Abstract
Special approaches are required for integration of global navigation satellite system (GNSS) with a strap-down inertial navigation system (INS), particularly based on low-cost micro-electro mechanical system (MEMS)-grade inertial sensors. The proposed approach should be computationally efficient, mathematically nonsingular, and executable on small digital signal processor (DSP) modules. This paper presents a new INS/GNSS navigation system based on a direct decentralized integration scheme. Based on the proposed QR-factorized cubature Kalman filter (CKF) structure, two cascade filters are implemented for separate estimation of the orientation attitude-heading angles and the 3-D position/velocity components. Owing to the QR-factorization, the numerical errors in the update process of estimation covariance matrix are removed. Considering the nonlinear dynamics of the strap-down INS as well as the large uncertainties included in stochastic model of the MEMS-grade inertial sensors, the QR-factorized CKF yields enhanced accuracy and reliability compared with the pure Kalman filter. The decentralized integration scheme provides the separate estimation of orientation components from the position and velocity vectors. Therefore, the propagation of the position and velocity estimation errors into the orientation filter section does not occur. The performance of the presented system is assessed through real data of vehicular field tests.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call