Abstract
This paper presents an algorithm for least squares estimation of the states of a dynamic system. This method based on a modification of the Gauss-Newton method posesses characteristics of both the steepest descent method and the Gauss-Newton method. Consequently, it has the advantage of quick convergence over the steepest descent method and the advantage of good stability over the Gauss-Newton method. This method is useful when the observer of the states of the system is very noisy and/or when the initial estimation errors are large. A computer simulation substantiates the above remarks. The application to the parameter estimation of a response model of the human operator in a control system is also shown.
