Abstract: A self-organizing sequential Monte Carlo filter for nonlinear state-space model is proposed. An expanded state-space model is defined by augmenting the state vector with the unknown parameters of the original state-space model. The self-organizing state-space model can also be applied to the self-tuning of the noises dispersion. A local likelihood is introduced and used to select the optimal parameter from a finite number of possible values. Examples of hysteretic system identification are shown to verify the effectiveness of the proposed method.