A 600-meter UWB Ranging Dataset with Radio Frequency Features and Deep Learning Applications

This paper presents an open-source Ultra-Wideband (UWB) ranging dataset (LR-UWB), with Radio Frequency (RF) channel features and Channel Impulse Response (CIR) features, challenging the conventional view of UWB as a short-range technology and underscoring its potential in long-range spatial perception. This paper makes the LR-UWB dataset publicly available to support more potential research endeavors and future perception applications, providing a comprehensive resource for community. All observation measurements in the LR-UWB dataset have undergone rigorous error annotation, enabling it to support learning-based study on ranging error compensation, channel state assessment. Moreover, this paper, using this dataset, demonstrates two deep learning applications with advanced outcomes: ranging error compensation and single-anchor mobile positioning. Taking ranging error compensation, improvements of nearly 80% are observed over the raw 85% error lines. For single-anchor mobile positioning, the final accuracies across three sets, each tracking unknown targets over hundreds of meters, achieve approximately 10 cm precision. Notably, this study, adopting forward-thinking academic perspectives, investigates the LR-UWB dataset’s additional functional potential and offers critical guidance for the advancement of UWB in long-range intelligent spatial perception. The key contribution of this paper lies in bridging the gap between short- and long-range applications of UWB by providing scarce data support, advancing research on UWB technology for long-distance, high-precision localization, and laying a foundation for future multi-sensor fusion and intelligent perception. The LR-UWB dataset is publicly available at: https://github.com/futurexlab2012/LR-UWB-Dataset.