Robust Multiple Measurement Model Filtering With Guaranteed Stability Under Outliers and Measurement Loss

This paper presents a robust multiple measurement model filter that accounts for outliers and measurement loss. A measurement model set including Gaussian distribution, Multivariate Laplace (ML) distribution and a binary indicator is established to describe the normal measurements, outliers, and measurement loss. To obtain a closed-form solution, the state and other parameters are derived using variational Bayesian inference. It is proven that the scale parameter of the ML distribution is bounded. Building on this result, the boundedness of the error covariance and estimation errors are further analyzed using the concept of collectively uniform detectability. The simulation results show the excellent adaptability of the proposed algorithm in the presence of both outliers and measurement loss.