Identification of continuous-time systems: direct or indirect?
This paper summarizes and discusses the results of an extensive simulation based exercise in quest of dependable approaches to identification of continuous-time models from sampled input/output data of continuous-time dynamical systems. Two well-established approaches are considered together with several related techniques of parameter estimation. One is an indirect approach in which well-established discrete-time techniques are applied to first estimate a discrete-time model for the original continuous-time system and the model is then transformed into a continuous-time version. The other is a direct approach in which a continuous-time model is estimated straightaway using well-known continuous-time methods. On the surface, the choice between the two approaches may seem trivial but this paper underlines the need to establish dependability of any approach in terms of certain criteria. The results of extensive simulations clearly show that the direct approach is more dependable than the indirect route.
Bibliogr. 23 poz., wykr.