Graphical stochastic models for tracking applications with variational message passing inference

In this paper we present a novel, highly-adoptable, state-estimation filter based on the framework of graphical stochastical models and variational message passing inference. We evaluate our method on both real and simulated data for tracking applications. Our experimental results show that the proposed approach offers qualitative and computational advantages over established filter methods in practical situations, where the noise within a process is not simply a Gaussian noise, but rather described by a more complex distribution.