An international team of researchers has developed a robotic hand that overcomes a long-standing hurdle in robotics: the lack of a sensitive, human-like sense of touch. The development, dubbed the “F-TAC Hand,” has now been presented in the renowned journal Nature Machine Intelligence and demonstrates how crucial rich tactile perception is for interacting with the real world. Scientists from Queen Mary University of London in the United Kingdom, the Beijing Institute for General Artificial Intelligence, and Peking University in China are behind the project.The central challenge of previous robotic hands was their relative clumsiness in unpredictable situations. Despite sophisticated mechanics, they lacked the direct feedback to adequately respond to subtle touches or displacements of an object. The F-TAC Hand addresses this problem with unprecedented sensor density.
The innovation lies in the complete sensory coverage. A full 70 percent of the palm is equipped with high-resolution, vision-based sensors. Imagine it this way: Inside the fingers and palm, tiny cameras record the deformation of a flexible elastomer skin as soon as it touches an object. A neural network analyzes these camera images in real time and reconstructs a precise 3D geometry of the contact surface. With a spatial resolution of 0.1 millimeters and a density of 10,000 tactile measurement points (“taxels”) per square centimeter, this system far surpasses previous commercial solutions. Despite this comprehensive sensor technology, the hand retains its full, human-like mobility with 15 degrees of freedom and can perform all 33 documented human gripping techniques.
The system’s performance was demonstrated not only in theory, but also in 600 real-world experiments. In one task, the hand was required to grasp and hold several objects one after the other. If an unexpected contact occurred that would have resulted in a collision with the next object, the hand recognized the situation through its tactile feedback. Within just 100 milliseconds, it adjusted its strategy and selected an alternative, safer grip.
Systems without this feedback inevitably fail in such cases. “The massive spatial resolution combined with the enormous coverage is truly novel and hasn’t been possible before,” Professor Kaspar Althoefer of Queen Mary University of London toldÂ
Techxplore magazine . This ability for dynamic adaptation is the crucial step from a mere gripping machine to an intelligent manipulation tool. The potential fields of application are diverse. They range from flexible manufacturing in Industry 4.0 and use in telemedicine to a new generation of prosthetics and assistive robots that can operate safely in everyday human life. The F-TAC Hand provides empirical evidence that sophisticated “embodied AI” requires more than just computing power—it requires precise physical perception.
