Streamlined Indoor UAVs Localization Using a Dense and Size-Heterogeneous Tags Map

A significant challenge in the employment of UAV platforms for indoor inspection and maintenance operations lies in the problem of finding a portable and cost-effective way to accurately localize aerial vehicles in GNSS-denied environments. Focusing on the visual-based positioning paradigm, we outline a pose estimation procedure whose accuracy is achieved by leveraging the potential offered by a dense and size-heterogeneous map of tags. The proposed indoor UAVs localization rests on i) hierarchical tag selection, ii) outlier removal, and iii) multi-tag estimation averaging, to facilitate visual-inertial reconciliation. We assess the performance of the outlined positioning system through ad-hoc experimental tests that highlight the localization accuracy improvement as compared with other existing state-of-the-art solutions.