Input-output data fitting methods are often used for unknown-structure nonlinear system modeling. Based on model-on-demand tactics, a multiple model approach to modeling for nonlinear systems is presented. The basic i...Input-output data fitting methods are often used for unknown-structure nonlinear system modeling. Based on model-on-demand tactics, a multiple model approach to modeling for nonlinear systems is presented. The basic idea is to find out, from vast historical system input-output data sets, some data sets matching with the current working point, then to develop a local model using Local Polynomial Fitting (LPF) algorithm. With the change of working points, multiple local models are built, which realize the exact modeling for the global system. By comparing to other methods, the simulation results show good performance for its simple, effective and reliable estimation.展开更多
本文研究了一类具有不同采样率的分布式多传感器动态系统的数据融合问题,针对一类采样率呈有理数倍关系的动态系统,提出一种基于多源异步采样数据的新融合算法.新算法首先是将来自各个传感器的测量值在融合中心的坐标系中和时钟下进行...本文研究了一类具有不同采样率的分布式多传感器动态系统的数据融合问题,针对一类采样率呈有理数倍关系的动态系统,提出一种基于多源异步采样数据的新融合算法.新算法首先是将来自各个传感器的测量值在融合中心的坐标系中和时钟下进行映射统一;其次,以对目标状态下一时刻的预测值与目标在该时刻状态的估计值之差为基础,建立起描述该融合周期内各个观测点处的目标状态向量之间的动态模型;然后,以该时刻目标状态基于全局信息的估计值为条件,结合建立的新模型和传统的K a lm an滤波器,利用本周期内按序到达的各传感器观测值,依次对各个观测点处目标的状态进行估计和更新;最后,在顺序得到本周期内各个观测点处目标估计值的同时,也将获得下一时刻目标状态基于全局信息的估计值或预测估计值.文中在给出新算法基本思想的同时,也较为详细地对融合算法进行了推导,并通过计算机仿真的方法,将新算法与基于时间校准的算法在估计精确度上进行了比较,从而验证了新算法的有效性.展开更多
基金This project was supported by National Natural Science Foundation (No. 69934020).
文摘Input-output data fitting methods are often used for unknown-structure nonlinear system modeling. Based on model-on-demand tactics, a multiple model approach to modeling for nonlinear systems is presented. The basic idea is to find out, from vast historical system input-output data sets, some data sets matching with the current working point, then to develop a local model using Local Polynomial Fitting (LPF) algorithm. With the change of working points, multiple local models are built, which realize the exact modeling for the global system. By comparing to other methods, the simulation results show good performance for its simple, effective and reliable estimation.
基金Supported in part by the University of Colorado, the US National Science Foundation (Grants CMS-9625086,CMS-0201459, IIS-9711936, and HRD-0095944) the US Office of Naval Research (Grants N00014-97-1-0642 and N00014-02-1-0136) the Colorado Center for Information Storage, the Colorado Advanced Software Institute, Maxtor Corporation, Quantum Corporation, Storage Technology Corporation, and Data Fusion Corporation
文摘Research in control systems, sensor fusion and haptic interfaces is reviewed.
文摘本文研究了一类具有不同采样率的分布式多传感器动态系统的数据融合问题,针对一类采样率呈有理数倍关系的动态系统,提出一种基于多源异步采样数据的新融合算法.新算法首先是将来自各个传感器的测量值在融合中心的坐标系中和时钟下进行映射统一;其次,以对目标状态下一时刻的预测值与目标在该时刻状态的估计值之差为基础,建立起描述该融合周期内各个观测点处的目标状态向量之间的动态模型;然后,以该时刻目标状态基于全局信息的估计值为条件,结合建立的新模型和传统的K a lm an滤波器,利用本周期内按序到达的各传感器观测值,依次对各个观测点处目标的状态进行估计和更新;最后,在顺序得到本周期内各个观测点处目标估计值的同时,也将获得下一时刻目标状态基于全局信息的估计值或预测估计值.文中在给出新算法基本思想的同时,也较为详细地对融合算法进行了推导,并通过计算机仿真的方法,将新算法与基于时间校准的算法在估计精确度上进行了比较,从而验证了新算法的有效性.
