基于部分扩维的无迹卡尔曼滤波算法研究

UKF扩维算法往往因为比较复杂而导致计算量较大,为此EW等学者提出了当过程噪声和量测噪声都是白噪声时UKF的简化算法。证明了这种简化算法在线性条件下不能达到最优估计;即线性条件下与Kalman滤波不等效;同时证明了UKF部分扩维的算法即只将过程噪声进行扩维后仍能达到最优估计;且又在一定程度上减少了计算量。最后在GPS/INS组合导航中进行了仿真验证,仿真结果进一步证明了理论分析的正确性。

Research on the Part Augmented Filtered Based on UKF

The augmented UKF adds the computational complexity. To reduce the complexity EW proposed a simplified UKF method when the process noises and measurement noises are purely white. However this simplified method has been proven not to be optimal under linear conditions in this paper. In other words it is not equivalent to Kalman filter under linear conditions which means the method is not correct. And at the same time it is also given that the partially augmented UKF which only adds process noised to the state vector could achieve optimal estimation and also reduces the computational complexity to some certain extent. Finally the simulation results in GPS/INS integrated navigation system have further proved the theoretical analysis is correct.