状态估计中的采样型滤波器

Sampling-Based Filters for State Estimation

基于采样方法的滤波器在现代非线性滤波领域内得到了广泛应用。其共同特点是利用抽样粒子点模拟系统状态的概率分布,从而不受状态先验分布假设(如高斯假设)的约束,拥有更高的滤波精度和更广的应用范围。论文在抽样意义下重新论述了确定性采样滤波器如高斯和型滤波器和UKF,以及基于随机模拟的粒子滤波器,并对这三种滤波器及其衍生方法在状态估计领域(滤波)的应用进行了精度和计算负荷分析。某Benchmark信号处理算例验证了该类方法的估计精度和扩展能力。给出基于实际系统需求一般性评价和选用原则。

Sampling-based filtering approaches are widely used for the on-line estimation problem of non-Gaussian nonlinear systems，They，including GHF and UKF as positive sampling algorithms and particle filter as a stochastic sampling algorithm,commonly take advantages of matching the prior density by special sampling methods.In this paper,theoretical analysis and experimental results on these three filters are proposed to show performance on coping with non-gauss nonlinear filtering problems;moreover ，guidance or suggestions may be adopted on accordance with practical appliance requisitions。