Monte Carlo Gaussian Sum Filter For State Estimation of Nonlinear Dynamical Systems∗

This work presents a novel nonlinear/non-Gaussian state estimation algorithm, named as, Monte Carlo Gaussian Sum Filter (MC-GSF). The proposed approach combines the elements of Monte Carlo (MC) sampling and design choices in recently developed Unscented Gaussian Sum Filter (UGSF). While the MC sampling retains the sampling benefits in capturing moments of non-Gaussian densities, the design choices in UGSF improves the ability of MC samples by means of sum of Gaussians representation. Further, the design choices in UGSF also overcomes the potential degeneracy issues persisting with Particle filters and Gaussian Sum Filters. We demonstrate the superiority of proposed approach by implementing on an illustrative case study.