UAS based autonomous visual inspection of airplane surface defects

Abstract

Today’s maintenance tasks are time-consuming and therefore cost-intensive, in particular the manual inspection of commercial airliners. The joint project ”AI-Inspection Drone” aims to provide a complete process chain for the surface damage detection of airliners based on an unmanned aerial system (UAS). To achieve this goal, visual data is gathered, which can later be evaluated by artificial intelligence models. The process chain is explained in detail, beginning with the simulation of the hangar environment, followed by insights about the indoor navigation of the UAS. Finally, it is validated through test flights around the airplane, respecting strict security and safety requirements. The 3D-simulation based on Gazebo was expanded with hardware-in-the-loop testing functionality by utilizing a camera-based motion capture system to track the UAS’s position in real-time and feed the position data back into the simulation, to test different inspection tasks. For the deployment of the UAS in the hangar, a 3D-LiDAR based SLAM algorithm is used to provide position and orientation data in relation to the airplane. Using a 3D model, which can be gathered beforehand with LiDAR scans, the airplane’s surface area is estimated to determine the mission waypoints and the corresponding inspection views for a high-resolution camera. A path planning algorithm controls the procedure of the inspection by evaluating an efficient path based on these waypoints and enables obstacle avoidance based on LiDAR data. With the proposed autonomous aerial inspection platform, the ground time of airplanes can be reduced, thus increasing the efficiency of the airplane inspection process.