Experimental evaluation of some classical and adaptive iterative learning control schemes on a 5DOF robot manipulator

Abstract

In many process industries (e.g., VLSI production lines, Automotive industries, IC welding process, inspections, manipulations), robot manipulators are used to perform the same tasks repeatedly over a finite time interval. The ultimate goal of robotic research is to design intelligent and autonomous robot control systems to perform repetitive tasks that are dull, hazardous, or require skill beyond the capability of humans. The nonlinear nature of the robot dynamics has made this problem a challenging one in robotics research. This highly demanding control problem of driving an industrial robot to follow a desired trajectory perfectly under constrained or unconstrained environment has led to the application of sophisticated control techniques. From the classical or modern control view point, it is a very difficult task to design an intelligent robot control system that can achieve perfect tracking over a finite time interval due to the effect of highly coupled robot dynamics and the presence of the unmodeled dynamics such as friction and backlash that are usually exhibited in the robot system during actual operation.