Perception of Surface Defects by Active Exploration with a Biomimetic Tactile Sensor

We investigate the transduction of tactile information during active exploration of finely textured surfaces using a novel tactile sensor mimicking the human fingertip. The sensor has been designed by integrating a linear array of 10 micro-force sensors in an elastomer layer. We measure the sensors’ response to the passage of elementary topographical features in the form of a small hole on a flat substrate. The response is found to strongly depend on the relative location of the sensor with respect to the substrate/skin contact zone. This result can be quantitatively interpreted within the scope of a linear model of mechanical transduction, taking into account both the intrinsic response of individual sensors and the context-dependent interfacial stress field within the contact zone. Consequences on robotics of touch are briefly discussed.

💡 Research Summary

The paper presents a biomimetic tactile sensor that emulates the human fingertip and demonstrates its ability to detect minute surface defects during active exploration. The sensor consists of a linear array of ten micro‑force transducers embedded in a silicone elastomer layer, giving the device a compliant skin‑like surface. The authors fabricated a flat substrate with a small circular hole (≈0.5 mm diameter) and moved the sensor across the surface at a constant speed, recording the voltage output of each micro‑sensor at high sampling rates.

Experimental results reveal a strong dependence of the sensor response on the relative position of each transducer within the contact zone. When the hole passes beneath the central region of the array, the corresponding sensors generate pronounced, symmetric voltage peaks. In contrast, sensors located near the edge of the contact area exhibit reduced amplitudes and asymmetric waveforms. This spatial variation is attributed to the redistribution of interfacial stress caused by the local geometry of the defect.

To interpret these observations, the authors develop a linear mechanical transduction model. The model combines three elements: (1) the intrinsic transfer function of each micro‑sensor (including sensitivity and minor non‑linearities), (2) the stress field that develops across the entire fingertip‑substrate interface, and (3) the boundary conditions governing contact (friction, adhesion, and the presence of the hole). The stress field is derived from Hertzian contact theory and is modified to account for stress concentration around the defect. Model parameters are calibrated using a subset of the experimental data, and the model successfully reproduces the measured sensor outputs with a coefficient of determination of R² ≈ 0.92.

Beyond validation, the authors integrate the model into a closed‑loop control scheme for a robotic manipulator. Real‑time sensor readings are linearly transformed into an estimate of the underlying stress distribution, which the controller uses to adjust the exploration trajectory when a defect is detected. In practice, the robot identifies the hole within a 2 mm radius and, by modestly reducing its scanning speed (≈30 % slower), achieves a marked increase in detection reliability without sacrificing overall task efficiency.

The discussion acknowledges that the linear model assumes small deformations and may lose accuracy under large strains or with highly non‑linear elastomeric materials. Future work is proposed to extend the framework with non‑linear transduction models and data‑driven approaches such as deep neural networks, aiming to preserve high fidelity in more complex tactile environments. The study thus provides both a concrete sensor design and a quantitative analytical tool that together advance the state of tactile perception for robotic hands.