An Online Parameter Estimation Method for a class of Nonlinear Systems Based on Particle Filter

Nonlinear system identification is one of the most important topics all over the word. Until now, there are many of off-line identification methods which exhibit well performance. The online approach with non-Gaussian noise, however, is still a challenge. For a class of nonlinear systems where all of the candidate parameters are contained in a definite parameter set, an online parameters and sates estimation method is proposed based on particle filter and Bayes theorem as the following steps. Firstly, regarding all of the candidates, the states are estimated by particle filter (PF) algorithm. Secondly the posterior probabilities of all of candidates are calculated according to the Bayes theorem; then the weights of all of the candidates are obtained through normalization. Lastly, the parameters and sates are estimated ultimately according to the weighted sum of all of the candidates and states. Numerical illustrations are presented to exhibit the application of the method proposed herein, and the performance of the method is examined.