Made by AI, Directed by David Stein

Search
Research & Innovation

A 3D-Printed 'Skin' Gives Humanoid Robots a Sense of Touch, With Just 16 Electrodes

New research uses electrical impedance tomography — the same imaging principle behind some medical scanners — to turn a printable TPU shell into a whole-body pressure sensor, without embedding a dense grid of individual touch sensors.

2 min read

Giving a robot a sense of touch usually means embedding a grid of discrete sensors across its surface — more coverage requires more sensors, more wiring, and more failure points. A new paper, “Toward Geometry-Scalable Whole-Body Touch for Humanoids: A 3D-Printed Conformal EIT Skin,” takes a different approach: rather than distributing individual sensors, it turns the skin itself into the sensor, using a technique borrowed from medical imaging.

Imaging pressure instead of measuring it point by point

The method is electrical impedance tomography (EIT) — the same underlying principle used in some medical scanners, which reconstructs an internal image from surface electrode measurements of electrical resistivity. Applied here, the skin is a flexible TPU (thermoplastic polyurethane) layer with conductive fabric patches attached to a top TPU cover. Touching or pressing the skin disturbs the resistivity field, and the EIT algorithm reconstructs where and how hard the contact happened from those disturbances — a continuously updating pressure map derived from relatively few physical measurement points, rather than one sensor reading per touch location.

A small number of electrodes, a genuinely useful map

The prototype uses just 16 electrodes and still produces what the paper describes as a fairly accurate pressure map — a meaningfully different scaling story than a discrete-sensor approach, where resolution is directly tied to sensor count. That’s the “geometry-scalable” part of the paper’s title: because the skin itself is doing the sensing rather than an embedded grid, the same basic approach can in principle be conformed to the irregular, curved surfaces of a humanoid body — hands, joints, a torso — without redesigning a sensor layout for every new shape. Hackaday’s own read on reproducibility is candid about where the real difficulty sits: “characterization of the TPU porosity and such along with the EIT algorithm… probably being the biggest hurdles for hobbyist recreations” — this is closer to a real fabrication and signal-processing problem than a weekend build.

Why it matters beyond the lab

Touch sensing is one of the more persistent gaps between humanoid robots that can walk and manipulate objects and humanoid robots that can safely, naturally share physical space with people — reacting to an unexpected bump, modulating grip force, or noticing incidental contact the way skin does automatically. A whole-body sensing approach that scales with printable geometry rather than sensor density is a real step toward humanoids that can respond to touch anywhere on their body, not just at a few instrumented contact points like fingertips.

Related Reading