针对无迹卡尔曼滤波(UKF)计算量大的问题,提出一种加性超球体平方根UKF算法,即ASSRUKF算法。该算法通过引入加性非扩展形式减少状态维数,并采用超球体单形采样减少采样点的数量,有效降低了算法计算量;同时采用协方差阵的平方根代替协方差阵参加递推运算,以提高滤波算法的计算效率和数值稳定性。建立了加性噪声下基于微机械惯性测量单元和磁强计的无人机姿态模型,并采用ASSRUKF算法进行姿态估计。仿真结果表明,本算法的精度与UKF相当,而执行时间仅为UKF的36.8%,有效降低了算法的计算复杂度。
Aiming at the overload computational complexity in an unscented Kalman filter (UKF) , an additive spherical simplex square root UKF (ASSRUKF) was proposed. To decrease the computational complexity, the algorithm for the proposed filter, called the ASSRUKF algorithm, used an additive non-augmented unscented transform and a spherical simplex sampling to reduce the state dimension and the number of sigma points, respectively. Meantime, the covariance matrix was replaced with a new matrix whose entries were square roots of the covariance matrix in the process of estimation to ensure the efficiency and stability of the filter. Under the condition of additive noise, the proposed algorithm was applied to the attitude determination model of an unmanned aerial vehicle ( UAV), which combined a MEMS inertial measurement unit and a magnetometer. The simulation results showed that the estimation precision of the proposed algorithm was similar to the standard UKF, while the computational time was only 36.8% of the UKF, which effectively reduced the computational complexity.