Abstract
Avian feather intricate adaptable architecture to wing deformations has catalyzed interest in feathered flapping-wing aircraft with high maneuverability, agility, and stealth. Yet, to mimic avian integrated somatic sensation within stringent weight constraints, remains challenging. Here, we propose an avian-inspired embodied perception approach for biohybrid flapping-wing robots. Our feather-piezoelectric mechanoreceptor leverages feather-based vibration structures and flexible piezoelectric materials to refine and augment mechanoreception via coupled oscillator interactions and robust microstructure adhesion. Utilizing convolutional neural networks with the grey wolf optimizer, we develop tactile perception of airflow velocity and wing flapping frequency proprioception. This method also senses pitch angle via airflow direction and detects wing morphology through feather collisions. Our low-weight, accurate perception of flapping-wing robot flight states is validated by motion capture. This investigation constructs a biomechanically integrated embodied perception system in flapping-wing robots, which holds significant promise in reflex-based control of complex flight maneuvers and natural bird flight surveillance.