A GEO-pivoted Adaptive Extended Kalman Filtering Method in Low-latitude Dense Building Environment

Abstract Non-line-of-sight (NLOS) or multipath signals can cause unexpected measurement noise, which will degrade the solution performance of extended Kalman filtering (EKF). Traditionally, a method of covariance matching is used to make the actual residuals consistent with their theoretical covariance. However, this innovation-based adaptive estimation (IAE) of the measurement noise often uses multiple epoch observations, which will be insensitive to the rapidly changing scene. To solve the issue, an improved IAE is proposed, which considers the high elevation angle of geosynchronous earth orbit (GEO) satellites in low-latitude regions. Correlation coefficients between the actual error and single differencing innovations verify the method. The real car tests show that 3-D positioning and velocity determination accuracy are improved from 2.00 m to 1.44 m, and 11.27 cm/s to 7.03 cm/s, respectively. In addition, this method does not store multiple epoch observations, which is convenient for real-time positioning of heavy NLOS and multipath environments.