Optimal input design for autoregressive model discrimination with constrained output variance

Abstract

The optimal input design problem is considered for discriminating autoregressive models when the output variance is constrained. It is shown that a Ds -optimal input design can be achieved with an input comprised of a finite number of sinusoidal components with frequencies and amplitudes determined by solving a system of nonlinear equations. Digital simulation studies illustrate the usefulness of the Ds -optimal input for autoregressive model discrimination.