摘要
针对粒子滤波中重采样过程与优化提议分布的处理方式导致的粒子溃退和算法实时性下降问题,通过广义UT变换原理和卡尔曼滤波预测更新机制的引入,实现当前量测信息对于状态估计结果的直接优化,给出了一种基于广义UT变换的粒子滤波算法.另外,将改进后算法与交互式多模型相结合,进而提出了一种基于广义UT变换的交互式多模型粒子滤波算法.理论分析和仿真结果表明:新算法在计算复杂度方面与标准粒子滤波相近,在滤波精度方面优于标准粒子滤波及其改进算法.
To solve the particles impoverishment and the real-time decline caused by re-sampling and the proposal distribution optimization,a novel particle filtering algorithm based on generalized unscented transformation is proposed.The generalized unscented transform and the one-step state prediction and observation update of Kalman filter are introduced to realize the optimization of state estimation by the latest observation information.Then,by means of combining the improved algorithm with interacting multiple model,the interacting multiple model particle filtering algorithm based on generalized unscented transformation is proposed.The theoretical analysis and experimental results show that the new algorithm is close to the standard particle filter in computational complexity and superior to the standard particle filter and its improved algorithms in precision.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2010年第6期1443-1448,共6页
Acta Electronica Sinica
基金
国家自然科学重点项目(No.60634030)
国家自然科学基金(No.60702066)
高等学校博士学科点专项科研基金(No.20060699032)
关键词
混合系统
粒子滤波
交互式多模式
广义UT变换
hybrid system
particle filtering
interacting multiple model
generalized unscented transformation
