Using black box and machine learning approach in controlling a machine

Considering the system as a black box, this paper studies backlash error and positional error maps in mechanical systems. The paper proposes implementation of machine learning methodology in conjunction with using error maps to compensate for the positional errors in the system. A new generation of the proportional keen approximator is introduced in the cylindrical coordinate system. The earlier version of this learning methodology was first implemented to the Cartesian coordinate system.