Bayesian Kalman filtering in the presence of unknown noise statistics using factor graphs

摘要

We propose an optimal Bayesian Kalman filtering framework that provides optimal performance relative to the posterior distribution of unknown noise parameters obtained from incorporating data into the prior distribution. The structure of the proposed filter is similar to that of classical Kalman filtering except the use of posterior effective noise statistics. We introduce a factor-graph-based approach to compute the likelihood function required for computing the posterior effective statistics. The performance of the proposed method is verified by applying it to a target tracking example.