Variational Bayesian Multiple Model Filter in the Presence of Outliers

In this article, we present a multiple model (MM) filter based on variational Bayesian (VB) in the presence of outliers. The multiple process and measurement models (M-PMMs) are taken into account to form a hybrid system, where the likelihood function is modeled as Gaussian scale mixture (GSM) distribution. The VB is employed to derive the general framework of state estimation. As two special cases of GSM, Gaussian distribution and multivariate Laplace (GML) distribution are used to make up measurement model sets and implement state estimation. The VB is applied to obtain a closed-form solution. Simulation results show the effectiveness of the proposed algorithm.