TUM Flyers: Vision—Based MAV Navigation for Systematic Inspection of Structures

Abstract

A large amount of buildings and structures in Europe’s industry and infrastructure require regular inspection and maintenance. Inspection nowadays often is a costly and time-consuming process. Inspection teams need to climb through the structures, or special heavy equipment such as access lifts need to be employed. Current approaches using micro aerial vehicles (MAVs) rely on GPS for autonomous navigation and require highly skilled and well trained pilots to maneuver the MAV close to structures. It is often difficult for the team to perform a full and systematic inspection, which may cause deficiencies to be overlooked. Finally, the inspection team typically has no accurate position information to exactly reference deficiencies, such that comparisons across multiple inspections are difficult to achieve. The aim of our project is to develop novel vision-based localization, 3D reconstruction, and navigation technologies for increasing the level of autonomy of MAV inspection systems and the systematic quality of inspections. The inspection with MAVs of a large structure such as a bridge is a highly challenging real use-case for which we have developed a feasibility demonstration in this project. Our scenario combines a variety of open problems for the vision-based localization, mapping and autonomous navigation for MAVs in large-scale outdoor scenarios, which are shared with many other tasks like the inspection of buildings, industrial plants, wind power plants, or oil platforms. The technologies pursued in this project target at further automating and simplifying the use of MAVs in GPS-restricted or GPS-denied environments. These technologies have the potential to increase cost effectiveness and the quality of inspection and maintenance operations.