Cyber-physical interaction between a drone and deep learning-based perceptual control for automated measurement of structural vibration

Aerial drones offer accessible structural vibration measurements, but manual operation limits efficiency and consistency. This paper presents a drone-based cyber-physical system to enable automated measurement of structural vibration. The physical components are a drone equipped with an optical camera and a LiDAR for sensing, and the cyber component is a machine learning-based framework that integrates deep reinforcement learning, computer vision, environment mapping, and pose adjustment, enabling the drone to automatically access and inspect the structure. Cyber-physical interaction supports perceptual control of the drone for structural inspection. The cyber-physical system has been implemented into a lab-scaled four-story frame model excited by a shake table, and modal analysis has been performed to evaluate the natural frequencies and mode shapes of the frame. Measurement results are comparable with the results from a fixed camera and finite element analysis, revealing satisfied efficacy and efficiency of the cyber-physical system.