UKF与EKF在卫星姿态估计应用中的比较

Comparison of Unscented Kalman Filtering and Extend Kalman Filtering for Satellite Attitude Estimation

针对卫星的姿态和角速度估计问题,分别给出基于Unscented卡尔曼滤波(UKF)与推广卡尔曼滤波(EKF)的估计算法,并做了相应比较。为了避免欧拉角带来的奇异问题,UKF选用Rodrigues参数而EKF选用四元数参数法来描述姿态误差。考虑卫星的非线性模型,UKF采用Unscented变换而EKF采用线性化方法对姿态误差进行估计。利用陀螺和磁强计的测量信息,UKF和EKF都可得到三轴稳定卫星的姿态估计值,但UKF的收敛速度高于EKF。数值仿真结果表明,当初始姿态存在大偏差时,所给出的UKF的滤波算法性能明显优于EKF。

Two algorithms are presented to solve the satellite attitude and rate estimation problem. One is based on unscented Kalman filter and the other is based on extended Kalman filter. Then a comparison between them is given. To avoid the singular problem from Euler angle, the UKF chooses the Rodrignes parameters and EKF uses the quaternion parameterization to describe the quaternion - error vector. Considering the satellite nonlinear model, un- scented transformation is used in UKF and the linearization technique is used in EKF. Although each of filters could derive the attitude estimations based on gyros and magnetometer measurements , UKF converges more rapidly than EKF. Simulation results indicate that the UKF is more efficient than EKF under large initial attitude error condition.