A stretchable tactile sensor with deep learning-enabled 3D force decoding for human and robotic interfaces

Tactile sensing technology is crucial for applications in robotics, human-computer interaction, health monitoring, and prosthetics, aiming to replicate the sensitivity of human skin. In this work, we present a stretchable, multilayered tactile sensor system that emulates the complex structure of human skin, enabling precise detection of both static and dynamic pressure, as well as the intricate deformation patterns generated by joint and muscle movement. The sensor integrates a SEBS-based sensing layer, a graphene–PEDOT:PSS composite electrode, and an elastomeric encapsulation layer. This sensor architecture synergistically enhances sensitivity, mechanical robustness, and electrical responsiveness, addressing key limitations of earlier designs in terms of durability and strain-dependent performance. Leveraging a deep learning algorithm combining convolutional neural networks (CNN) and transformer models, the system accurately resolves three-dimensional force distributions and decouples normal and shear forces across multiple directions. We validate the sensor's performance through applications in American Sign Language (ASL) gesture recognition and wrist motion tracking, achieving classification accuracies of 97 % and 100 %, respectively. Additionally, the sensor supports stable robotic grasping by providing real-time force feedback. These results underscore the sensor's potential in wearable electronics and intelligent tactile interfaces, bridging the gap between complex human biomechanics and digital systems.