Estimation of the parameters of sampled-data systems by means of stochastic approximation†

Abstract

This paper discusses the application of stochastic approximation to the estimation of sampled-data system parameters, including the sampling interval. The sampled-date systems considered have a closed-loop configuration, continuous input and output signals, and error sampling. A continuous dynamic system follows a zero-order hold. Sufficient conditions are given for the mean-square convergence of a stochastic approximation algorithm of the Kiofer-Wolfowitz type which is used to estimate all the parameters of the sampled-data system. Simulation results are presented which illustrate the theoretical results. In addition, simulations indicate that good estimates of the mean values of parameters are still possible when the parameters have random components.