Development and Characterization of an AI-Driven Textile-Based Sensor for Tactile Sensing in Soft Robotics

Soft collaborative robots are increasingly used to enhance human capabilities in industrial and service settings. However, ensuring safety, autonomy, and user acceptance requires the development of compliant and adaptable sensing technologies. Textile-based sensors have emerged as a cost-effective solution for mechanical sensing, particularly in wearable applications, and offer a promising method for integrating tactile sensing into soft collaborative robots. However, e-textiles face challenges such as hysteresis, deformation, and drift of sensitivity over time, limiting their application to research demonstrators. This study investigates the electromechanical properties of a commercially available conductive textile to assess its suitability for pressure and texture sensing. The selected conductive thread was chosen for its low cost and its use in various studies. To evaluate its electromechanical behaviour, the thread was subjected to controlled mechanical stretching, performing static and hysteresis analyses while monitoring resistance, elongation, and applied force. These findings informed the design of a flexible AI-driven pressure and texture sensor, consisting of an orthogonal grid of embroidered conductive threads on a scuba textile substrate, held in place by a 3D-printed frame and connectors. The fabricated sensor was integrated with an AI classification system for object geometry recognition, enabling tactile sensing for robotic grippers. After systematic data acquisition, the sensor achieved over 85% accuracy in AI-driven object shape classification. Its adaptability, low cost, and flexibility make it suitable for both robotic tactile sensing and wearable applications.