An Autonomous Aircraft Inspection System using Collaborative Unmanned Aerial Vehicles

The maturity in autonomous navigation of Unmanned Aerial Vehicles (UAVs) provides the possibility to deploy UAVs for different kinds of inspection jobs. Aircraft inspection is a well-known periodic process in aviation history, which is long, costly, and subjective. Manual inspection usually takes several hours, where multiple sensors on the aircraft's outer surface, dents, lightning strikes, paint, etc. are mainly checked. The main advantage of a UAV-based inspection is to minimize the turnaround time, which reduces the cost. Deployment of multiple collaborative UAVs is a must because most of the off-the-shelf UAVs do not have endurance for more than 30 minutes. There exist multiple challenges, e.g., safety of the multi-million dollar aircraft while multiple UAVs navigate, accurate identification of defects in the millimeter range, accurate localization of the defect on the aircraft body surface, etc. Moreover, the solution should be independent of the aircraft model and scalable for any model with minimal effort. To this end, we present a visual inspection system for aircraft using collaborative autonomous UAVs that is auto-adaptable to any aircraft model with minimal manual intervention. The system allows each UAV to start from any random location in the vicinity of the aircraft and uses a novel registration algorithm to collaborate between them and navigate using LiDAR-Inertial measurements. The navigation algorithm creates multilayered zones for safe navigation and to avoid obstacles. The system uses a low-cost RGB-D camera for inspection and detects defects. To the best of our knowledge, there is no open data for such a task, and given the limitations in creating it with a real aircraft, we evaluate our proposed system in a Gazebo simulation with an anonymous aircraft model. The proposed approach is able to complete the inspection task within 10 minutes using two UAVs.