Multi-sensor fusion for robust indoor localization of industrial UAVs using particle filter
Year: 2024
Type: article
Abstract: Robotic platforms including Unmanned Aerial Vehicles (UAVs) require an accurate and reliable source of position information, especially in indoor environments where GNSS cannot be used. This is typically accomplished by using multiple independent position sensors. This paper presents a UAV position estimation mechanism based on a particle filter, that combines information from visual odometry cameras and visual detection of fiducial markers. The article proposes very compact, lightweight and robust method for indoor localization, that can run with high frequency on the UAV’s onboard computer. The filter is implemented such that it can seamlessly handle sensor failures and disconnections. Moreover, the filter can be extended to include inputs from additional sensors. The implemented approach is validated on data from real-life UAV test flights, where average position error under 0.4 m was achieved.
Authors: Eduard Mráz, Adam Trizuljak, Matej Rajchl, Martin Sedláček, Filip Štec, Jaromír Stanko, Jozef Rodina
Institution: Slovak University of Technology in Bratislava
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
Award: G8139692357
OpenAlex ID: W4401459700