High‐Load Capable Soft Tactile Sensors: Incorporating Magnetorheological Elastomer for Accurate Contact Detection and Classification of Asymmetric Mechanical Components

Abstract

Soft tactile sensors are soft and sufficiently flexible for attachment to a robot's gripper to enhance human‐like sensory capabilities. However, existing tactile sensors exhibit large size and a limited force measurement range. This article presents a novel design of a new soft tactile sensor for a robotic gripper, incorporating a sandwich‐like multilayered structure, together with a deep learning (DL) model, which overcomes the limitations of traditional sensors. The structure consists of three distinct layers: a 15 wt% iron magnetorheological elastomer, a flexible printable circuit board layer equipped with three‐dimensional Hall sensors (TLE493D; Infineon), and permanent magnets. Additionally, a multilayer perceptron network that can classify the loading state is adopted for the DL model. This new tactile sensor is capable of performing three distinct functions simultaneously: measurement of normal forces up to 3.73 kgf, identification of the precise location of force occurrence by subdivision into intervals of 2.5 mm, and differentiation between a wide (≈8 mm) and narrow (≈2 mm) contacted surface area. This newly developed soft tactile sensor has considerable potential for improvement in the performance of robotic grippers through its high accuracy, resolution, and large measurement range, as demonstrated by experimentation with the sensor attached to a real gripper.