摘要
介绍了粒子滤波及随机分布的代表点理论,将统计学中的数论方法应用于粒子滤波,使用随机分布的均方差代表点,对粒子滤波中关键的初始粒子生成、重点密度采样及再采样过程给出了相应的代表点算法,得到了一个包含最少随机操作的、使用非等权值粒子的确定性粒子滤波算法。仿真结果表明,确定性粒子滤波算法不仅是可行的,而且其滤波性能及计算效率均有不同程度的提高。
The principles of the particle filter and the representative-points of random distribution are introduced.The number-theoretic method in statistics is applied to the particle filter, The corresponding rep-points algorithms, which include the generation of initial particles, the sampling of the importance density and re-sampling, is given using the MSE rep-points. Then a deterministic particle filter algorithm, with the fewest random operative and unequally weighted particles, can be built. The simulation results show that not only the algorithm is feasible, but the performance and computational efficiency can be also improved to a certain extent.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2004年第1期72-76,共5页
Journal of National University of Defense Technology
