Robust and Efficient CIR-Based Outlier Rejection for Tightly Coupled UWB–ODO–Inertial Navigation

Abstract

The ultrawideband (UWB) technology has been proven to achieve continuous absolute positioning for wheeled robots. However, its accuracy is often degraded by non-line-of-sight (NLOS) propagation in complex environments. In this article, a channel impulse response (CIR)-based method is proposed to improve the NLOS outlier rejection performance and the positioning accuracy. The proposed method is integrated into a tightly coupled UWB–odometer (ODO)–inertial navigation system, named CIR-UOINS, as an application instance. Multiple CIR feature parameters are constructed, and feature selection is performed by using probability density distributions, the Bhattacharyya distance, and the correlation analysis. Based on the selected CIR feature, a likelihood ratio test is employed to identify NLOS outliers directly at the signal level. Experiments are conducted in a semi-indoor rooftop with an area of about 1000 m2. Sufficient results demonstrate that the proposed CIR-UOINS achieves a positioning accuracy of about 0.19 m in the complex scenario, which is comparable to that of the state-of-the-art (SOTA) random sample consensus (RANSAC)-based method. Moreover, the outlier rejection time is reduced by 96.2 % compared to the SOTA RANSAC-based method. Hence, CIR-UOINS achieves a better balance between the positioning accuracy and the computational efficiency.