Using Machine Learning Approaches to Improve Ultra-Wideband Positioning

Abstract

An ultra-wideband (UWB) positioning system consists of at least three anchors and a tag. Via the UWB transceiver mounted on each device in the system, we can use some techniques to obtain the distance between each anchor and the tag. Then we can further realize the tag localization by some classic algorithms. However, in the real environment, the uncertain measurement may bring incorrect distance as well as positioning information. Therefore, in this research, we intend to reconsider the positioning issue by incorporating some machine learning approaches with uncertain measurement in the real environment. Particularly, we utilize the concept of machine learning for overall consideration instead of using a model to evaluate the uncertainty. The experimental results show that our method can be applied to different cases, and some interesting properties in the practical experiments are presented.