Identification of linearized continuous-time models of mechanical systems from sampled data

An important special problem arising from identification of the linearized continuous-time parametric model of a mechanical system is that all parameters of the model are not independently identifiable. The commonly encountered constraints on model parameters include the symmetry of the inertia matrix and some zero locations in the parameter matrices. In this paper, the identification problem of a linearized continuous-time model of a general n-DOF mechanical system with a priori information is formulated as parameter estimation problem with equation constraints, and a systematical method of solving such a problem is developed in the framework of the least squares equation error. A simplified method is further given which reduces the original problem into several subproblems of much lower dimensions, and is applied to the identification of a linearized continuous-time model of a planar robot manipulator for illustration.