Variational Bayesian Kalman Filtering Algorithm for GPS/INS Integrated Navigation System

In the micro-inertial-based satellite navigation system, the traditional Kalman filter-based fusion algorithm has been able to meet the needs of most application scenarios. However, when GPS signal lock is lost or the signal is interfered with, the corresponding measurement errors can change significantly in a short period. At this point, the traditional Kalman filter algorithm cannot effectively estimate the real-time changes in measurement variance, leading to significant errors. Based on the above issues, this paper proposes an adaptive Kalman filter algorithm based on variational Bayesian learning. By dynamically updating the measurement noise matrix of the system using the shape and scale parameters in the inverse Gamma distribution, this approach effectively addresses the problem of the dynamic variation of the measurement noise matrix in practical applications. Simulation and test vehicle experiments show that the Kalman filter algorithm based on variational Bayesian learning can perform real-time estimation of the measurement noise matrix, significantly improving the calculation accuracy of the Kalman filter algorithm in dynamic environments.