Particle Gaussian Mixture Filters: Application and Performance Evaluation

The particle Gaussian mixture filters are a new class of Bayesian estimation techniques that have been proposed for the general multimodal nonlinear filtering problem. In this paper, we evaluate the estimation performance of the particle Gaussian mixture filters on a collection of benchmarking problems that have been selected from recent literature. The problems are chosen to facilitate comparisons with the estimation results of other recently proposed general purpose nonlinear filters. We investigate the effect of coupling, the dimensionality of the problem and the number of particles on the estimation performance. Our results indicate that the performance of the particle Gaussian mixture filters are at par with feedback particle filters and log homotopy based Daum Huang particle flow filters on the test problems.