A Generalized Maximum Correntropy Extended Kalman Filter for UWB/IMU Localization

The accuracy of ultrawideband (UWB) localization is challenged in complex urban environments due to multipath, non-line-of-sight (NLOS), and sensor placement. The inertial measurement unit (IMU) is unaffected by the environment and NLOS conditions, providing reliable short-term state estimation. In this article, we propose a new extended Kalman filter (EKF) localization framework to solve the noise adaptation problem under typical and severe NLOS conditions. Specifically, we introduce an adaptive density-based spatial clustering of applications with noise (ADBSCAN) algorithm to detect clusters and outliers in the raw UWB ranges. In addition, a new EKF is developed to deal with different NLOS scenarios. First, a generalized maximum correntropy (GMC) with variable kernel bandwidth and shape (VKBS) is proposed based on sliding window NLOS detection. Then, a GMC-EKF integrated with VKBS called GMC-VKBS-EKF is developed for robust UWB/IMU fusion localization. Finally, the performance of the proposed GMC-VKBS-EKF is verified in different NLOS scenarios. The experimental results show that the proposed method is able to suppress the effect of different NLOS noises and achieve higher localization accuracy than the traditional EKF algorithm using maximum correntropy (MC).