一种基于传统粗化策略的改进粒子滤波算法
|
摘要
粒子滤波算法是一种重要的处理非线性、非高斯模型的推理算法,针对标准粒子滤波算法中存在的样本贫化问题,提出了一种新的基于传统粗化策略的改进粒子滤波算法。该算法首先利用带有簇类合并的K-means聚类算法对粒子进行聚类,再计算每个簇内状态变量的两个估计值之间的向量差值,将其作为粒子移动的指引方向。仿真结果证明,带有簇类合并的K-means算法的聚类效果优于标准K-means算法聚类效果,而且提出的粒子滤波改进算法加快了带有传统粗化策略的标准粒子滤波算法的收敛速度。
A novel particle filter algorithm based on traditional roughening strategy is proposed in this paper.At first,the algorithm clusters the particles by using K-means algorithm with clusters merging,then the algorithm calculates the vector difference between the two estimates of the state variable in each cluster,and takes them as the direction of the particle movement.Simulation results show that the clustering effect of K-means algorithm with clusters merging is better than that of the standard K-means algorithm,and the novel particle filter algorithm accelerates the convergence speed of standard particle filter algorithm with traditional roughening strategy.
A novel particle filter algorithm based on traditional roughening strategy is proposed in this paper.At first,the algorithm clusters the particles by using K-means algorithm with clusters merging,then the algorithm calculates the vector difference between the two estimates of the state variable in each cluster,and takes them as the direction of the particle movement.Simulation results show that the clustering effect of K-means algorithm with clusters merging is better than that of the standard K-means algorithm,and the novel particle filter algorithm accelerates the convergence speed of standard particle filter algorithm with traditional roughening strategy.
引文
[1]杜正聪,辛强,邓寻.基于权值优化的粒子滤波算法研究[J].重庆师范大学学报(自然科学版),2015(3):124-129
[2]Arnaud Doucet,Simon Godsill,Christophe Andrieu.On Sequential Monte Carlo Sampling Methods for Bayesian Filtering[J].Statistics and Computing,2000,10(3):197-208
[3]李小婷,史健芳.基于重采样技术改进的粒子滤波算法[J].微电子学与计算机,2016,33(9):164-168
[4]Li T,Sun S,Sattar T P,et al.Fight Sample Degeneracy and Impoverishment in Particle Filters:A review of Intelligent Approaches[J].Expert Systems with Applications,2014,41(8):3944-3954
[5]乔端瑞.基于K-means算法及层次聚类算法的研究与应用[D].长春:吉林大学,2016
[6]Davies D L,Bouldin D W.A Cluster Separation Measure[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2009,PAMI-1(2):224-227
[2]Arnaud Doucet,Simon Godsill,Christophe Andrieu.On Sequential Monte Carlo Sampling Methods for Bayesian Filtering[J].Statistics and Computing,2000,10(3):197-208
[3]李小婷,史健芳.基于重采样技术改进的粒子滤波算法[J].微电子学与计算机,2016,33(9):164-168
[4]Li T,Sun S,Sattar T P,et al.Fight Sample Degeneracy and Impoverishment in Particle Filters:A review of Intelligent Approaches[J].Expert Systems with Applications,2014,41(8):3944-3954
[5]乔端瑞.基于K-means算法及层次聚类算法的研究与应用[D].长春:吉林大学,2016
[6]Davies D L,Bouldin D W.A Cluster Separation Measure[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2009,PAMI-1(2):224-227