A tactile sensor that turns pressure into color — with no latency

A team from Queen Mary University of London and three Italian universities has shown a tactile sensor for robot fingertips in which the material itself changes color under pressure — red at a light touch, through green, to blue at the strongest deformation. Force is encoded in the color at the material level, so reading it requires no extra computation. IEEE Spectrum reported the work, which the university announced in early July 2026.

Key takeaways

Spatial resolution of 100 micrometers, with no computational latency
The material's color shifts from red (light touch) through green to blue (strongest deformation)
The core is a Bragg reflector — polymer layers with different refractive indices
The light-sensitive layer is set by a 7-minute exposure to a 635 nm red laser
Authors: Giacomo Sasso and Federico Carpi's group (Queen Mary University of London, plus universities in Florence, Trieste and Trento)

How a "skin" that sees touch works

Camera-based tactile sensors — like the widely used GelSight — work indirectly: a camera watches a gel deform, and software then computes the force and contact shape from that image. It works, but the processing adds latency.

The new sensor flips that logic. Pressure is not computed from an image — it is directly visible as color. A Bragg reflectorBragg reflector: A stack of alternating layers with different refractive indices that reflects a specific wavelength of light; its color shifts as it deforms. does the work: a mechanochromicMechanochromism: A material changing color in response to mechanical pressure or deformation. material that reflects different wavelengths of light depending on deformation.

As the layer compresses, the spacing between its internal layers changes — and so does the reflected color. The sensor is built from three layers: a protective black silicone layer for contrast, the Bragg reflector itself, and a transparent silicone fingertip with an embedded LED and a tiny camera that captures the color changes from the inside.

The core aspect is that we're essentially moving the sensing element to the material level.

Giacomo Sasso, postdoctoral researcher in Federico Carpi's group, Queen Mary University of London.

A laser instead of a print

The layers with different refractive indices are not deposited mechanically — they form within the material itself. A light-sensitive film is exposed for 7 minutes to a red laser at a wavelength of 635 nm. This sets alternating layers of different density in the polymer, which give the sensor its optical behavior.

The 100-micrometer resolution means the sensor captures very fine detail. In demonstrations, the team mapped the topology of a human fingertip, the relief of a U.S. penny, and the texture of a leaf — a level of detail closer to human touch than to coarse contact detection.

Where it fits

High-resolution touch is one of the weakest points of today's robots, especially humanoids and manipulators. Precise grasping of soft or slippery objects requires sensing force and slip in real time — and the longer the delay in the control loop, the harder that becomes.

Humanoid makers are moving toward richer sensing in the hands — we recently covered the new NEO hands with 25 degrees of freedom and tactile sensing. The Queen Mary approach represents a different philosophy: rather than multiplying separate sensors, it writes the measurement into the "skin" itself.

For now, this is research, not a product. The team names sensing non-flat surfaces and surgical-instrument applications — where both precision and low latency matter — as its next targets.

Why it matters

The real value of this sensor is not the resolution itself but where the measurement happens. Moving the sensor's "intelligence" from software into the material shifts the balance of the whole system: less computation in the control loop means faster response and simpler integration.

In manipulation robotics, tactile latency is a genuine constraint — it decides whether a robot reacts before an object slips from its grip. If the mechanochromic approach can be carried onto curved, multi-sensor surfaces and hold up under repeated use, it could become an alternative to the gel-based sensors that dominate today. It is also a sign of a broader trend in embodied AI: some of the work now done by models and processors is moving back into the physics of the material.

What's next?

The team plans to extend the sensor to non-flat surfaces — a prerequisite for use on curved fingertips and robot grippers
Surgical instruments are named as a target application, where low latency and high resolution matter directly
Open questions remain around the durability of the mechanochromic layer and scaling production beyond the lab