Dynamic identification of a mitsubishi pa10-6ce robot using motion capture

This paper discusses a method to identify dynamic parameters in a rigid-link flexible-joint robot model using motion capture. The procedure is divided into two parts. First, the robot dynamics are parameterized in a new form to exclude the stiffness torque, and are solved with a rich set of data. The stiffness torque is then reconstructed from the new parameter set, and a torsional spring model is fit to the measured data. This method offers a better conditioned system of equations for weakly-flexible systems, and more flexibility in the choice of torsional spring model. The two-step identification method is experimentally evaluated on a Mitsubishi PA10-6CE.