An Optically Driven Learning Control for Industrial Manipulators

Abstract

Automation of manufacturing processes often involves integration of some type of robotic manipulator. Automatic brushing, painting, deburring, welding, and seam tracking are examples of tasks that manipulators should be able to perform accurately and autonomously. However, there are dynamic interactions between the manipulator and its environment that produce uncertainties and the manipulator controller has to be capable of compensating for them. Furthermore, due to unmodelled dynamics and flexibilities in the manipulator itself, the task of the controller becomes even more complicated. This paper presents an optical-based learning controller that utilizes past experience and sensory information about the current state to overcome these difficulties. This scheme tries to mimic the way a human brain compensates for such uncertainties. The optical sensor uses an LED emitter and eight detectors to provide distance information normal to the surface of the workpiece. The resulting scheme is implemented on a GE-P50 robot and used to track different unknown surfaces.