes on command, and the user watches it to know what it is doing, because it tells them nothing. Every grasp is done by eye. Add a force sensor to the fingertip and a vibration motor on the upper arm whose pattern follows the force, and the user can grasp without looking, feel the egg before it cracks, and, after weeks, start to feel the hand as theirs. HapticHaptic feedback Information delivered through touch, such as vibration on the skin that tells a prosthesis user how hard they are gripping. Glossary entry feedback is the cheapest way to close the loop, the encoding is the whole design problem, and the local lab that works on the invasive version is the one to show your results to.
Predict before you look
You map grip force linearly to vibration amplitude on the upper arm. The user says it 'buzzes' and they ignore it within a day. What went wrong?
The encoding. Skin receptors fire a burst when contact is made, a burst when it is released, and only a weak sustained response between. Biomimetic encoding, which mimics that pattern (strong pulses at contact and release, gentle modulation during hold), feels like an event rather than a buzz, is not habituated to, and improves task performance more than proportional encoding in published comparisons. The design problem is not the motor; it is the code.
Hardware
a loop, closed at both ends
EMG on the forearmthe user commands open or close
→
The handa myoelectric or a printed robotic hand
→
Force sensoran FSR or a load cell on a fingertip
→
Encodingbursts at contact and release, not proportional buzz
→
Skin on the upper armwhere the user feels it
Test it blindfolded. If the feedback only helps when the user can also see the hand, you have measured attention rather than feedback.
Figure 1. The loop, closed. The bottom half of this chain is the whole project: without it, the user's eyes are the only feedback path, and looking away means dropping the cup.
The hand. A commercial myoelectric hand if you are working with a user and their prosthetist. Otherwise a 3D-printed hand (the e-NABLE or OpenBionics designs) driven by a servo, controlled by your Phase 1 EMGElectromyography (EMG) Recording the electrical activity of muscles, hundreds of microvolts to millivolts, much larger than EEG. Related: Myoelectric prosthesisMyoelectric prosthesis An artificial limb controlled by EMG from the remaining muscles. , Electroencephalography (EEG)Electroencephalography (EEG) Recording the brain's electrical activity from electrodes on the scalp, a few tens of microvolts of blurred, summed cortical activity. . Glossary entry, so that you have a full loop to test on yourself.
The sensor. A force-sensitive resistor or a thin-film load cell on the thumb and index fingertip. Sample at 200 Hz; force events are fast.
The stimulator. Two or three coin or linear-resonant vibration motors on a band on the upper arm or shoulder (where sensation is intact and the site is out of the way), driven by a motor driver from the same microcontroller.
Encoding
Implement three and compare. Proportional: amplitude follows force. Discrete: three levels, light, medium, hard. Biomimetic: a strong 100 ms pulse at contact onset (force rising past a threshold), a strong pulse at release, and a low-amplitude modulation proportional to force in between, with the onset pulse’s intensity scaled by how fast the force rose. Add a second motor for a second finger, or for slip (a rapid drop in force while grasping).
Testing on yourself
Blindfolded, with the printed hand: grasp objects of different stiffness and identify them (sponge, apple, wooden block). Pick up a paper cup without crushing it, with and without feedback. Count crushed cups. Detect when an object is being pulled from the grasp (slip) and re-grip. Score each encoding over twenty trials. Then wear the system doing ordinary tasks for an hour a day for a week and log whether it starts to feel like the hand rather than the arm.
Testing with a user
Through a prosthetist or the NeuroRobotics lab. The same tasks, plus the ones the user names. Embodiment questionnaires exist; use one. Report task scores by encoding, and what the user said.
Build the hand, sensor, and stimulator, and confirm the force trace and the motor respond.
Implement the three encodings behind a single switch.
Blindfolded object-stiffness and cup tests on yourself, twenty trials per encoding. Score.
Wear it an hour a day for a week. Log.
If possible, a session with a user through their prosthetist. Report.
Write it up as a comparison of encodings with the published literature alongside.
Stop here if you like. A closed loop from fingertip to arm with three encodings compared on real tasks. That is a result the local lab will want to hear about.
Recall
What is biomimetic haptic encoding and why does it beat proportional encoding?
Strong pulses at contact and release with weak sustained modulation in between, mimicking how skin receptors fire. It feels like a touch event, resists habituation, and improves task speed and grasp control more than continuous proportional vibration.
Recall
Why put the stimulator on the upper arm or shoulder?
Sensation is intact there, it is out of the way of the socket and the EMG electrodes, and it is close enough to the hand for the brain to map the feedback onto it with practice.