Feedforward Unscented Kalman Filter for INS/GPS Tightly Coupled Integration System

Abstract

This paper deals with INS/GPS tightly coupled integration algorithm using the feedforward unscented Kalman filter (UKF). In the tightly coupled approach, nonlinear pseudorange measurement models are used to update the Kalman filter. An EKF is usually applied for this task, however it may diverge due to poor functional linearization of the nonlinear measurement in case of the feedforward filter type which corrects the navigation errors without updating the INS. The UKF that approximates a distribution about the mean using a set of calculated sigma points is applied to overcome the flaw of the EKF. We introduce the generalized scaled unscented transformation (GSUT) which modifies the sigma points themselves to reduce these existence boundaries. Then, we apply the GSUT to calculate pseudorange measurements and estimate the navigation errors with the UKF. Real van test and simulation are carried out to estimate the performance of the UKF-based tightly coupled algorithm with the feedforward type.