Abstract
Dynamics identification studies have been carried out using Runge-Kutta-Gill neural networks for a seven-link industrial robot manipulator called PA-10,manufactured by Mitsubishi heavy industries Ltd. Experimental data of the reference input of joint angle, its velocity, and their output data were used for the actual dynamics identification. Using the identified dynamics, a dynamic controller has been designed so that, for a given position and orientation vector of the tip of the manipulator, the controller computes a desired joint angular velocity vector that can be applied to the PA-10 standard servo controller without changing any hardware. This method can easily be extended for position-force control applications. Promising results are obtained to prove the ability of the proposed method.
