Sequential auxiliary particle belief propagation

This paper discloses a novel algorithm for efficient inference in undirected graphical models using sequential Monte Carlo (SMC) based numerical approximation techniques. The methodology developed, titled "auxiliary particle belief propagation", extends the applicability of the much celebrated (Loopy) belief propagation (LBP) algorithm to non-linear, non-Gaussian models, whilst retaining a computational cost that is linear in the number of sample points (or particles). Furthermore, we provide an additional extension to this technique by analyzing temporally evolving graphical models, a problem which remains largely unexplored in the scientific literature. The work presented is thus a general framework that can be applied to a plethora of novel distributed fusion problems. In this paper, we apply our inference algorithm to the (sequential problem of) articulated object tracking.