Particle filters for partially observed Markov chains
Abstract
We consider particle filters in a model where the hidden states and the observations form jointly a Markov chain, which means that the hidden states alone do not necessarily form a Markov chain. This model includes as a special case non-linear state-space models with correlated Gaussian noise. Our contribution is to study propagation of errors, stability properties of the filter, and uniform error estimates, using the framework of Le Gland and Oudjane.