基于逆Gamma分布的变分自适应滤波算法

Variational Adaptive Filter Based on Inverse Gamma Distribution

针对Bayesian滤波在组合导航量测噪声随机模型不准确时引发的估计精度下降问题,提出了一种基于逆Gamma分布优化的变分自适应滤波算法。该算法借鉴变分Bayesian学习理论,通过逆Gamma分布进一步精化了Bayesian滤波随机模型,准确高效地实现了量测噪声协方差的自适应估计,显著改善了滤波估计性能。最后通过紧耦合组合导航数据仿真实验,结果表明本文所探讨的优化算法能实时跟踪量测噪声变化,保障滤波估计精度,且运算量小速度快,易于工程实现,为今后研究工作于时变噪声环境下的导航系统及其扩展应用提供一定的理论支持。

Aiming at solving the problem of decreasing estimation accuracy caused by Bayesian filter when the random model of integrated navigation measurement noise is inaccurate,this paper proposes a refined algorithm of adaptive filter based on inverse Gamma distribution from variational Bayesian learning algorithm.The algorithm further refines Bayesian filter random model,and significantly improves the filtering estimation accuracy based on the variational learning theory.The results show that the optimized algorithm can realize real-time tracking of measurement noise changes,and easy to be further improve the estimation accuracy,which is also computationally fast and less burdensome.This research provides some theoretical support for the application and expansion of navigation system under noisy environment.