HMN 2026: How Artificial skin enables robots to simultaneously sense temperature and pressure like humans

A research team led by Prof. Seung Hwan Ko of Seoul National University College of Engineering’s Department of Mechanical Engineering has developed an artificial skin technology that enables robots to sense temperature and pressure simultaneously, similar to human skin.

The team created a novel multimodal tactile sensor capable of detecting both thermal and mechanical stimuli within a single ultrathin device. Inspired by the way human skin processes sensory information, the sensor is designed to efficiently extract temperature and pressure data from a single integrated platform.

Using a single attachable sensor combined with a wireless switching board and artificial intelligence, the researchers demonstrated the ability to identify 20 everyday objects with high accuracy comparable to human tactile perception.

The study, published in Nature Materials, further confirmed that the technology can be extended to achieve high-resolution sensing at levels comparable to human touch. As such, the technology is expected to serve as a key enabler for emerging Physical AI systems.

Recently, “Physical AI,” which enables robots and artificial intelligence systems to interact with the physical world, has gained increasing attention. Physical AI goes beyond simple computation, allowing machines to see, touch, feel and make decisions based on their environment. Sensors capable of simultaneously detecting multiple tactile inputs—such as temperature and pressure, similar to human skin—are considered essential for realizing such systems.

Limits of stacked sensor designs

Human skin can process diverse stimuli, including temperature and pressure, rapidly and precisely. However, existing multimodal sensory devices that attempt to replicate this functionality have typically relied on combining multiple sensors or stacking multiple functional layers.

These approaches result in complex system structures, increased device size, slower response times due to reactive elements, and difficulty precisely detecting multiple stimuli at the same location.

Therefore, there has been strong demand for a new artificial tactile platform that can process complex stimuli quickly using a single thin, flexible sensor. In particular, multimodal tactile sensing technologies that mimic human skin are crucial for Physical AI, enabling robots to perceive their surroundings in a humanlike manner.

A single layer that switches modes

To address this challenge, Ko’s team developed a device based on a core-shell nanowire network composed of a silver (Ag) core and a copper oxide (Cu?O) shell. The device can switch between thermal sensing mode (T mode) and mechanical sensing mode (M mode) 16 times per second within a single structure.

Thanks to its ultrathin single-layer design, the sensor achieves rapid response times—sub-microsecond for mechanical stimuli and millisecond-level for thermal stimuli.

In object classification experiments, the team trained an AI model using interleaved signals from both sensing modes. The classification accuracy improved significantly from approximately 65% (when using only thermal or mechanical signals) to 95%. Even with reduced data input, the model maintained a high accuracy of 94.53%.

In further validation using a fingertip-mounted sensor integrated with a wireless measurement board, the system achieved an accuracy of 83% across 20 everyday objects.

From object sensing to artificial skin

The researchers also developed a multi-array platform capable of measuring temperature and pressure distributions at spatial resolutions comparable to human skin. This demonstrates that the technology can be scaled beyond single-device sensing to full artificial skin systems with humanlike spatial resolution.

The multimodal artificial tactile sensor developed in this study is expected to be widely applied in fields such as prosthetics, wearable electronic skin, soft robotics, robotic grippers and human-machine interfaces.

It is anticipated to become a core technology for tactile perception in next-generation robots and Physical AI systems. Because the device can process multiple stimuli within a single ultrathin layer—without requiring complex stacking of multiple sensors—it offers significant advantages in system simplification and high-resolution sensing. This positions it as a promising foundation for next-generation intelligent tactile platforms.

Ko stated, “This study is significant in that it demonstrates, for the first time, the ability to process both thermal and mechanical stimuli within a single ultrathin device without stacking multiple sensors. We expect this technology to evolve into a core solution for enabling human-level tactile perception in robots and to be widely applied in wearable electronic skin, prosthetics and soft robotics.”

The study’s co-first authors, Kyun Kyu Kim and Junhyuk Bang, received their Ph.D. degrees from SNU’s Department of Mechanical Engineering. Kim is currently working at Apple in the United States, while Bang is continuing his research as a postdoctoral researcher at the California Institute of Technology.

The content is provided for information purposes only.