Parameter identification and convergence analysis for a class of non-linear systems with time-varying random parameters

Abstract

New developments for parameter identification are presented for discrete stochastic non-linear systems with linear time-varying random parameters, which include bilinear systems. Hammerstain systems, Volterra systems, etc. These identification algorithms arc based on the least square method. Convergence of both identifying parameters and output error are studied and preliminary persistent exciting conditions arc given. Under some conditions, it is proved that identifying parameters and the mean square value of the difference between identifying parameters and random parameters are ensured converging to expectations of these random parameters and the averaging square derivation value of random paramctes, respectively.