Ph.D. Thesis Summary

A Study of Dynamic Pursuit of Moving Objects with Hand-Eye Coordination

Yifei Qian

June 1995

This thesis addresses the problem of grasping moving objects from a vibratory feeder with robotic hand-eye coordination. The process of dynamic pursuit of moving objects has been divided into three phases, namely, (1) searching for a target of interest, (2) pursuing the target found, and (3) grasping the target once certain specified conditions are satisfied. The system is a combination of three subsystems- a robotic system, a vision system, and a vibratory feeder. Each subsystem has its independent controller.

To deal with the problem, a hierarchical structure was developed for the system. A central coordinator was constructed to provide the system with a task plan, decompose it into subtasks, and assign them to each subsystem in the course of the process development. It works at the highest level or the system level to govern the whole process. The three subsystems work at a lower level or the execution level to perform the tasks assigned by the central coordinator. At the same time, the problem was formulated in the context of Prey Capture, in which the robot behaves as a "pursuer", and a moving target a passive "prey" or "evader." The formulation creates an opportunity to use linguistic rules based on "rules-of-thumb" experiences and engineering judgments for specifications of control laws, and apply experimental or heuristic knowledge as basis for logical inference. This strategy is particularly advantageous since the pursuit process is highly nonlinear and is impractical to model accurately.

A flexible integrated vision system was used to feed back the information about moving targets together with the scene observer designed. The optics of the vision system was first discussed. The algorithm to determine the pose (position and orientation) of a moving target was then presented. It was concluded that both motion parameters and image features of moving targets are necessary and useful for the pursuit process. The motion parameters, i.e. position and orientation, are needed in the pursuit phase of the process to estimate the motion of a moving target while image features are required for the search and grasp of the moving target in the phases of target search and target grasp of the process.