Information estimates in the problem blind identification of dynamic systems

V. V. Karelin

Saint Petersburg State University, Universitetskaya nab. 7-9, Saint Petersburg, 199034

Abstract: The convergence and stability problems of the Bayesian estimations are considered for identification of stochastic control systems. The main tools to prove the convergence is the information measure of the mismatch between the given distribution and the estimation. As a measure, the so-called information number of Kulbaka–Leiblera is taken. The convergence of the estimation of the transitive function to the process to the nonstationary transitive function is established.

Keywords: Bayesian estimation, identification, information number.

UDC: 639.3