QUKF算法及其在SINS初始对准中的应用  

Quaternion unscented Kalman filter and its application to initial alignment of SINS

作　　者：周卫东[1] 乔相伟[1] 任蕾[1] 吉宇人[1] 

机构地区：[1]哈尔滨工程大学自动化学院,哈尔滨150001

出　　处：《仪器仪表学报》2010年第2期264-269,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60674087)资助项目

摘　　要：针对捷联惯导系统初始对准过程中的大失准角情况,建立了乘性四元数姿态误差方程。结合UKF算法,提出了基于四元数的UKF算法(QUKF)。以姿态矩阵为对象,通过构造姿态矩阵代价函数的方法来求取四元数的一步预测均值,保证了均值四元数的规范化;利用乘性误差四元数表示四元数Sigma点与均值四元数的距离,以求取四元数的预测协方差矩阵。四元数状态更新中,采用乘性四元数更新保证了滤波过程中四元数的规范化。基于该算法的SINS在粗对准水平失准角为小角度、方位失准角为大角度条件下的仿真实验结果表明,与常规EKF相比,纵、横摇角对准精度均略有提高,而航向角误差估计精度提高显著。Aiming at the large misalignment angle of strap-down inertial navigation system（SINS） in initial alignment,a multiplicative quaternion attitude error equation is established.Combined with the UKF algorithm,a quaternion UKF algorithm（QUKF） is proposed.In order to ensure normalization of mean quaternion,an attitude matrix cost function is derived to compute the predicted mean quaternion;the multiplicative quaternion error,which represents the distance between quaternion Sigma points and the predicted mean quaternion,is used to obtain the predicted covariance matrix of the quaternion.The quaternion updates are performed using quaternion multiplication,which guarantees the quaternion normalization in the filtering process.Simulations were performed under the condition that the horizontal misalignment angle was small and the azimuth misalignment angle was large.Simulation results show that compared with normal EKF,the proposed algorithm improves the alignment precision of both roll and pitch angles slightly and enhances the estimate precision of yaw angle significantly.

关 键 词：捷联惯导系统 初始对准 四元数无迹卡尔曼滤波 乘性误差四元数 

分 类 号：U666.12[交通运输工程—船舶及航道工程]