Artificial finger can identify what common material things are made of

Smart finger uses sensors to detect substances such as glass, silicon and wood with more than 90 per cent accuracy, which could be useful for robotic manufacturing tasks

By Alex Wilkins

An artificial finger can identify different materials with more than 90 per cent accuracy by sensing their surface. The technology could be useful for automating robotic manufacturing tasks, such as sorting and quality control.

Touch sensors that can gain information about surfaces, such as pressure or temperature, aren’t new, but sensors that can recognise the type and roughness of surfaces are less common.

Dan Luo at the Chinese Academy of Sciences’s Beijing Institute of Nanoenergy and Nanosystems and his colleagues have developed a finger that can identify what a material is made from by using “triboelectric” sensors to test its ability to gain or lose electrons, and discern its roughness, without causing damage to it.

When trialled on hundreds of samples of 12 substances such as wood, glass, plastic and silicon, and combined with a machine learning-based data analysis, the finger achieved an average accuracy of 96.8% and at least 90 per cent accuracy for all of the materials.

The device consists of four small square sensors, each made of a different plastic polymer, chosen for their different electrical properties. When the sensors move close enough to the surface of an object, electrons from each square interact with the surface in a slightly different way, which can then be measured.

These sensors, housed in a finger-like case, are then attached to a processor and an organic LED screen, which displays the name of the detected material type. In an industrial setting, the processor could be connected directly to a manufacturing control mechanism. “Smart fingers could help robots check whether products meet manufacturing standards, in terms of composition and surface structure,” says Luo. “Our system could also play an important role in industrial material sorting.”

If it is shown to be robust over many thousands of tests, the sensor’s ability to differentiate between materials could make it well-suited for tasks like quality control in manufacturing, says Ben Ward-Cherrier at the University of Bristol, UK. However, it would probably prove more effective when combined with other sensors that can detect things such as edges or friction.

Luo and his team also suggest that the device could be used for artificial prosthetics, but it is unlikely it would be useful for that, says Tamar Makin at the University of Cambridge. “For technology that is human controlled, we don’t need this level of sophistication,” she says. “Imagine you’re an amputee and you’re reaching out for a cup of coffee. You have so much life experience, and online experience with your intact hand, to have a very good estimate of the material that you’re about to reach.”

Journal reference: Science Advances, DOI: 10.1126/sciadv.abq2521