摘要
针对粒子滤波算法存在的粒子退化现象和重要性密度函数难以选取等问题,在研究交互式多模型滤波算法的基础上,设计一种基于随机加权自适应IMMUPF算法。首先,该算法在无迹粒子滤波的采样过程中融合了随机加权和交互式多模型滤波的优点,利用无迹卡尔曼滤波算法得到[k]时刻各模型估计最新量测信息的粒子;然后,对该组粒子进行输入交互作为各模型的输入,再经过模型匹配、重采样以及模型概率更新过程;最后,对各模型相对应的粒子进行输出交互,得到所有粒子的随机加权自适应和的表达式,循环更新粒子实现状态估计。将设计的算法应用于GPS/DR组合导航系统中进行仿真计算,结果表明,该算法计算得到的位置误差较UPF和IMMUPF有所减少,东向位置误差控制在[-8m,+6m],北向位置误差控制在[-8m,+8m],提高了GPS/DR组合导航系统定位的解算精度。
Under the particle degradation performance and difficulty in selecting the importance density function of particle filter,it designs a random weight adaptively Unscented particle filter based on interacting multiple model algorithm.First,this method absorbs the advantages of the random weight and interacting multiple model filter in the sampling process of Unscented particle filter.It uses the Unscented kalman filter to get the latest measurement information at [k]time for each model.Then,it can input interaction of the corresponding particle as input of each model,after matching the model and using resample,and updating the probability of model.Finally,it can output interaction of the corresponding particle for each model,and sum all particles by random weight in order to continuously update the particle state estimations.The proposed algorithm is applied to the GPS/DR integrated navigation system. Simulation results demonstrate that the proposed algorithm calculated error estimations are better than UPF and IMMUPF,the east position error estimation is in the range [[-8m,+6m]] and north position error estimation is in the range[-8 m,+8 m],thus improving the position calculation accuracy of the GPS/ DR integrated navigation system.
作者
薛丽
杨一
高怡
毛艳慧
XUE Li;YANG Yi;GAO Yi;MAO Yan-hui(School of Physics and Electronic-Electrical Engineering,Ningxia University,Yinchuan 750021,China;School of Electronic Engineering,Xi'an Shiyou University,Xi'an 710065,China)
出处
《电子设计工程》
2019年第13期133-138,共6页
Electronic Design Engineering
基金
国家自然科学基金资助项目(51604226)
宁夏自然科学基金资助项目(NZ17044)
宁夏大学博士启动基金资助项目(BQD2014014)
关键词
交互式多模型
无迹粒子滤波
随机加权
组合导航
interacting multiple model
unscented particle filter
random weight
integrated navigation
