A new solution of Bayesian filtering in autoregression model with Laplace distributed innovations is presented in the thesis. An algorithm leading to an effective solution of the problem is proposed.