Electrical Impedance Tomography Based Finger-Shaped Soft Artificial Skin

Obtaining dense contact information for feedback control is vital for robotic manipulation. However, existing tactile sensing technologies have a large footprint, making seamless integration with current robotic devices difficult. This paper presents a novel multi-layer Electrical Impedance Tomography (EIT) sensor architecture designed to provide distributed, high-density tactile sensing with a small form factor. Using our multilayer structure, we address a common issue in other EIT-based tactile skins that prevents electrodes from being placed distally from the sensing surface. Our innovative multi-layer design enables the development of complex-shaped soft sensing skins without any electronic components on the sensing surface, achieving very high accuracy. To demonstrate practical applications, we fabricated a finger-shaped three-dimensional (3D) skin and conducted experiments to collect real-world data. We developed a perception model for the tactile sensor by employing data-driven machine learning methods to predict press localization and force with high accuracy based on EIT signals. Our work presents a significant step towards developing whole body full soft tactile sensors with a small form factor.