Researchers from the University of Cambridge and University College London (UCL) have developed a new type of electronic skin for robots. This flexible, conductive material can be shaped into complex forms and is capable of sensing and processing various physical inputs. The aim is to enhance robotic interaction with the physical world.
Unlike existing robotic touch solutions that use sensors embedded in specific areas, this entire electronic skin acts as a sensor, similar to human skin. While not as sensitive as human skin, it detects signals through over 860,000 tiny pathways, allowing it to recognize different types of touch and pressure. These include finger taps, temperature changes, damage like cuts or stabs, and multiple simultaneous touches.
The team used physical tests combined with machine learning techniques to teach the robotic skin which pathways are most significant for sensing contact efficiently. The potential applications extend beyond humanoid robots or prosthetics to sectors such as automotive and disaster relief. Their findings are published in Science Robotics.
Dr. David Hardman from Cambridge’s Department of Engineering stated: “Having different sensors for different types of touch leads to materials that are complex to make.” He emphasized their goal was to create a single material capable of detecting multiple touch types simultaneously.
Co-author Dr. Thomas George Thuruthel from UCL added: “At the same time, we need something that’s cheap and durable, so that it’s suitable for widespread use.”
The researchers employed multi-modal sensing using a gelatin-based hydrogel shaped like a human hand. They experimented with various electrode configurations to optimize information collection about different touches. From just 32 electrodes at the wrist, they gathered over 1.7 million data points across the hand due to the conductive material's pathways.
The skin underwent testing against heat guns, fingers, robotic arms, gentle touches, and scalpel cuts. Data from these tests trained a machine learning model to interpret various touches accurately.
“We’re able to squeeze a lot of information from these materials – they can take thousands of measurements very quickly,” said Hardman. He noted that although it's not yet on par with human skin sensitivity, their method surpasses existing alternatives in flexibility and ease of construction.
Thuruthel commented: “Our method is flexible and easier to build than traditional sensors.”
Future research aims at improving durability and conducting further real-world tests on robotic tasks.
This project received support from Samsung Global Research Outreach Program, the Royal Society, and UK Research Innovation's EPSRC branch. Professor Fumiya Iida co-authored the study alongside David Hardman and Thomas George Thuruthel.
Reference:
David Hardman et al., ‘Multimodal information structuring using single layer soft sensory skins and high-density electrical impedance tomography.’ Science Robotics (2025). DOI: 10.1126/scirobotics.adq2303