Machine-Learning-Assisted Recognition on Bioinspired Soft Sensor Arrays

Soft interfaces with self-sensing capabilities play an essential role in environment awareness and reaction. The growing overlap between materials and sensory systems has created a myriad of challenges for sensor integration, including the design of a multimodal sensory, simplified system design capable of high spatiotemporal sensing resolution and efficient processing methods. Here we report a bioinspired soft sensor array (BOSSA) that integrates pressure and material sensing capabilities based on the triboelectric effect. Cascaded row + column electrodes embedded in low-modulus porous silicone rubber allow rich information to be captured from the environment and further analyzed by data-driven algorithms (multilayer perceptrons) to extract higher level features. BOSSA demonstrates the ability to identify 10 users (98.9%) and identify the placement or extraction of 10 objects (98.6%). Moreover, its scalable fabrication facilitates large-area sensor arrays with high spatiotemporal resolution and multimodal sensing abilities yet with a less complex system architecture. These features may be promising in the development of immersive sensing networks for intelligent monitoring and stimuli response in smart home/industry applications.