Multi-parameter e-skin based on biomimetic mechanoreceptors and stress field sensing

Electronic skin for robotics must sense multiple tactile modalities with high accuracy, but traditional multimodal approaches require dense, heterogeneous sensor arrays that complicate fabrication and integration. Field-based strategies reconstruct spatially continuous fields (e.g., stress, temperature, magnetic) from limited data to infer tactile information, yet prior optical stress-field methods are hard to apply on curved, stretchable surfaces, and other material-based methods yield only coarse classifications. Inspired by human skin, which converts contact forces into internal stress fields sensed by distributed mechanoreceptors, the authors propose biomimetic mechanoreceptors (BMRs): a layered e-skin with a soft deformation layer sampled by top (“epidermal”) and bottom (“dermal”) pressure-sensing arrays. The research question is whether analyzing the sampled stress field enables precise 3D force reconstruction and hardness perception with simple, modular hardware that is robust to assembly variations and curvature. The study aims to demonstrate accurate multi-parameter tactile sensing, improved effective spatial resolution, and modularity akin to human skin.

The paper reviews two main tactile sensing paradigms. (1) Direct multimodal sensing using combined arrays for pressure, temperature, humidity, etc., which improves accuracy with higher density and variety but suffers from complexity and limited sensor types. (2) Field-based approaches that reconstruct stress, temperature, or magnetic fields. Vision-based (e.g., GelSight-like) stress-field sensing can infer shape and hardness but is limited by camera size/rigidity for curved, stretchable settings. Approaches that use the sensing material to carry stress/temperature fields enable motion classification but lack precise physical outputs and are not aligned with human tactile interpretation. The authors motivate a skin-inspired, field-sensing strategy that combines deformable mechanics and distributed pressure sampling to overcome these limitations.

Device design: BMRs adopt a sandwich structure: a soft elastomeric deformation layer (Ecoflex Gel, Shore 000-35, 7 mm thick) between an epidermal and a dermal pressure-sensing array. Each sensing array comprises PDMS substrate, upper Cu/PI serpentine electrode, porous TPU/CB piezoresistive sensor array, protective patterned silicone, and lower Cu/PI serpentine electrode, in an island-bridge layout (1.5 × 1.5 mm islands, 5 mm spacing) with 150 µm-wide serpentine interconnects. Arrays have 8 × 8 sensing units made by laser direct writing. Patterned silicone walls around each unit, inspired by Meissner corpuscles, mechanically isolate and protect units and maintain separation of the resistive layers when unloaded. Sensing material: Porous TPU/CB foam (100–200 µm thick sheets, central 100 µm used) is prepared via a NaCl-templated process; it provides high sensitivity, ~35 ms response time, and stable performance. Assembly: Cross-alignment of top/bottom arrays with patterned silicone creates isolated cells. The Ecoflex Gel layer is cast in a 3D-printed mold and bonded to the arrays to complete the BMRs. 3D force sensing principle: External 3D force induces a stress field in the deformation layer. Pressure distributions at the top and bottom surfaces are sampled by the epidermal and dermal arrays. The centroids of the two distributions are computed; the deviation vector between dermal and epidermal centroids encodes force magnitude and direction: deviation length correlates with force magnitude and polar angle, and deviation direction equals azimuth angle. Experimental setup: A custom 3D force platform with XYZ gantry (0.01 mm repeatability), a force sensor (0.05 N sensitivity), a 2-DOF servo for polar angle and a manual rotation stage for azimuth, applies controlled loads. Tests used 5 N force, polar angles 40°–140° (90° vertical), azimuth 0°–90° (fourfold symmetry), pressing at corrected positions. Data acquisition via a 16 × 16 passive-matrix readout: dual 74HC595 column drivers, CD74HC4067 row selector, LM324-based virtual ground to suppress crosstalk, sampled by Arduino Mega at 14 fps, 8-bit digitization. Raw 8 × 8 maps are interpolated (cubic spline) for visualization/analysis. Hardness sensing: Hemispherical indenters (radius 1.5 cm) of varying hardness (e.g., Shore 0018, 0030, 0065, 0080) are pressed at the same speed until 5 N. Two metrics are used: (i) pressure sum (total grayscale of pressure map) and (ii) diffusion length, defined as the mean distance of pixels from the centroid (DL = ΣΣ d_ij / ΣΣ), computed separately for top and bottom arrays over time/force. Modularity and calibration: To emulate assembly errors, epidermal and dermal arrays are intentionally misaligned (rotation + translation). An affine transformation is calibrated using two test points to align epidermal and dermal coordinate frames, restoring centroid overlap. The device is also mounted on curved robot surfaces (arm and elbow) to demonstrate robustness to curvature-induced local displacements. Modeling: Finite element analysis (COMSOL 5.4) models a 50 × 50 × 7 mm deformation layer as Yeoh hyperelastic material (C1=1000, C2=2670, C3=6.39). A rigid cubic applicator loads external forces to approximate static friction during pressing. Additional materials processing and tests: Detailed fabrication steps for TPU/CB films, BMR assembly, indenter preparation, and basic sensing unit characterization (sensitivity and fatigue) are provided, including laser etching parameters and silicone dispensing for structural modulation.

- Accurate 3D force reconstruction from stress-field analysis using centroid deviation between epidermal and dermal pressure maps. - Angular resolution: polar angle standard error as low as ~1.8° at 40°, increasing to ~5° near vertical (90°); azimuthal angle standard error ~3.5°. - Spatial precision: deviation length standard error ≤ 0.16 mm overall; for vertical force, ~0.07 mm, corresponding to up to 71× improvement over the 5 mm sensor pitch. - Hardness sensing: For hemispherical indenters pressed to 5 N at constant speed, pressure sum curves are clearly separated by hardness (harder indenters reach target force faster). Diffusion length increases with force; harder indenters yield larger diffusion lengths and a smaller diffusion-length ratio between epidermal and dermal layers, enabling discrimination of object hardness near skin hardness via simple press actions. - Modularity and robustness: Intentional misalignment between epidermal and dermal arrays (rotation + translation) can be calibrated via an affine transformation using two points, restoring coincident centroids and yielding 3D force results consistent with perfectly aligned devices. The BMRs function consistently when mounted on curved robotic arm and elbow surfaces, indicating insensitivity to array placement and surface curvature. - Simple, exchangeable components: Because sensing depends primarily on the deformation layer’s mechanics and stress field continuity, the epidermal/dermal arrays can be modular, with different sensing principles or mechanical properties tailored to specific applications (e.g., fingertip sensitivity vs. joint durability).

The study demonstrates that sensing and analyzing the internal stress field of a soft deformation layer allows precise extraction of multiple tactile parameters—3D force direction/magnitude and object hardness—from comparatively simple hardware. By sampling pressure at the top and bottom surfaces and using centroid-based features, the approach achieves high angular accuracy and submillimeter effective spatial precision far exceeding the physical sensor pitch. This field-centric strategy reduces the need for densely integrated heterogeneous sensors, simplifying fabrication and improving robustness. The method also tolerates assembly errors and surface curvature, since the stress field is governed by mechanical properties rather than exact sampling locations, and misalignments can be corrected via affine calibration. Together, these results directly address the challenge of accurate, general-purpose tactile sensing in robotics with a modular, scalable e-skin architecture inspired by biological mechanoreceptors.

The authors present a biomimetic electronic skin (BMRs) that uses a soft deformation layer and dual distributed sensing arrays to reconstruct stress fields for multi-parameter tactile perception. The system measures 3D forces with polar and azimuthal angle resolutions of about 1.8° and 3.5°, achieves up to 71-fold improvement in effective spatial resolution (down to ~0.07 mm deviation length error under vertical loads), and identifies object hardness via pressure-sum dynamics and diffusion-length metrics. Its modular design is resilient to misalignment and applicable on curved surfaces, enabling exchangeable components tailored to different robotic tasks. This work provides a simple yet powerful and broadly applicable strategy for versatile tactile sensing in robotics by leveraging stress field continuity and biomimetic mechanoreceptor design.