广义联邦滤波器的全局最优性

基于分散化滤波算法和信息分配原理,建立了广义联邦滤波器设计理论。证明了联邦滤波器当其主滤波器和局部滤波器的维数都相同时,其全局滤波和集中卡尔曼滤波等价,是最优的;同时提出当主滤波器维数和局部滤波器维数不相同时,达到全局滤波最优的解析补偿方法,其附加计算量小,并可作为一种性能指标用于子系统的软故障检测。在组合导航系统中运用此方法对非公共状态信息进行补偿,仿真结果验证了该方法的有效性。

Global Optimality for Generalized Federated Filters

Based on the decentralized filtering algorithm and information-sharing principle, the paper constructs a design theory for generalized federated filters. It is proved that the global filtering of the federated filter and centralized Kalman filtering are equivalent when the dimension of the master filter is equal to that of local filters for federated filters. If the dimension of the master filter is not equal to that of local filters, an analytic compensation approach for the global filtering to achieve optimality is presented. The approach with less additional computation load could be used as a performance index for soft-fault detection of subsystems. In the integrated navigation systems, the approach is used to compensate non-common state information and its validity has been verified by the simulation results.