摘要
为了进一步认识基于状态空间投影的一类动态多尺度系统的融合估计算法本质,本文对该算法进行了分析．首先,将该融合估计算法和在最细尺度上直接进行卡尔曼滤波的算法性能进行了比较,并用仿真进行了验证．结果表明,在最细尺度上,融合估计效果比直接进行卡尔曼滤波的效果好．其次,从计算过程和计算量方面,与一般的时间配准方法进行了对比分析．结果表明,该融合估计算法用比较严谨的数学模型代替了时间配准,可以在每个尺度上获得基于全部观测信息的最优估计,但计算量比时间配准方法大．本文的研究为基于状态空间投影的一类动态多尺度系统的融合估计算法的实际应用奠定了基础．
The fusion estimation algorithm of a class of dynamic multiscale system based on state space projection is explained for clearly understanding of its essence in this paper. Firstly, the fusion estimation algorithm is compared with that directly performing Kalman filter at the finest scale. Simulation results show that performance of the fusion estimation algorithm outperforms that of directly performing Kalman filter at the finest scale. Secondly, the fusion estimation algorithm is compared with general time registration method in terms of algorithm process and computational cost. It is shown that, time registration is replaced with rigorous mathematical model in the fusion estimation algorithm, and optimal estimation based on measurement information at all scales can be obtained, and at the same time, the computational cost is larger than that of time registration method. It thus lays a foundation for practical application of the fusion estimation algorithm for a class of dynamic multiscale system.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
2007年第1期84-89,共6页
Control Theory & Applications
基金
国家自然科学基金重点资助项目(60234010)
中国博士后基金资助项目(2005037353)
国家"863"计划资助项目(2006AA042168)
