Optimal Parameter Discretization Strategy for Multiple Model Adaptive Estimation and Control.

摘要

A method is proposed for designing multiple model adaptive estimators (MMAE) to provide combined state and parameter estimation in the presence of an uncertain parameter vector. It is assumed that the parameter varies over a continuous region and a finite number of constant-gain filters is available for the estimation. The estimator elemental filters are chosen by minimizing a cost functional representing the average state prediction error autocorrelation, with the average taken as the true parameter ranges over the admissible parameter set. A second-order example is used to illustrate the increase in performance over previously accepted filter selection methods. By minimizing a cost functional representing the average parameter prediction error autocorrelation, a parameter estimator is designed which is different from the state estimator. The parameter estimator found with this method provides lower average mean square parameter estimation error than previously accepted design methods. An analogous method is proposed for designing multiple model adaptive controllers to provide stabilizing control in the presence of an uncertain parameter vector. A finite number of constant-gain controllers is used to regulate a system with a parameter vector that varies over a continuous region of the parameter space. The controller elemental filters are chosen by minimizing a cost functional representing the average regulation error autocorrelation, with the average taken as the true parameter ranges over the admissible parameter set. Keywords: Numerical optimization, Adaptive control systems, Theses. (aw)