Experimental neural decomposed dynamic adaptive control of robot manipulator

We investigate a new neural network decomposed dynamic adaptive controller for the robot manipulator. The robot controller is divided into two components. The dynamic action is based on the decomposition of dynamical model into fixed and variable torque. The fixed torque represents the nominal torque and the variable torque, based on the parameter estimation obtained with an ANN, tracks the change in the dynamical system. To estimate the variable parameters we used an ANN parameter estimator. Two schemes were derived from this approach and applied successfully to a SCARA robot