Benchmarking Resilience and Sensitivity of Polyurethane-Based Vision-Based Tactile Sensors

TL;DR

This work tackles the durability limitations of vision-based tactile sensors that use silicone gels by introducing a polyurethane gel and a model-free benchmarking framework that separately evaluates mechanical resilience and sensitivity. The authors fabricate both silicone and polyurethane VBTS gels on a DIGIT platform, and subject them to cyclic compression, cyclic shear, abrasion, and force/spatial sensitivity tests, using MAE and SNR as key metrics. Results show polyurethane gels offer superior resilience across loading modes and while silicone delivers higher sensitivity at low loads, polyurethane can achieve strong performance under higher-load conditions, suggesting a use-case dependent material choice. The study provides a practical, model-free evaluation method to enable direct, hardware-centric comparisons across VBTS gels, guiding material selection for real-world deployments.

Abstract

Vision-based tactile sensors (VBTSs) are a promising technology for robots, providing them with dense signals that can be translated into an understanding of normal and shear load, contact region, texture classification, and more. However, existing VBTS tactile surfaces make use of silicone gels, which provide high sensitivity but easily deteriorate from loading and surface wear. We propose that polyurethane rubber, used for high-load applications like shoe soles, rubber wheels, and industrial gaskets, may provide improved physical gel resilience, potentially at the cost of sensitivity. To compare the resilience and sensitivity of silicone and polyurethane VBTS gels, we propose a series of standard evaluation benchmarking protocols. Our resilience tests assess sensor durability across normal loading, shear loading, and abrasion. For sensitivity, we introduce model-free assessments of force and spatial sensitivity to directly measure the physical capabilities of each gel without effects introduced from data and model quality. Finally, we include a bottle cap loosening and tightening demonstration as an example where polyurethane gels provide an advantage over their silicone counterparts.

Figures (6)

• Figure 1: Our polyurethane VBTS gel is capable of performing grasps without failure on rugged, heavy objects, including a 23.5 N piece of aluminum extrusion (left), a 40 N stool (center), and a 44.5 N weight (right).
• Figure 2: We perform four resilience tests: (A) cyclic compression on an indenter, (B) cyclic local shear on an indenter, (C) cyclic transverse shear on a flat surface, and (D) abrasion. The setup in (A) is also used to characterize the sensor's force sensitivity.
• Figure 3: Our model-free spatial sensitivity evaluation uses the sensor reading when pressed onto a ridged surface. The image is preprocessed via background subtraction and bandpass filtering before being cropped and run through a series of 1D FFTs. The resulting power spectral densities are averaged. The signal power is compared to a noise power, defined by the same frequency range from a flat, ridgeless surface, to obtain a signal-to-noise ratio (SNR). The background subtraction image is amplified for visualization purposes.
• Figure 4: Results for cyclic compression (A), cyclic shear on an indenter (B), cyclic shear on a flat surface (C), and abrasion (D) tests across three gel samples for silicone (SI) and polyurethane (PU). The first row represents sensor images captured at each cycle after unloading, while the second row shows results from loaded images. The raw sensor images in the third row depict the final unloaded sensor image (after 1000 cycles) from sample 1 of each material. Different sets of gels are used for each test.
• Figure 5: Results for force sensitivity (A) and spatial sensitivity (B) tests. In (A), the mean absolute error (MAE) is calculated with respect to the unloaded image to quantify signal change across force. In (B), the first and second rows show SNR under 2 N and 10 N loads, respectively. The left and right columns represent tests with varying amplitude (constant period) and period (constant amplitude).