摘要
针对粒子滤波的重要性密度函数选择问题,提出一种基于集合卡尔曼滤波(Ensemble Kalman Filter,EnKF)的改进粒子滤波算法。该方法利用集合卡尔曼滤波产生粒子滤波在每一时刻各粒子的重要性密度函数,在融合最新观测信息的同时,使重要性密度函数更加符合状态的真实后验概率分布。为消除样本枯竭现象,对重采样后的粒子进行马尔科夫链蒙特卡洛处理。在仿真实验中,将新算法用于GPS/DR组合定位系统,与粒子滤波、扩展卡尔曼粒子滤波以及无迹粒子滤波进行比较。仿真结果表明,该算法的估计精度高于传统粒子滤波算法,同时其能够有效控制计算量,并且在粒子数目较少时仍能保证较好的估计性能。
An improved particle filtering algorithm based on the ensemble Kalman filter (EnKF) was proposed in this paper starting with the selection of importance density function of the particle filter. At each time instant, the impor- tance density function is generated by EnKF which fuses the latest observation information and propagates the system states by using a collection of sampled state vectors, called an ensemble. In this way, the importance density function can be very close to the true posterior probability. Furthermore, to avoid the particle impoverishment problem, the Markov Chain Monte Carlo method was introduced after resampling process. In the simulation, the developed filter was com- pared with standard particle filter, extended Kalman particle filter and unscented particle filter in GPS/DR integrated system. The simulation results demonstrate the validity of the developed algorithm. Under the same conditions, the new filter is superior to other particle filtering algorithms with the respect to estimation accuracy, as well as it controls the computational load effectively. It is also found that the new filter can obtain outstanding performance even with a small number of particles.
出处
《计算机科学》
CSCD
北大核心
2016年第9期218-222,共5页
Computer Science
基金
国家自然科学基金(61273291
61305073)
山西省高校科技创新项目(2014104)
山西省回国留学人员科研资助项目(2012-008)资助
关键词
粒子滤波
重要性密度函数
集合卡尔曼滤波
组合定位系统
Particle filter, Importance density function, Ensemble kalman filter, Integrated localization system