Combined UKF/KF for Fast In-motion Attitude Determination of SINS

摘要

In this paper, we propose a combined Unscented Kalman Filter and Kalman Filter (UKF/KF) structured attitude determination method that aims to handle large initial attitude errors on a moving base at much lower computational cost. The proposed algorithm includes modified UKF that reduces initial attitude errors to a small range during coarse alignment and KF to proceed fine alignment that further diminish attitude errors. A Measurement-based Feedback algorithm is proposed to accelerate the convergent speed of UKF. Handover strategy for stable transition from UKF to KF is investigated. Moreover, block matrix multiplication is adopted, which reduces about 10.4% of the computation burden of KF. To improve system observability, position and velocity information from Global Positioning System (GPS), together with magnetic measurements are utilized. Numerical results of real-world simulated trajectory show that the proposed UKF/KF structure outperforms alternative UKF method in convergent speed as well estimation accuracy for both tactical-grade SINS and low-cost MEMS-based SINS.