A variational Bayesian approach to robust sensor fusion based on Student-t distribution

In this paper, a robust sensor fusion method is proposed where the measurement noise is modeled by a Student-t distribution. The Student-t distribution has a heavy tail compared to the Gaussian distribution and is robust to outliers. We formulate sensor fusion as a state space estimation problem in the Bayesian framework. Both batch and recursive variational Bayesian (VB) algorithms are developed to perform this non-Gaussian state space estimation problem to obtain the fusion results. Computer simulations show that the proposed approach has an improved fusion performance and a lower computation cost compared to methods based on Gaussian and finite Gaussian mixture models.