摘要
针对多分辨分析方法用于降低粒子滤波样本集容量时,因多次迭代引起的样本数急剧下降而导致滤波性能不稳定的问题,提出一种通过监测拟测量误差相关统计量来预警可能出现的误差扩大,当系统处于性能临界区时,在原样本空间上进行准蒙特卡罗增量采样或复制原样本集来控制样本集容量,规避滤波发散的风险.仿真实验表明该算法保持了粒子滤波算法的估计性能,同时有效降低了粒子滤波样本数目,提高了计算效率.
When the multiresolutional analysis method is applied to reduce the number of particles in particle filter,it may cause the drop of filtering accuracy.Aiming at this problem,a size control-based multiresolutional particle filter is proposed.The method predicts the possible enlarging of error by observing the quasi-measurement error probability density,and then increases the size of sample set when the filtering is becoming unsteady by quasi-Monte Carlo sampling or multiplying the original particle set to avoid the possible divergence.The simulation results show that the algorithm maintains the estimation performance of particle filter,meantime also reduces the number of particles and improves the computation efficiency.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2010年第11期2664-2668,共5页
Acta Electronica Sinica
基金
国家自然科学基金(No.60773067)
关键词
粒子滤波
多分辨分析
小波变换
拟测量误差
准蒙特卡罗采样
particle filter
multiresolutional analysis
wavelet transform
quasi measurement error
quasi-Monte Carlo sampling
