Abstract
This paper presents a new approach in estimating the parameters of a dynamic system from an initial and relatively inaccurate state-space model. The proposed approach is based on determining an optimal control law that minimizes a quadratic performance index for a free final state condition by means of solving a two-point boundary-value problem. The control input to both the model and the physical system is assumed to be the difference between their outputs. The proposed approach is applied to a second order system that has a modeling error of about 40%. The new approach succeeded in reducing the modeling error to about 5%.