Robotic taste sensing via triboelectric and deep learning

Abstract

The ability to endow robots with a sense of taste is a long-standing challenge in the field of robotics and sensory technology. Traditional chemical measurement devices for liquid identification are often bulky, slow, and costly, limiting their integration into mobile robots. To address this, we propose a novel Liquid Identification Taste System (LITS) that utilizes the triboelectric effect combined with deep learning for precise and efficient liquid identification. By exploiting the unique triboelectric signals generated at the solid-liquid interface, LITS accurately identifies a wide range of liquids based on their physical and chemical properties. Leveraging the Long Short-Term Memory (LSTM) algorithm, the system achieves a high classification accuracy of 96 % for six common liquids. Additionally, the system demonstrates the potential to distinguish inedible liquids, such as HCl and NaOH, further highlighting its versatile application. This integration of triboelectric sensing with deep learning not only enhances the sensor’s resolution and sensitivity but also provides a scalable, cost-effective solution for real-time liquid identification, making it a promising step toward enabling robots to acquire a sense of taste.