A bioinspired vision-based airflow sensing architecture for flow-guided robotic control

Flow sensing is fundamental to both animals and robots for navigation, stability, and interaction with the environment. While animals achieve this through specialised mechanosensory structures, flow-guided autonomy in robotics remains underdeveloped due to the limitations of existing sensors, which often lack directional sensitivity, robustness, or scalability. To address this gap, we present a bioinspired, vision-based airflow sensing architecture that employs vertically suspended flexible filaments that respond passively to airflow, from which flow information is inferred using onboard computer vision. The sensor's utility was demonstrated on a quadruped robot, where it reliably detected flow speed and direction during both static and locomotion experiments. A mathematical model describing the aerodynamic and gravitational forces acting on the filaments was developed and experimentally validated, showing strong agreement with empirical data. The results highlight the potential of this novel sensing architecture in robotic applications. Beyond robotics, the sensor's low-cost and modular design makes it broadly applicable from scientific research and environmental monitoring to education and public engagement in flow-sensing principles.