基于非线性混合效应状态空间模型的粒子滤波

Particle Filters Based on the Nonlinear Mixed Effect State Space Models

粒子滤波算法是一种用于解决非线性系统问题的新型算法。通常粒子滤波利用重要性重抽样算法,选用先验分布,但是其易受外部观测值影响,从而导致权重变化较大。为此,文中引入辅助粒子滤波算法进行改进,该算法优势在于前一时刻的样本在抽取时以当前的观测数据为条件,这样得到的样本更加接近真实状态。最后,通过仿真实例,进一步分析验证了辅助粒子滤波算法比采样重要性重抽样更为有效。

Particle filter algorithm is a new algorithm for solving nonlinear system problems. Typically the particle filter uses importance resampling algorithm,which selects a priori distribution. But it is easily affected by external observation,leading to larger changes in weights. This paper introduces an auxiliary particle filter algorithm to improve this. The advantage of this algorithm is that the sample at the previous time is based on the current observational data,thus the sample obtained being closer to the true state. Simulation shows that the auxiliary particle filter algorithm is more effective than the sampling importance resampling.