Aspects of input signal design for order and parameter estimation in linear dynamical systems

It is shown that a close relationship between the input signal design and model order estimation procedures (possibly with an expansion to structure determination procedures of multivariable systems) exists in linear dynamical systems. This implies that the input sequence is selected in order to improve the accuracy for the parameter estimates (e.g., in D-optimality or A-optimality sense) and at the same time can make the order tests more powerful. An overall identification algorithm, including the amplitude constrained input sequence design, estimation procedure, and an alternative instrumental determinant ratio test is proposed. Simulation results show that the algorithm works satisfactorily for the relatively short identification horizon.