MD-RadioMap: Multi-Drone Radio Map Building via Single-Anchor Ultra-Wideband Localization Network

摘要

Drones are very promising for the future of logistic applications, which require high communication and localization capabilities. A radio map is vital for drones since it describes the cellular signal quality between drones and ground base stations as well as the Global Navigation Satellite System (GNSS) positioning accuracy. Currently, radio maps have been already utilized to guide the design of drone delivery routes, but the radio map building process relies on manual detection, which is costly and unsuitable for large-scale intensive data gathering. Therefore, we introduce the use of drones for autonomous map construction. Firstly, a pipeline for the multi-drone radio map building task is presented, in which we introduce a novel single-anchor ultra-wideband (UWB) localization network. Then we propose a multi-drone coverage path planning algorithm that ensures the line-of-sight connectivity between drones during radio data collection. Using Unreal Engine 4, we demonstrate our proposed pipeline in a realistic low-altitude urban environment. Experimental results show that our planning algorithm is superior in terms of mission completion time and an application of the radio map in delivery routes design is presented. Moreover, we deploy the UWB-based localization network on real robots and it achieves decimeter-level localization accuracy, proving its potential for use on drones.