Stiffness‐Gradient Auxetic Metamaterials Engineering Enables High Linearity of Strain Sensor for Facial Expression Recognition

ABSTRACT Flexible strain sensors are key components for humanoid robots to achieve high‐precision facial expression monitoring and gesture recognitions. However, they still face challenges in realizing high linearity under high sensitivity, which is mainly due to the uncontrollable reconstruction of the conductive network of sensitive materials under large deformations. Here, we propose a flexible strain sensor with a stiffness‐gradient re‐entrant anti‐tetrachiral auxetic metamaterial (SG‐RATCAM), which achieves high linearization by leveraging the complementary effect of the nonlinear responses of sensitive units in different stiffness regions. The sensor is fabricated by in‐situ printing using laser‐assisted direct ink writing (DIW) technology and its local stiffness is regulated by the laser power density (LPD). By exploiting the differences in the sensing characteristics of RATCAM sensors with different stiffnesses, the SG‐RATCAM is forward‐designed based on the theoretical model. Thus, the sensor exhibits high linearity (GF = 104.68, R 2 = 0.99) within the strain range of 0%–25%, and achieves a minimum detection limit of 0.04%. As a proof of concept, this sensor has been successfully applied to facial expression recognition and dynamic monitoring of dielectric elastomer actuator (DEA), providing a reliable solution for human‐machine interaction and integrated sensing and actuation in soft robotics.