自适应网格交互多模型不敏粒子滤波算法  

Adaptive grid interacting multiple model unscented particle filter algorithm

作　　者：许江湖[1] 刘忠[1] 党玲[2] 

机构地区：[1]海军工程大学电子工程学院,湖北武汉430033 [2]海军大连舰艇学院,辽宁大连116018

出　　处：《舰船科学技术》2012年第8期18-22,共5页Ship Science and Technology

基　　金：中国博士后科学基金资助项目(20090461460);湖北省自然科学基金资助项目(2009CDB301)

摘　　要：将一种变结构多模型算法——自适应网格交互多模型(AGIMM)算法和不敏粒子滤波(UPF)算法相结合,提出了自适应网格交互多模型不敏粒子滤波算法(AGIMMUPF)。该算法通过自适应网格实现了模型自适应,从而以较小的模型集合覆盖了目标大范围的机动,并以此来克服固定结构交互多模型粒子滤波(IMMPF)算法存在的缺陷,同时各模型滤波算法采用不敏粒子滤波(UPF)算法,使重要性密度函数融合了最新量测信息,更好地逼近真实状态的后验概率分布。通过计算机仿真证明,提出的算法可以有效提高IMMPF的费效比。A variable structure multiple-model algorithm, i.e., adaptive grid interacting multiple model（AGIMM） algorithm is combined with .unscented particle filter （UPF）, adaptive grid interacting multiple model unscented particle filter （AGIMMUPF）is presented. The algorithm achieves model-set adaptation by adaptive grid, many different maneuvers can be covered by a small set of models, then the limitation of fixed structure interacting multiple model particle filter（IMMPF） is overcomed. The unscented particle filter （UPF） is adopted in every model of the IMM, then the latest observation is integrated into importance density functions that match true posterior more closely. Simulation indicates that this algorithm can effectively improve the cost-effectiveness of IMMPF.

关 键 词：自适应网格 变结构多模型算法 不敏粒子滤波 

分 类 号：TN957[电子电信—信号与信息处理]