Adaptive Maximum Correntropy Unscented Kalman Filter Based on IMU and UWB Data

Abstract

Ultra-wideband (UWB) systems are often impacted by non-Gaussian time-varying noise in indoor positioning applications because of non-line-of-sight (NLOS) and multipath impacts. In this paper, a UWB and Inertial Measurement Unit (IMU) tightly coupled fusion structure is built to eliminate the IMU accumulated error and to enhance the dynamic response of localization. To complete the data fusion, an adaptive maximum correntropy unscented Kalman filter (AMCUKF) is suggested. On the one hand, the AMCUKF incorporates the maximum correntropy criterion to suppress the non-Gaussian noise (NGN). On the other hand, by modifying the traditional Sage-Husa estimator, the effect of NGN is further reduced, and the localization accuracy and robustness are improved. Finally, simulations and hardware experiments were used to demonstrate the algorithm effectiveness, which can perform highaccuracy localization in complex environments.