A novel UWB/INS tight integration model based on ranging offset calibration and robust cubature Kalman filter

Abstract

Ultra-wideband (UWB) is treated as a competitive positioning technology for indoor location-based applications. However, the non-line-of-sight (NLOS) condition would lead to a UWB solution with low accuracy and poor stability. Such an influence of NLOS errors on UWB positioning accuracy can partially be degraded after integrating UWB range measurements with measurements from an inertial navigation system (INS). Compared with the loose integration based on UWB positioning solutions and the INS-predicted solutions, such a tight integration of UWB/INS has significant advantages in operating with the number of available UWB measurements less than three and detecting and reducing the influence of NLOS errors. However, the accuracy of such a UWB/INS tight integration in the vertical direction is still poor due to the distribution of UWB base stations and UWB ranging offsets. To solve these problems in UWB positioning, this paper proposes a robust and reliable UWB/INS tight integration method based on a ranging offset calibration model and an improved robust cubature Kalman filter (IRCKF), in which the impacts of UWB ranging offsets, NLOS errors, and linearization errors are constrained. According to the results based on a set of experiment data, about 4.3 % and 54.2 % position improvements in horizontal and vertical directions can be obtained by applying ranging offset calibration in UWB positioning mode. Such positioning improvements obtained by applying the IGG-III robust model are about 51.0 % and 2.8 % in horizontal and vertical directions. Furthermore, the CKF offers about 16.0 % and 6.1 % positioning improvements in horizontal and vertical directions, compared to that of EKF. In general, the IRCKF-based UWB/INS tight integration has an accuracy of 0.082 m, 0.841 m, 0.493°, 0.468°, and 5.785° in horizontal position, vertical position, roll angle, pitch angle, and heading angle, respectively, which offers about 65.8 %, 58.9 %, 27.3 %, 26.4 %, and 46.5 % improvements percentages compared to the solutions from the EKF-based UWB/INS tight integration.