Estimation of parameters of one-step predictor with particle filter method

This paper is focused on estimation of the parameters of a system with non-Gaussian noise. Firstly, the Bayesian inference is described and the method of the particle filters is introduced which is directly based on the Bayesian inference. The particle filters method numerically solve a problem of a recursive Bayesian state estimator. Secondly, the method for transformation of a random variables is introduced which changes the relative likelihood of the particle filters according to the distribution of the measurement noise. Thirdly, recursive least square method is derived and linear one-step predictor is described. Fourthly, parameters of the one-step predictor are estimated online with two methods that were mention before. The outputs of both methods are compared and results are discussed. The particle filters method with random variables is analyzed.