Unlocking Dynamic Subtle Stimuli Tactile Perception: A Deep Learning-Enhanced Super-Resolution Tactile Sensor Array with Rapid Response

Unlocking Dynamic Subtle Stimuli Tactile Perception

A lightweight and fast-moving ping pong ball bounces on our designed flexible tactile sensor array, whose topology is optimally designed to maximize the sensing area with a minimal number of sensors. When the ball contacts the sensor array surface, three adjacent tactile sensors are activated, and 23-channel tactile data are transmitted to a super-resolution localization model. Beneath the sensor array is the graph convolutional neural network-based super-resolution model, designed based on the sensor array’s topological structure. This network effectively infers the contact position by extracting spatiotemporal features from the data, where the localization process and the generation of virtual tactile elements (taxels) are performed at the lowest layer. The combination of a well-optimized array topology and a super-resolution algorithm introduces virtual taxels into the physical tactile sensor array, ultimately achieving super-resolution localization. In the background, numerous sensor units are used to pave the area, further illustrating the sensor’s structure and its scalability. More details can be found in article number 2400913 by Geng Yang and co-workers.