【摘要】 预测滤波器是一种基于非线性系统模型的滤波方法,它通过使输出一步前向预测误差最小来估计模型误差,具有较高的估计精度。当预测滤波器应用到多传感器系统时,存在维数多、计算量大等缺点。本文提出了一种预测滤波器的分散化方案,将分散化滤波的优点引入到预测滤波器中。该方案中子滤波器各自给出模型误差的局部估计,并传递给主滤波器。主滤波器由此计算全局最优估计,并反馈重置各子滤波器。通过多传感器卫星测姿系统的仿真,证实分散化之后的预测滤波器与集中的预测滤波器具有相同的精度。同时,分散化的结构具有维数低、矩阵运算量小等优点,而且有利于增强系统的故障检测能力,提高系统的容错性。更多还原

【Abstract】 The Predictive Filter is an estimation method based on nonlinear system model,which determines the optimal model error using a one-step ahead control approach to provide accurate state estimations.To multi-sensor systems,the calculation requirements increase with states and measurements,like other centralized methods.In this paper a new decentralized structure of predictive filter is proposed.The local filters with low dimension measurement equations transfer the local model error estimations to the master filter,through which the master filter can get the global optimal estimation of states.For the global optimality,the states of local filters must be reset to those of the master filter.The benefits of the decentralized structure are not only decreasing the burden of computation but also increasing the ability of fault tolerance for the reason that the fault sensor/local filter can be isolated from other sensors/local filters.This new algorithm was used to estimate the attitude angles of a satellite.The simulation results showed that the decentralized predictive filter had the same accuracy as the centralized one with the advantages of lower dimensions and smaller amounts of matrix computations.Further more,it could improve the ability of fault detection.更多还原