State estimation based on Generalized Gaussian distributions

This paper presents a novel strategy of particle filtering for state estimation based on Generalized Gaussian distributions (GGDs). The proposed strategy is implemented with the Gaussian particle pilter (GPF), which has been proved to be a powerful approach for state estimation of nonlinear systems with high accuracy and low computational cost. In our investigations, the distribution which gives the complete statistical characterization of the given data is obtained by exponent parameter estimation for GGDs, which has been solved by many methods. Based on GGDs, an extension of GPF is proposed and the simulation results show that the extension of GPF has higher estimation accuracy and nearly equal computational cost compared with the GPF which is based on Gaussian distribution assumption.