基于T分布变分贝叶斯滤波的SINS/GPS组合导航

为了解决组合导航中由于野值存在而导致传统滤波算法性能下降的问题，针对SINS/GPS组合导航系统模型提出基于T分布的变分贝叶斯高斯滤波算法，充分考虑野值所导致的噪声厚尾特性，将观测噪声建模为T分布.对系统状态和自举变量进行估计，并且在每个滤波时刻借助变分贝叶斯学习对状态估计进行迭代，以逼近真实后验分布.针对噪声存在野值的场景进行仿真验证，结果表明，在SINS/GPS组合导航系统中，当噪声存在野值时，基于T分布的变分贝叶斯组合导航滤波方法具有一定的鲁棒性，并且精度优于传统组合导航滤波方法.
Abstract: In order to resolve the problem of performance degradation caused by the outliers in the traditional filtering methods, a variational Bayesian Gaussian filtering (VBGF) algorithm based on Student-t distribution was proposed for SINS/GPS integrated navigation model. The proposed algorithm took full account of the characteristics of the heavy-tailedness caused by outliers, and the measurement noise was modeled as Student-t distribution. State variables were estimated with latent variables, and the estimation was iterated at each time to approximate the real joint posterior distribution of state and latent variables using the variational Bayesian learning. The results of simulation with outliers show that the proposed filtering algorithm is robust with outliers to a certain degree and reaches a higher precision than the traditional methods in the SINS/GPS integrated navigation system.