Bio-inspired fiber-optic-neural network enabled multi-physical sensing for tissue-safe robotic adhesion

Smart perceptive suction cups capable of non-destructively adhering to tissues are in high demand in robot-assisted surgery. However, current surgical suction cups often over-adhere to tissues due to a lack of perception ability, causing tissue damage and posing surgical risks. To address this issue, we propose a bionic suction cup incorporating a multi-physical sensing method for tissue damage monitoring. The system integrates fiber optic receptors with a physical model guided multitask cascade neural network, enabling real-time detection of multiple safety relevant mechanical parameters, such as mechanical compression, contact force, adhesion force, and vacuum level during suction adhesion. These measurements provide essential feedback for dynamic regulation of the adhesion state, thereby preventing tissue damage. We believe the proposed sensing method offers an effective perceptual framework for the development of next generation adhesion based surgical instruments and robotic systems, paving the way for more robust and non-destructive tissue manipulation.