A thin and soft optical tactile sensor for highly sensitive object perception

TL;DR

The paper addresses the need for safe, high-resolution tactile sensing in soft robotics by tackling the limitations of vision-based tactile sensors that require bulky, alignment-sensitive optics. It proposes a thin, alignment-free speckle-based optical tactile sensor built from a soft silicone elastomer, a single optical fiber, and a compact camera, with a CNN decoding tactile information from cropped speckle images. Key results include a force sensing RMSE around $0.04\ \mathrm{N}$ and a nine-class texture recognition accuracy of $93.33\%$, demonstrating strong performance while maintaining a compact, conformable form factor. The work offers a scalable, low-complexity tactile sensing paradigm suitable for soft robotics and wearable haptic interfaces, reducing optical alignment constraints while delivering high sensitivity and texture discrimination.

Abstract

Tactile sensing is crucial in robotics and wearable devices for safe perception and interaction with the environment. Optical tactile sensors have emerged as promising solutions, as they are immune to electromagnetic interference and have high spatial resolution. However, existing optical approaches, particularly vision-based tactile sensors, rely on complex optical assemblies that involve lenses and cameras, resulting in bulky, rigid, and alignment-sensitive designs. In this study, we present a thin, compact, and soft optical tactile sensor featuring an alignment-free configuration. The soft optical sensor operates by capturing deformation-induced changes in speckle patterns generated within a soft silicone material, thereby enabling precise force measurements and texture recognition via machine learning. The experimental results show a root-mean-square error of 40 mN in the force measurement and a classification accuracy of 93.33% over nine classes of textured surfaces, including Mahjong tiles. The proposed speckle-based approach provides a compact, easily fabricated, and mechanically compliant platform that bridges optical sensing with flexible shape-adaptive architectures, thereby demonstrating its potential as a novel tactile-sensing paradigm for soft robotics and wearable haptic interfaces.

Figures (7)

• Figure 1: (a) Schematic of the sensor operation. The camera captures the speckle pattern induced by laser illumination in the soft material. A central $128 \times 128$ region is cropped and input to a decoding model (CNN) for tactile information extraction. (b) Structure of the proposed sensor. (c) Speckle patterns collected under different contact conditions. (d) Proposed sensor on a human hand.
• Figure 2: Decoding model based on CNN.
• Figure 3: (a) Experimental setup for position recognition. (b) Proposed sensor placement on a flat surface. (c) Proposed sensor placement on a curved surface. (d) Confusion matrix of the position recognition results. The left side shows the result placed on a flat surface; on the right is the result placed on a curved surface.
• Figure 4: (a) Experimental setup for force measurement. (b) Three locations (labeled “A”, “B”, and “C”) used for force measurement. (c) Scatter plots of the predicted forces versus the ground-truth values. RMSE and MAE denote the root-mean-square error and mean absolute error, respectively. $R^2$ is the coefficient of determination. (d) Histograms of the absolute errors.
• Figure 5: (a) Experimental setup for Mahjong tile classification and the tactile sensor. The tactile sensor was designed to be mounted on a robotic gripper and used to grasp different tiles. (b) Mahjong tiles used in this experiment. The tiles are rigid cubes measuring $1.9,\text{mm} \times 1.5,\text{mm} \times 2.6,\text{mm}$, with different patterns engraved on their surfaces.