Real-time remote measurement of distance using ultra-wideband (UWB) sensors.

Distance measurement is significant for ensuring the safety and serviceability of engineering structures. Recently, ultra-wideband (UWB) sensors have offered an alternative real-time remote sensing solution to measure distances. UWB sensors exhibit small size, low cost, low energy consumption, and high robustness to weather conditions, but their ranging accuracy is still limited. This paper presents two machine learning approaches to achieve high accuracy and high frequency, simultaneously. The first approach integrates a convolutional neural network, a long short-term memory module, and a regression module. The second approach integrates two random forest models. These two approaches were implemented into measurement from UWB sensors deployed on a highway bridge and outperformed the state-of-the-art approaches in terms of measurement accuracy and output frequency. The configurations and key parameters of the two approaches were evaluated and improved. This research enhances the capability of measuring distances and deformations for structural health monitoring. [Display omitted] • A real-time remote measurement method of distance using UWB sensors is developed. • A machine learning approach is presented based on a long short-term memory (LSTM) model. • A machine learning approach is presented based on random forest models. • The two machine learning approaches achieve sub-millimeter measurement accuracy. • The upper frequency of distance measurement reaches 100 Hz using the LSTM-based model. [ABSTRACT FROM AUTHOR]