Development of a Soft Tactile Sensor Array for Contact Localization Estimations

Contact localization is an essential feature of tactile sensors for biomedical, rehabilitation, surgical, and service robot applications where interaction with the environment is needed. In this work, we have developed a piezoelectric (PZT) based soft tactile sensor using the layer-by-layer fabrication technique. We placed
$4\times 4$
mm2 size PZT elements in
$3\times 3$
matrix form with 3 mm spaces and embedded them in silicone layers. Although the sensor has discrete brittle sensing elements, the silicone layer distributes the contact force on the surface area, which enables continuous contact localization on the sensor. Here, we studied the estimation using the collected and conditioned output voltage signals of PZT elements. First, we used a logic-based algorithm that gives results qualitatively using a developed user interface. Then we used machine learning algorithms to estimate the contact position coordinates numerically. Finally, we applied Artificial Neural Network (ANN), Decision Tree (DT), Support Vector Machine (SVM), and k-Nearest Neighbor (KNN) algorithms and compared the results. We got our best results using a two-layer ANN algorithm which provides a mean deviation of 0.08 mm for the X-axis and 0.07 mm for the Y-axis. Besides, we had 97.20% of testing estimations with an error lower than 0.35 mm. Therefore, we believe our tactile sensor gives promising results and can be used in small-scale applications where contact localization estimation is desired.