A Multilevel Optimised Algorithm for UWB Positioning in Indoor Environment

Abstract

The proliferation of the Internet of Things has precipitated an escalating demand for indoor positioning and navigation systems that exhibit a confluence of heightened precision and economic viability. However, non-line-of-sight has an impact on the accuracy of ultra-wideband indoor location. To address this issue, we proposed a multilevel optimised algorithm based on particle filter and Bayesian unscented Kalman filter (PF-BUKF) to approach the nonlinear state and then achieve accurate three-dimensional position estimation. This approach comprises two stages. Firstly, the PF is utilised to determine the tag's coordinate's state vector and covariance as the initial optimised values. Then, the results are employed as the prior information for BUKF in order to anticipate the state of tag. The process of two steps utilises discrete points to approach the true state, which enhances the robustness and accuracy of the positioning system. Furthermore, we investigated the effect of time step size on the precision of positioning. Experimental results reveal a substantial improvement over traditional positioning methods, with mean absolute error and root mean square error values of 8.84 and 2.70 cm, respectively, as opposed to 19.02 and 8.45 cm using conventional algorithms in a nonlinear system.