捷联惯导初始对准的超球体采样SRUKF算法

SINS Initial Alignment Based Spherical Simplex Transformation SRUKF Algorithm

平淡卡尔曼滤波(UKF)在捷联惯导系统静基座大方位失准角初始对准中计算量大,且滤波数值不稳定.针对这一问题,该文将超球体采样策略与平方根UKF(SRUKF)算法相结合,提出一种改进的SRUKF算法.该算法在保证其滤波精度和UKF算法相当的前提下,通过引入超球体采样减少了采样点数,提高了计算速度.并以协方差阵的平方根矩阵代替协方差阵参加递推运算,减少了计算机舍入误差,提高了滤波数值稳定性.仿真结果表明,该算法在保证初始对准滤波精度的前提下降低了计算量,提高了滤波性能.

To reduce the amount of calculation and solve the instability problem in the strapdown inertial navigation system（SINS） initial alignment of large azimuth misalignment, this paper proposes a modified square root unscented Kalman filter （SRUKF）. The filter uses the spherical simplex unscented transformation to reduce the number of sigma points to speed calculation without sacrificing the filter precision. To reduce rounding errors and improve stability, the covariance matrix is replaced with a new matrix whose entries are square roots of the covariance matrix. Simulation results show that the spherical simplex SRUKF algorithm can significantly reduce the computation load and provide better filtering performance with accurate alignment.