Recently, Associate Professor Zhimiao Yan and Professor Benlong Wang's team from the Department of Engineering Mechanics at Shanghai Jiao Tong University published a paper titled "Avian-inspired embodied perception in biohybrid flapping-wing robotics" in Nature Communications. The research team developed a bird-inspired biohybrid flapping-wing robot with embodied perception technology, utilizing the vibrational structures of feathers and flexible piezoelectric materials as mechanoreceptors. Through deep learning, the system achieves the recognition of multiple flight parameters, including flapping frequency, wind speed, and pitch angle, which was validated through experiments involving untethered flight. The team has established a bio-material and smart-material integrated embodied perception system within flapping-wing robots, holding significant potential for feedback-based complex flight maneuver control and monitoring of natural avian flight.

The remarkable agility of birds in flight is largely attributed to their multi-degree-of-freedom wing structures. By modulating wing movements, birds can execute complex aerial behaviors such as efficient cruising, agile turns, and rapid dives. In addition to the flexible skeletal structure of their wings, the unique arrangement of feathers on the wing surface plays a crucial role in avian flight. The high adaptability between avian feathers and wing deformation behaviors has sparked interest in developing flapping-wing robotics that exhibit high maneuverability, agility, and stealth. However, simulating avian flight perception remains challenging due to stringent weight constraints.

Embodied perception plays a vital role in controlling and regulating the trunk and limbs of biological entities. As components of embodied perception, both tactile and proprioceptive feedback are essential. Tactile perception enables organisms to sense external pressure, vibrations, and variations in temperature and humidity, while proprioception provides information regarding body movement and position. These two sensory modalities, distinct yet intricately interrelated, work in concert to regulate an organism's perceptual system, inspiring robotic sensing designs. Similar to mechanoreceptors in vertebrate skin that detect surface pressure and vibrations, PVDF piezoelectric materials exhibit rapid and robust responses to pressure and vibrations. The Feather-PVDF biohybrid mechanoreceptor serves as a sensory medium for the wings of the flying robot, functioning analogously to sensory nerves by providing feedback on sensation and movement.

This work leverages the biological materials and structures of natural feathers to enhance the differentiation of mechanical perception, integrating lightweight PVDF films to simulate the functions of mechanoreceptors found in avian wings. The team investigated the heterogeneous interfacial connection characteristics of the Feather-PVDF biohybrid mechanoreceptors, including adhesion strength, fatigue durability, and electromechanical properties. Based on the bird-inspired flapping-wing robot (FWR), they established a biohybrid perception system that collects corresponding voltage signals under varying conditions such as flapping frequency, wind speed, pitch angle, and wing shape. Using a sliding window approach for data segmentation, they employed convolutional neural networks to achieve intelligent recognition of various flight parameters and validated the accuracy of recognition during temporally varying flapping behaviors.

To further validate the feasibility of the biohybrid embodied perception method in real flight environments, the team developed a flapping-wing robot capable of untethered flight. This robot comprises a gear-driven mechanism, flight control board, piezoelectric signal acquisition, high-impedance wireless transmission module, and flapping wings, with a total mass of 28.465 g, of which the PVDF component constitutes only 0.79%. They achieved the perception of flapping frequency, relative flow speed, and pitch angle during untethered flight, demonstrating effective identification of flight parameters in indoor environments. This biohybrid perception design contributes to the monitoring and feedback control of flight states in miniature flapping vehicles, offering new insights for the development of lighter, more integrated, and stealthier biomimetic flapping-wing robotics.

The first author of the paper, Qian Li, is a master's student in the Department of Engineering Mechanics at Shanghai Jiao Tong University, which is the sole affiliated institution for the paper.

This team has been invited by the editor to write one blog to introduce interesting points on Nature Research Communities. See the blog address for details：

https://communities.springernature.com/posts/nature-s-flight-sense-the-bio-mechanical-perception-with-embodied-ai-insights