X-Hinges prints motion sensors into the joints of everyday objects

A 3D-printed glove can relay finger movements to a robot hand; a folded lamp can change brightness when squeezed. Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory and Tianjin University demonstrate both with X-Hinges, a design method that builds sensing into the flexible joints of an object as it is printed.[1][2]

The central idea is to give different jobs to different conductive materials. One forms a sensing element whose electrical resistance changes with deformation. Another carries the signal through a trace with much lower resistance. Separating those functions helps the system distinguish movement at the joint from unwanted changes along the electrical connection.[2]

Designing movement and sensing together

The software lets a designer choose an object’s geometry, stiffness and intended movements, then embeds a suitable sensing arrangement. The structures can support lateral and vertical bending as well as compression. Instead of attaching a separate sensor to a finished moving part, the designer produces the mechanism and its sensing elements in the same multi-material print.[2][1]

The resulting electrical readings still need interpretation. A data-acquisition system collects the signals, and regression models estimate movement from their history. That is where the machine-learning component contributes: it helps decode how a flexible, resistive structure is moving rather than asking a language model to interpret an object.[2]

In the glove demonstration, the printed joints record finger motion and transmit it to a robotic hand. The lamp uses its deformation as an input for brightness. A tactile matrix provides another variation, using the pattern of physical contact to recognise objects. Each demonstration applies the same sensing concept to a different interaction.[1][2]

Calibration makes a new print usable

Small differences in printing and materials change the relationship between movement and resistance. The paper tests a calibration process that uses a camera to track a printed marker while the hinge is moved. The paired visual and electrical observations allow the decoder to adapt to that particular object.[2]

In one transfer test, a decoder trained on an earlier specimen had a mean angular error of 10.02 degrees on a new print. With labelled calibration data, the error fell to 3.96 degrees. A separate stressed specimen improved from 8.20 to 3.71 degrees. Those single-degree-of-freedom results show why calibration belongs in the fabrication process; they are not universal precision figures for every geometry.[2]

The work, accepted for UIST 2026, is aimed at fabrication researchers, designers and engineers. It requires multi-material printing, dedicated electronics and calibration. The authors also note that flexible conductive materials retain effects such as hysteresis, in which the reading depends partly on how the material was previously deformed.[2]

For prototyping, the attraction is that shape, motion and input can be designed together. A creator can explore how an interface should feel in the hand while making its movement measurable. That opens a route to customised controls and responsive objects whose sensing is part of their structure.

Report an error or suggest a correction