基于自适应门限及分层容量的粒子滤波算法

Particle Filter Algorithm Based on Adaptive Threshold and Hierarchical Capacity

为了解决传统粒子滤波算法在重采样过程中因粒子匮乏而导致滤波精度降低的问题,提出基于自适应门限(T)及分层容量(C)的粒子滤波(particle filtering,PF)算法,简称TCPF算法。利用无迹变换在建议分布中融入最新量测信息,使得建议分布贴近概率密度函数的真实分布;构造自适应门限,减少在高斯混合中对相似组件单元聚类的运算步骤,提高聚类效率;设置分层采样比例容量,对连续概率密度函数进行分层,并对劣势层粒子权值进行优化组合,提高粒子多样性。6组系统模型测试结果表明,与其他滤波算法相比,TCPF的均方根误差均值和标准差均值分别平均下降29%和25%,从而验证了其在抑制粒子退化、避免样本匮乏上的优势,提高了滤波精度。

In order to solve the problem that the traditional particle filter algorithm reduces the filtering accuracy due to the lack of particles in the resampling process,a particle filter(PF)algorithm based on adaptive threshold(T)and layered capacity(C)is proposed.Unscented transformation is used to integrate the latest measurement information into the recommended distribution,so that the recommended distribution is close to the true distribution of the probability density function.The adaptive threshold is constructed to reduce the calculation steps of clustering similar component units in Gaussian mixture,so as to improve the clustering efficiency.The proportional capacity of stratified sampling is set to stratify the continuous probability density function,and the inferior layer particle weights are optimized and combined,in order to increase the particle diversity.The test results of six groups of system models show that,compared with other filtering algorithms,the mean root mean square error and mean standard deviation of TCPF are reduced by an average of 29%and 25%,respectively,which verifies the advantages of the TCPF in suppressing particle degradation and avoiding sample scarcity,and improve the filtering accuracy.