Tactile-sensitive artificial skin for multiaxial force detection and texture recognition

Abstract

Tactile perception, particularly at the fingertips, is fundamental to human dexterity, enabling fine motor control and reliable manipulation of objects through the precise real-time modulation of normal and shear forces based on encountered frictional conditions. To bridge this capability gap in robotics, a novel tactile-sensitive artificial skin is designed to significantly enhance robot-object interaction and environmental recognition. The artificial skin, fabricated from a 2-mm thick silicone elastomer membrane embedded with polymer optical fiber Bragg grating (FBG) sensor array, achieves large measurement range (detecting forces up to 10 N normal and ±4 N shear) and high sensitivity for multiaxial forces. The skin's performance was evaluated through tests involving normal force loading of up to 10 N and shear force loading around ±4 N using a three-axis translation gantry. Additionally, the study examines slip-induced vibrations on various textured surfaces. A multi-input multi-output convolutional neural network (MIMO-CNN) was developed to simultaneously estimate force and recognize textures based on multichannel FBG inputs. The MIMO-CNN achieved an R-squared value of 0.96 for force estimation and classification accuracy of 94% for texture recognition across 20 fabric samples. These findings highlight the potential of tactile-sensitive artificial skin to enhance robotic perception and manipulation, paving the way for more advanced humanoid robotic systems.