Recursive Estimator and Stability of a Class of Time Varying Difference Equations

摘要

This report develops a recursive least-squares estimator in an identification observer format with the property that the need for estimating initial conditions is eliminated for time limited data. Estimators are developed that work with observed input/output data corrupted by deterministic, piecewise-deterministic, or stochastic noise. Conditions are established for the asymptotic stability of a class of nonautonomous linear difference equations which apply to the deterministic recursive estimators. The simulations for the recursive identifier indicate that a valid estimator has been developed. Simulation experience has shown, however, that from a practical standpoint the estimator is more powerful in its deterministic or piecewise-deterministic form than in either of its stochastic forms.