Human-Inspired Robotic Tactile Perception for Fluid

Abstract

In order to achieve natural tactile sensation and satisfactory perception for robots in complex environments, this article presents a novel human-inspired robotic tactile sensing system for fluid. Specifically, a biomimetic fluid-sensitive handlike sensor (BFHS) is designed by mimicking the perception mechanism of human skin. A deep learning model is then developed to establish a mapping between tactile sensor offset and two key properties of fluid, namely, fluid flow direction and velocity. Furthermore, a feature correlation reconstruction (FCR) method is proposed to derive a new feature to improve the interpretability among features. Finally, an experimental system is constructed to validate the proposed BFHS. The experimental results demonstrate that its average accuracy in sensing fluid direction and velocity significantly outperforming human tactile perception. This can be viewed as a breakthrough finding in the field of robotic tactile perception.