Neural Network-assisted Capacitive Sensor for Multi-directional Force Detection

The development of flexible tactile sensors is of great interest for various application areas such as artificial skins, medical surgery and robots. For the requirement of multidirectional force detection, several sensing mechanisms have been proposed for normal and shear force sensing. However, the extraction of force direction and value is still challenging due to the nonlinear relation between sensor output to the force direction. Herein, we develop a microstructures-based capacitive sensor with overlapped sensing units and implement the neural network for both force value and direction sensing. The dielectric layers of two sensing units are designed with different microstructures, i.e., rectangular pyramid and hemisphere microstructures, respectively. Specially, these two sensing units exhibit quite different sensitivity relation to the applied force direction. Therefore, the combination of then could be used for the integrated force value and direction detection by the assistance of neural work training. Such sensing mechanism offers a simple and universal approach for multi-directional force sensing, providing a new insight for tactile sensing design.