Bioinspired textured sensor arrays with early temporal processing for ultrafast robotic tactile recognition

Abstract

Rapid tactile processing is one of the most effective and direct strategies for robots to interact with surrounding environment. However, achieving both fast and accurate tactile recognition remains a challenge due to the inherent trade-off between sensor sensitivity and reaction time. In this study, we developed a bioinspired textured sensor array (TSA) using a circular grid arrangement, which could provide rich information on dynamic tactile processes in a self-powered manner. Early tactile process model (ETPM) was introduced to prioritize early-stage tactile data, which enables ultrafast decision-making speed without compromising classification accuracy. Specifically, our system achieved early predictions of object classification with an accuracy of 92 % while using only the initial 19 % (48 ms) of tactile data. The practicability of this system was examined through integration into a robotic arm. An ultrafast reaction time of 89 ms was achieved in real-time object property prediction, which is even faster than human hands. This advancement provides a robust foundation for rapid and precise tactile recognition in robotic perception systems, improving the robot’s response speed, reliability, and intelligence in real-world applications, including collaborative manufacturing, assistive technologies, and interactive service environments.