A Robust GNSS Navigation Filter Based on Maximum Correntropy Criterion with Variational Bayesian for Adaptivity

Abstract

In this paper, an advanced satellite navigation filter design, referred to as the Variational Bayesian Maximum Correntropy Extended Kalman Filter (VBMCEKF), is introduced to enhance robustness and adaptability in scenarios with non-Gaussian noise and heavy-tailed outliers. The proposed design modifies the extended Kalman filter (EKF) for the global navigation satellite system (GNSS), integrating the maximum correntropy criterion (MCC) and the variational Bayesian (VB) method. This adaptive algorithm effectively reduces non-line-of-sight (NLOS) reception contamination and improves estimation accuracy, particularly in time-varying GNSS measurements. Experimental results show that the proposed method significantly outperforms conventional approaches in estimation accuracy under heavy-tailed outliers and non-Gaussian noise. By combining MCC with VB approximation for real-time noise covariance estimation using fixed-point iteration, the VBMCEKF achieves superior filtering performance in challenging GNSS conditions. The method's adaptability and precision make it ideal for improving satellite navigation performance in stochastic environments.