摘要
当观测值给定的条件下,为了提取未知参数的信息,文中基于Gibbs与Metropolis-Hasting两种重要采样提出了多尺度粒子滤波算法。该算法在目标状态空间上一条马尔科夫链上,采用不同的粗细尺度进行交替采样来传递目标的状态信息和参数信息,从而搜索目标状态的最大后验分布函数。从而实现对机动目标状态的最优估计。其中细尺度采样保持了估算的精度,粗尺度提高了运算效率。仿真表明,该新算法实现了算法精度和效率的良好折衷。
Based on the given observation data,the multi-scale filtering algorithm was proposed for extracting optimum value with respect to the target state by using different fine and coarse scales in a single chain to explore a maximum posterior likelihood distribution function of the state information.Hence,the state information concerning maneuvering target optimum estimation can be achieved via the novel algorithm based on both Gibbs and Metropolis-Hasting important sampling.The fine scales guaranteed the precision of estimation while the coarse scales enhanced computational efficiency.The simulation shows the fact that the good tradeoff between the estimation accuracy and computational efficiency.
出处
《弹箭与制导学报》
CSCD
北大核心
2011年第4期214-217,共4页
Journal of Projectiles,Rockets,Missiles and Guidance
关键词
多尺度粒子滤波
目标状态信息
最大后验分布函数
重要采样
马尔可夫链
multi-scale particle filtering
target state information
maximum posterior distribution function
important sampling
Markov chain