A parametric learning control method with application for robot manipulators

Abstract

The application of a parametric learning method for the expert tuning of controller parameters is discussed. A multimode structure is used for knowledge acquisition and/or skill refinement. The method is applied in control strategies based on conventional proportional, integral, and derivative (PID) control loops but it can also be used in other multivariable and adaptive controllers. The method can be implemented at low cost and offers advantages when compared with the knowledge acquisition approach used in pattern recognition rule-based expert controllers. The method is applied to the automatic tuning of robot joint controllers in repetitive operations. The learning method is applied to compensate for the inaccuracies in the model as well as variations in working conditions. Sonic experiments with a simulated two-joint manipulator are discussed