CMAC based iterative learning control of robot manipulators

An iterative learning control scheme is presented. It incorporates a version of the cerebellar model articulation controller (CMAC) memory for the torque sequence generation. A learning rule is constructed by utilizing a gradient descent algorithm, and a map which updates old data stored in a distributed form is defined. It is shown that the training factor should be less than two for error convergence in the case of high-gain feedback