Sinusoidal Curve Fitting and Moving Horizon Estimation Based Adaptive Unscented Kalman Filter for Geomagnetic Based Roll Attitude Estimation

Abstract

The classical Unscented Kalman filter (UKF) algorithm will be deteriorated or even divergent while the prior information of the noises statistical characteristics of a nonlinear dynamic system is unknown or inaccurate. In order to avoid these constraints of classical UKF applied in geomagnetic based projectile roll attitude estimation, t a new adaptive UKF strategy by introducing in the sinusoidal curve fitting (SCF) and moving horizon estimation (MHE) is proposed in this paper. This strategy establishes an estimation principle of systematic and observing noise statistics with MHE and SCF combination. Based on the estimated noise statistics, the adaptive UKF with the estimated statistics as input parameters is designed. Consequently, a new optimized technique for geomagnetic based projectile roll attitude estimation is developed. The proposed adaptive UKF solution is able to realize the online computation of the noise statistics and enhance the stability and adaptation to variable application conditions of traditional filters. The advantages of the proposed adaptive UKF strategy are clearly illustrated by projectile roll attitude estimation results in the trajectory simulations.