A Bayesian Nonparametrics based Robust Particle Filter Algorithm
This paper is concerned with online estimation of a nonlinear dynamic system from a series of noisy measurements. The focus is on cases wherein outliers are present in-between normal noises. We assume that the outliers follow an unknown generating mechanism which deviates from that of normal noises, and then model the outliers using a Bayesian nonparametric model called Dirichlet process mixture (DPM). A online mixture learning algorithm is derived for sequential inference for the DPM model, which works in tandem with a particle filter (PF) based state estimation procedure. The resulting algorithm is thus termed DPM based robust PF (DPM-RPF). The nonparametric feature makes this algorithm allow the data "speak for itself" to determine the complexity and structure of the outlier model. Simulation results show that the presented algorithm performs remarkably better than two state-of-the-art robust PF methods especially when outliers appear frequently along time.
READ FULL TEXT