Direct and indirect least squares methods in continuous-time parameter estimation (Q1101058)

This paper extends well-established discrete-time parameter estimation techniques to continuous-time nonlinear models. It is shown that, in most cases, the direct integral least squares (DILS) and symmetric bootstrap (SB) estimators are numerically efficient and are statistically sound alternatives to the conventional indirect least squares (ILS) approach. In most applications a few iterations of the SB algorithm, started at the DILS estimates, yield results as good as the computationally more expensive ILS method. Thus the direct approach often results in a considerable saving of computational effort.