几种滤波算法的分析与比较

Analysis and Comparison of Several Filtering Algorithms

滤波算法常用来解决对系统状态估计的问题,主要有卡尔曼滤波、粒子滤波以及在此基础上改进的扩展卡尔曼,无迹卡尔曼,无迹粒子滤波算法等。对于线性高斯系统模型,卡尔曼滤波有着极强的处理能力,因此得到了广泛的应用。粒子滤波无须对系统状态做线性高斯假设,其应用范围大于卡尔曼滤波,但时间的消耗要远远大于前者。在介绍了常见滤波算法的原理与应用后,通过仿真实验对比了上述几种常见滤波方法的跟踪效果。实验表明,上述算法在非线性高斯模型下均有较好的准确性与较低的误差。

Filtering is used to estimate the system state,including Kalman filter(KF),particle filter and extended kalman(EK),unscented kalman(UK),unscented particle filter algorithm.For linear Gaussian system,Kalman filter has a strong processing ability,so it has been widely used in the field.Linear and Gaussian assumption is not necessary in particle filter contrast to KF,the application range is larger than the latter and the time consumption is stronger than the latter.introducing the principle and application of filtering algorithms,the tracking effect are compared through simulation experiments.Experimental resultsshows that improved algorithm have better accuracy and lower error in the model of nonlinear gaussian.