基于微分几何的两种曲率——参数影响曲率和固有曲率,给出了定量描述非线性滤波问题的非线性强度的方法,分别采用扩展Kalman滤波方法和Unscented Kalman滤波方法进行了模拟实验。结果验证了这些曲率确实能够度量非线性滤波问题的非线性强度,且能够评估非线性滤波算法的状态估计性能。
A quantitative measure approach to the degree of nonlinearity of the nonlinear filtering problems using the differential geometry based measures of nonlinearity such as parameter-effect curvature and intrinsic curvature is given.The simulation test using the extended Kalman filtering and the unscented Kalman filtering is implemented.The results verify that these curvatures indeed can measure the degree of the nonlinearity of the nonlinear filtering problem,and simultaneously can tell the state estimation performance of the nonlinear filtering algorithms.