再论联邦滤波在组合导航中的应用可行性CSCD

在联邦滤波算法中,针对子滤波器存在私有状态从而导致全局状态信息融合估计为次优估计的问题,通过一个简单的3维状态、2维量测系统的数值仿真以及速度+姿态传递对准组合导航仿真,验证了联邦滤波的次优估计结果误差太大,难以满足实际系统的高精度滤波需求。联邦滤波估计的次优程度没有经过严格证明,且难以证明,因而存在应用风险。对于以精度为主要性能指标的组合导航系统,不建议采用联邦滤波算法,否则其精度损失不足以弥补其带来的计算量降低和容错性提高的优势。

Further Discussion on the Application Feasibility of Federated Kalman Filter in Integrated Navigation

In the federated Kalman filter algorithm, aiming at the problem that the unique bias states of sub-system will lead to suboptimal estimation of common system states, a simple numerical simulation example consisting of 3 states and 2 measurements and the classical velocity+attitude transfer alignment integrated navigation simulation are carried, which verify that the suboptimal estimation errors of federated Kalman filter are too big, it is difficult to meet the high-precision requirements in practical systems. The suboptimal degree of federated Kalman filter estimation is not strictly proved and hardly to be proved, so there are application risks in real systems. For the integrated navigation system whose main performance index is accuracy, federated Kalman filter algorithm is not recommended, otherwise its accuracy loss is not enough to compensate for its advantages of calculation reduction and fault tolerance improvement.