WSU Researchers 3D Print Electronic Skin That Gives Prosthetics a Sense of Touch

Prosthetics That Cannot Feel

Prosthetic hands have gotten remarkably good at gripping and manipulating objects. What they still cannot do is feel them. A person using a myoelectric hand has to watch every interaction to judge how hard to squeeze or whether an object is slipping. That constant visual monitoring is exhausting and limits what the hand can realistically do in daily life.

Researchers at Washington State University have now built a customizable electronic skin that could close that gap. Their system combines high-resolution pressure sensing with temperature detection, and it is manufactured using 3D printing and laser cutting rather than clean-room semiconductor processes. The result is a sensing layer that can be shaped to fit any prosthetic, not just a flat test bench.

Why Existing E-Skin Falls Short

Most electronic skins that reach the research literature are built on rigid substrates in expensive fabrication facilities. They offer impressive sensitivity but only over small flat areas. Adapting them to the curved geometry of a prosthetic finger or palm means compromising on coverage or sensor density. Some systems leave the sides and backs of fingers completely unwired, which is precisely where slip detection matters most.

The WSU team took a different route. They start by scanning the target prosthetic with a structured-light 3D scanner, then use software to map sensor layouts directly onto the scanned geometry. Custom structural layers are produced with SLA 3D printing. Flexible electrical layers are laser-cut and assembled into sensor modules that snap together without adhesives. A damaged module can be swapped out individually, which is a practical consideration that most lab prototypes ignore.

How the Sensing Works

Each module is a multilayer sandwich less than two millimetres thick. Inside are pressure-sensing matrices made from piezoresistive film alongside miniature thermistor temperature sensors. The stacked layout lets both sensing systems occupy almost the same physical area without interfering with each other. That is what enables dense multimodal feedback across curved prosthetic surfaces.

A prototype finger and palm assembly incorporates 170 pressure-sensing pixels and 14 temperature sensors. That works out to roughly six pressure sensors per square centimetre, which is about ten times the spatial resolution of commercial glove-based sensors. The modular approach also means the layout can be tuned to the individual prosthetic rather than forcing the prosthetic to match a standard sensor array.

Neural Network Cleanup

Raw sensor data from a system this dense is noisy. Manufacturing variations mean every pixel behaves slightly differently, and without calibration the pressure readings drift by tens of kilopascals. The team trained a neural network to calibrate individual sensing elements and remove manufacturing noise. The network performs pixel identification and pressure estimation simultaneously, cutting average pressure estimation error from about 30 kPa down to 8.5 kPa. Classification accuracy exceeds 95 percent.

The system does more than measure static force. As a finger slides across a surface, the sensors pick up micro-vibrations that reveal texture. Temperature sensing adds another layer of material discrimination based on how quickly heat flows between the sensor and whatever it touches. Together those inputs mimic the multimodal sensing that human skin performs without conscious thought.

What Is Still Missing

The e-skin is only half of the equation. It detects pressure and temperature, but it cannot yet relay that information back to the wearer as a felt sensation. Haptic stimulation hardware is the other half of a future bionic skin, and integrating the two remains an open problem. Even so, having a custom-fit, high-resolution sensing layer that can be manufactured without clean-room facilities is a meaningful step. It lowers the barrier for prosthetics clinics and research labs that want to experiment with tactile feedback without needing a semiconductor budget.