基于多模型联合滤波的MIMU/GPS组合导航方法

MIMU/GPS integrated navigation method based on IMM-based combined filter

针对单一噪声假设和单一模型假设对组合导航精度的不利影响,提出了基于多模型联合滤波的MIMU/GPS组合导航方法。首先,用分布集合的概念进行误差建模,实现了随机噪声和有界噪声的融合建模;而后,推导了多椭球加权Minkowski和的外定界椭球的运算过程,并给出参数优化方法,从而实现了随机/有界双重不确定条件下的多模型交互,得到了交互多模型联合滤波算法的具体迭代过程;最后,将提出的方法用于MIMU/GPS组合导航。试验结果表明,交互多模型联合滤波算法可以将导航位置误差由最大超过10 m降低到1.5 m以下,导航精度显著优于单模型滤波方法,同时也优于标准交互多模型算法。特别是,由于交互多模型联合滤波算法充分利用了两种不同性质的噪声,既可以避免单一噪声假设导致的导航性能下降,又可以保证边界估计,对于后续的控制制导具有重要意义。

In view of the adverse effects of single noise assumption and single model assumption on integrated navigation accuracy,a MIMU/GPS integrated navigation method based on multi-model combined filter is proposed.Firstly,by modeling the error with the concept of distribution set,the fusion model containing both random noise and bounded noise is established.Then,the operation process and parameter optimization method of the outer bound ellipsoid of multi ellipsoid weighted Minkowski sum are derived,so that the multi-model interaction considering stochastic/bounded dual uncertainty is realized,and the specific iterative process of the interactive multi model combined filter algorithm is obtained.Finally,the proposed method is used in MIMU/GPS integrated navigation.The experimental results show that the multi-model interactive filtering algorithm can reduce the navigation position error from more than 10 m to less than 1.5 m,and the navigation accuracy is significantly better than the single model filtering method,and also better than the standard interactive multi model algorithm.Especially,because the proposed algorithm makes full use of two different kinds of noise,it can not only avoid the navigation performance degradation caused by single noise hypothesis,but also guarantee the boundary estimation,which is of great significance for the subsequent control and guidance.