Enhancing sensitivity and measurement range by Multi-layered Vision-Based Tactile Sensor (ML-VBTS): A parametric study and comparative benchmarking

Vision-Based Tactile Sensors (VBTSs) have significantly advanced robotic applications. However, they are often constrained by the inherent trade-off between sensitivity and measurement range, making the simultaneous enhancement of both parameters particularly challenging. This trade-off limits their applicability for applications demanding both subtle force detection and broad measurement capabilities. To address this challenge, we present a multi-layered VBTS (ML-VBTS) that integrates materials with different hardness levels within the tactile skin. Based on the proposed multi-layered design, we conduct a Finite Element Analysis (FEA) to optimize material properties and bridge the gap to real-world sensor performance using a Sim2Real transfer framework. Rigorous experiments on the ML-VBTS demonstrated its superior performance compared to the commercial GelSight mini, achieving approximately a 60% improvement in both sensitivity, with values of 0.9 N/px for normal forces and 0.4 N/px for shear forces, and in measurement range. This proposed sensor design not only enhances the precision and reliability of tactile feedback in advanced robotic applications but also paves the way for broader adoption of VBTS in various industries where sensitive yet broad force measurements are critical. • Developed ML-VBTS optimizing sensitivity & range via multi-hardness materials. • Used FEA and Sim2Real framework to optimize sensor design for real-world accuracy. • Ablation study confirmed each layer’s role in boosting sensor performance overall. • Acheived 0.9 N/px (normal) & 0.4 N/px (shear), 60% better than GelSight Mini.