Estimation of Gaussian noise parameters in nonlinear models using particle filters

Particle filters, which has been designed to find a solution to the problem of state estimation in highly nonlinear systems has been applied to many areas where Kalman filter or its variant are not successful. The success of particle filters also relies on prior knowledge of the model parameters. But in many applications it might not be easy to know or guess the all parameters of the model priori. In this study, it is aimed to make the particle filter adaptive by estimating the unknown noise parameters in Bayesian framework. The proposed method is efficient such that it uses the marginalization approach as in the marginalized particle filters and the conjugate priors are used in order to obtain analytical substructures.