Curvature-Aware Calibration of Tactile Sensors for Accurate Force Estimation on Non-Planar Surfaces

TL;DR

This work tackles the challenge of measuring contact force with flexible tactile sensors mounted on curved surfaces. It introduces a curvature-aware calibration framework that jointly estimates local surface curvature $\kappa$ from baseline sensor data via a residual MLP and then calibrates force using a curvature-aware model. The curvature prediction achieves $R^2=0.91$, and the force calibration attains $R^2\approx0.92$ on curved geometries, demonstrating improved accuracy and consistency over flat-surface calibration. The approach is validated on five everyday objects across 2–8 N of force and is released with open hardware and software to enable practical adoption in robotic and wearable applications.

Abstract

Flexible tactile sensors are increasingly used in real-world applications such as robotic grippers, prosthetic hands, wearable gloves, and assistive devices, where they need to conform to curved and irregular surfaces. However, most existing tactile sensors are calibrated only on flat substrates, and their accuracy and consistency degrade once mounted on curved geometries. This limitation restricts their reliability in practical use. To address this challenge, we develop a calibration model for a widely used resistive tactile sensor design that enables accurate force estimation on one-dimensional curved surfaces. We then train a neural network (a multilayer perceptron) to predict local curvature from baseline sensor outputs recorded under no applied load, achieving an R2 score of 0.91. The proposed approach is validated on five daily objects with varying curvatures under forces from 2 N to 8 N. Results show that the curvature-aware calibration maintains consistent force accuracy across all surfaces, while flat-surface calibration underestimates force as curvature increases. Our results demonstrate that curvature-aware modeling improves the accuracy, consistency, and reliability of flexible tactile sensors, enabling dependable performance across real-world applications.

Figures (8)

• Figure 1: Predicted Force Response on Flat and Curved Surfaces. (a) On a flat surface, the calibrated sensor predicts forces that closely match the ground truth. (b) On a curved surface, using the same flat calibration leads to significant deviation from the ground truth, highlighting the need for curvature-aware calibration.
• Figure 2: Tactile Sensing Hardware. The sensor was constructed by sandwiching a Velostat film between two orthogonal layers of conductive threads, forming a resistive array. A custom PCB handled row-column scanning, sequentially addressing each node and sending the measurements to an Arduino. The circuit schematic illustrates the scanning architecture.
• Figure 3: MLP Network Architecture. Residual multilayer perceptron that predicts surface curvature directly from baseline tactile sensor data, mapping 24 input features to a single curvature value.
• Figure 4: MLP Training Result. Prediction performance of the residual MLP on the test set, showing predicted versus true curvature values. Each curvature level includes five independent samples.
• Figure 5: Force Calibration Experimental Setup: (a) Tactile sensor mounted on a curved fixture. (b) 3D printed fixtures with different curvatures. (c) Indenter mounted on a force/torch sensor to apply controlled loads. (d) In each trial, the indenter pressed the four central nodes simultaneously, forming a node block, to collect sensor responses under varying surface curvatures.