(A) study on iterative learning contral methods for the robot manipulators

Abstract

In this thesis, a class of iterative learning control systems is investigated, and two different types of learning control methods are presented. In particular, the technique is shown to be usefull for the path control of robot manipulators. Specifically, the plant to be controlled is assumed to be linear and time-varying but also periodic so that the plant dynamics remains the same at each iterative operation. In order to find a control input which makes the system state track the desired trajectory within a given error bound for all time, an iterative learning control method is first proposed which is based on a parameter estimator together with an inverse system model to generate the control signal at each iteration. In particular, the parameter estimation is performed with respect to the domain of discrete iterative sequence of operation with time frozen, not with respect to the usual continuous time domain as done in conventional adaptive control techniques. Also, a convergency proof for the algorithm is given. Then the method is shown to be applicable for the motion control of a robot manipulator via linearized model. Another type of an iterative learning control method is proposed by employing a model algorithmic control together with an inverse system model. The proposed model algorithmic control is utilized with respect to the domain of discrete sequence of operation with time frozen, which is different from the conventional model algorithmic control techniques. Also a sufficient condition for convergency of the method is given and two numerical examples, one for a linear periodic system and one for a two link robot manipulator, are illustrated to show the validities of the second method. It is shown that both of the methods are capable of learning in the sense that the subsequent performance of the system is improved in the next operation with the aid of information of previous operation as well as current process data. Finally, it is further shown th...