M. S. Thesis Abstract

An Experimental Investigation on Dynamic Vision Guided Pick-Up of Moving Objects

James Douglas Downs

June 1997

This thesis addresses the problem of dynamic recognition and pick-up of objects in pseudo-random motion using position-based visual servoing to determine the initial state of a target object of interest and a recursive neural network based control to predict the final object state for the pick-up to occur.  A vibratory part feeder is used to generate the pseudo-random motion of various parts, while a FIVS vision system mounted on the wrist of a T3-786 industrial robot is used for visual feedback and actuation respectively.  The control strategy was developed and implemented in real-time in order to investigate the capabilities of such a design.