Tight integration of UWB and INS based on the predicted residual determined robust unscented Kalman filter

Abstract The accuracy and stability of an ultra-wide-band (UWB)-based positioning system would be degraded while suffering from the non-line-of-sight (NLOS) errors and UWB base station distribution significantly. Besides, the linearization of UWB observations could also bring visible linearization errors. To solve these problems, this paper presents a novel tight integration of UWB and the inertial navigation system (INS) (UWB/INS) by integrating the unscented Kalman filter (UKF) and an innovation and residual-based IGG-III robust model, which is to limit the impacts of the linearization errors and the NLOS errors. Meanwhile, different base-station distributions are designed to analyze the impact of different base-station distributions on the UWB-based positioning accuracy. Results from both the field test and simulation test demonstrated that (1) the accuracy of the UKF-based UWB/INS tight integration has been improved by 17.32% in the vertical direction, compared to the UWB/INS tight integration based on extended Kalman filter (EKF); (2) the UWB/INS tight integration based on the predicted-residual-based robust UKF can offer about 69.56%, 86.36%, and 73.13% improvements in the north, east, and vertical directions compared to the solutions of the EKF-based UWB/INS tight integration; (3) the geometric distribution and the number of UWB base-stations affect positioning accuracy of the UWB/INS tight integration. The coverage and strength of the UWB base station should be fully considered when the UWB base station is deployed.