Fusion localization for indoor airplane inspection using visual inertial odometry and ultrasonic RTLS

Abstract

In this paper, the fusion localization system for the visual inertial odometry (VIO) and ultrasonic real-time localization system (RTLS) for indoor airplane inspection using drones is proposed. In a hangar environment, either trilateration-based RTLS or vision-based localization shows disadvantages and neither can be used alone. In this research, we design a configuration of VIO suitable for hangar environment and outlier filter on ultrasonic RTLS for non-line of sight situations, so that both can be fused using graph optimization. The proposed solution can provide more accurate localization than the visual odometry-only system as well as continue estimating positions in the absence of RTLS data. Localization and real-time performance of the proposed algorithm are evaluated through experimentation in a hangar and a flight test in an outdoor space.