Generalized Incremental Predictive Filter for Integrated Navigation System INS/GPS in Tangent Frame

Abstract

The extended Kalman filter (EKF) is a widely used algorithm for nonlinear estimation of Inertial Navigation Systems (INS) and the Global Position System (GPS) integration.However, EKF has several limitations, such as linearization dependency, and the model error statistics are assumed as a zero-mean Gaussian noise with known covariance.Consequently, if EKF is not tuned correctly, the INS error predictions can quickly diverge.To overcome the limitations of existing Kalman algorithms, this paper derives a real-time predictive approach.The proposed method increases the accuracy and the reliability requirements of loosely INS/GPS integration by estimating the unknown model errors of sensors without augmenting the state space.Also, considering the insufficiency of the researches on the integrated navigation in tangent (launch) frame, this research derives the navigation equations in tangent frame and its error model is analyzed.The estimation performance of the predictive approach is analyzed.The performance is verified using an experimental data acquired from a land-vehicle test.The results of predictive filter demonstrate superior performance to the traditional EKF.The test results of land-vehicle navigation validate the advantages of the presented method, which increases the position accuracy by an amount of 70% and it decreases the computational cost to 50% and improves the estimation performance for the integrated INS/GPS better than the traditional EKF.This test is a fundamental step to determine the capability of the filter for robotics and aerospace applications in future.