THE ELEVENTH IEEE SENSOR ARRAY AND MULTICHANNEL SIGNAL PROCESSING WORKSHOP (IEEE SAM2020)

149 / 2019-12-15 09:12:00

A Novel Kalman Filter with Adaptive Measurement Bias Estimate for DVL/SINS Integrated Navigation

Integrated Navigation; biassed measurement noise; Kalman filter; Normal inverse Wishart; variational Bayesian

IEEE SAM 2020: Special Sessions > Multi-Sensor Data Fusion for Underwater Application

In this brief, a novel adaptive Kalman filter with adaptive estimate for measurement bias is proposed to handle the filtering problem with biassed measurement noise, which may be encountered in the application of DVL/SINS integrated navigation. The Variational Bayesian (VB) method is used in the proposed filter, and the one-step prediction probability density function (PDF) is modelled as a Gaussisan distribution with zero mean vector, and the measurement noise mean vector and measurement noise covariance matrix joint PDF is modelled as a Normal-inverse-Wishart (NIW) distribution. Based on the established hierarchical Gaussian model, the estimated state vector, the measurement noise bias vector as well as the measurement noise covariance matrix are corporately estimated. As compared with existing VB-based adaptive Kalman filter (VBAKF), the proposed filer has the better precision in the last simulation.