一种改进的自适应优化粒子滤波算法研究
|
摘要
粒子滤波算法在非线性滤波领域受到广泛关注,但是该算法存在样本退化问题.为了改进粒子滤波算法的性能,这里结合自适应优化机制对粒子滤波算法的建议分布选择机制及重采样技术进行改进.对于粒子滤波的建议分布选择,提出一种基于自适应退火参数优化的混合建议分布方法.通过混合建议分布不足的分析,利用退火参数来优化控制状态转移先验分布函数和观测似然函数之间的比例,同时,基于自适应参数优化机制来动态调整退火参数的值.对于粒子滤波的重采样,提出了基于部分分层重采样优化算法的自适应重采样技术.通过有效样本大小的评估来执行自适应重采样策略,此外,基于部分分层重采样算法,利用权重优化的思想对其重采样前后权重计算的方法进行优化.通过相关算法的性能比较,所提改进粒子滤波算法的有效性得以验证.
Although it has attracted widespread attentions in the nonlinear filtering field,particle filter algorithm exists the sample degradation problem.In order to improve the algorithm performance,proposal distribution choice and resampling technique of particle filter algorithm are improved by the adaptive optimization mechanism.For proposal distribution choice,it proposed a method of hybrid proposal distribution based on adaptive annealing parameter optimization.With the deficient analysis of hybrid proposal distribution,annealing parameter factor is utilized to adjust the mix ratio of state transition prior distribution and likelihood proposal distribution in hybrid proposal distribution.At the same time,adaptive parameter optimization mechanism is integrated into above hybrid proposal distribution to dynamically adjust the annealing parameter.For resampling technique,it presented an adaptive resampling based on an optimized partial stratified resampling(PSR) algorithm.Effect sample size is estimated so as to implement adaptive resampling.In addition,combined with PSR algorithm,weight computation before and after resampling algorithm is optimized based on weight optimization idea.By the comparison with other algorithms,the effectiveness of improved particle filter algorithm introduced in this paper is demonstrated.
Although it has attracted widespread attentions in the nonlinear filtering field,particle filter algorithm exists the sample degradation problem.In order to improve the algorithm performance,proposal distribution choice and resampling technique of particle filter algorithm are improved by the adaptive optimization mechanism.For proposal distribution choice,it proposed a method of hybrid proposal distribution based on adaptive annealing parameter optimization.With the deficient analysis of hybrid proposal distribution,annealing parameter factor is utilized to adjust the mix ratio of state transition prior distribution and likelihood proposal distribution in hybrid proposal distribution.At the same time,adaptive parameter optimization mechanism is integrated into above hybrid proposal distribution to dynamically adjust the annealing parameter.For resampling technique,it presented an adaptive resampling based on an optimized partial stratified resampling(PSR) algorithm.Effect sample size is estimated so as to implement adaptive resampling.In addition,combined with PSR algorithm,weight computation before and after resampling algorithm is optimized based on weight optimization idea.By the comparison with other algorithms,the effectiveness of improved particle filter algorithm introduced in this paper is demonstrated.
引文
[1]Gordon N J,Salmond D J,Smith A F M.Novel approach to nonlin-ear/non-gaussian bayesian state estimation[J].IEEE Proceedingson Radar and Signal Processing,1993,140(2):107-113.
[2]Thrun S,Fox D,Burgard W,et al.Robust monte carlo iocalizationfor mobile robots[J].Artificial Intelligence,2001,128(1-2):99-141.
[3]Li Di-ping,Zou Bei-ji,Fu Zi-gang.New approach to tracking objectmoving rapidly[J].Journal of Chinese Computer Systems,2011,32(8):1665-1668.
[4]Doucet A,Godsill S J,Andrieu C.On sequential monte carlo sam-pling methods for bayesian filtering[J].Statistics and Computing,2000,10(3):197-208.
[5]Arulampalam M S,Maskell S,Gordon N,et al.A tutorial on particlefilters for online nonlinear/non-gaussian bayesian tracking[J].IEEE Transactions on Signal Processing,2002,50(20):174-188.
[6]Hu Shi-qiang,Jing Zhong-liang.Principles of particle filter and itsapplication[M].Beijing:Science Press,2010.
[7]Yu Jin-xia,Tang Yong-li,Chen Xian-cha,et al.Choice mechanismof proposal distribution in particle filter[C].Proceedings of theEightth World Congress on Intelligent Control and Automation(WCICA2010),Jinan,China:IEEE Press,2010,7:1051-1056.
[8]Douc R,Cappe O.Comparison of resampling schemes for particlefiltering[C].Proceedings of Image and Signal Processing and A-nalysis,Zagreb,Croatia:IEEE Press,2005:64-69.
[9]Bolic M,Djuric P,Hong Sangjin.New resampling algorithms forparticle filters[C].Proceedings of the 2003 International Confer-ence on Acoustics,Speech,and Signal Processing,HongKong,Chi-na:IEEE Press,2003:589-592.
[10]Chen Jian,Yan Ping,Zhang Jing-yuan.Research on weight optimalcombination particle filter algorithm[J].Computer Engineering andApplications,2009,45(24):33-35,39.
[3]李弟平,邹北骥,傅自钢.结合粒子滤波与Mean shift的高速运动目标跟踪[J].小型微型计算机系统,2011,32(8):1665-1668.
[6]胡士强,敬忠良.粒子滤波原理及其应用[M].北京:科学出版社,2010.
[10]谌剑,严平,张静远.权值优化组合粒子滤波算法研究[J].计算机工程与应用,2009,45(24):33-35,39.
[2]Thrun S,Fox D,Burgard W,et al.Robust monte carlo iocalizationfor mobile robots[J].Artificial Intelligence,2001,128(1-2):99-141.
[3]Li Di-ping,Zou Bei-ji,Fu Zi-gang.New approach to tracking objectmoving rapidly[J].Journal of Chinese Computer Systems,2011,32(8):1665-1668.
[4]Doucet A,Godsill S J,Andrieu C.On sequential monte carlo sam-pling methods for bayesian filtering[J].Statistics and Computing,2000,10(3):197-208.
[5]Arulampalam M S,Maskell S,Gordon N,et al.A tutorial on particlefilters for online nonlinear/non-gaussian bayesian tracking[J].IEEE Transactions on Signal Processing,2002,50(20):174-188.
[6]Hu Shi-qiang,Jing Zhong-liang.Principles of particle filter and itsapplication[M].Beijing:Science Press,2010.
[7]Yu Jin-xia,Tang Yong-li,Chen Xian-cha,et al.Choice mechanismof proposal distribution in particle filter[C].Proceedings of theEightth World Congress on Intelligent Control and Automation(WCICA2010),Jinan,China:IEEE Press,2010,7:1051-1056.
[8]Douc R,Cappe O.Comparison of resampling schemes for particlefiltering[C].Proceedings of Image and Signal Processing and A-nalysis,Zagreb,Croatia:IEEE Press,2005:64-69.
[9]Bolic M,Djuric P,Hong Sangjin.New resampling algorithms forparticle filters[C].Proceedings of the 2003 International Confer-ence on Acoustics,Speech,and Signal Processing,HongKong,Chi-na:IEEE Press,2003:589-592.
[10]Chen Jian,Yan Ping,Zhang Jing-yuan.Research on weight optimalcombination particle filter algorithm[J].Computer Engineering andApplications,2009,45(24):33-35,39.
[3]李弟平,邹北骥,傅自钢.结合粒子滤波与Mean shift的高速运动目标跟踪[J].小型微型计算机系统,2011,32(8):1665-1668.
[6]胡士强,敬忠良.粒子滤波原理及其应用[M].北京:科学出版社,2010.
[10]谌剑,严平,张静远.权值优化组合粒子滤波算法研究[J].计算机工程与应用,2009,45(24):33-35,39.