The choice of perturbation inputs is critical in the identification and model building exercise. One of the major objectives of system identification is accurate estimation of the system parameters. Accurate identification requires that the input be persistently exciting so as to excite all modes of interest. The practitioner would however prefer a 'plant friendly' input signal. Typically, in identification, an input is chosen based on some criteria. We propose a unified multi-objective formulation for the input design. The input can be evaluated as a solution to a constrained multi-objective optimization problem. © 2003 International Federation of Automatic Control.