Fundamentals of Real-Time Collision-Avoidance Path Planning for Robots: The GhostMethod

摘要

The dissertation proposes a solution to the following problem: To use a robotmanipulator and visual aid to grasp a moving element belonging to a certain class. The elements (targets) are transported on a conveyor belt, and must be grasped while in motion by the robot gripper and placed at designed positions according to the class to which they belong. It begins with the vision system and introduces a 2-D methodology for a planar RR-type robot. It presents the kinematics, dynamics and low-level control of the robot manipulator. It then analyzes in detail various aspects of 2-D navigation and the velocity profile along the path. It presents an experimental analysis of parameter variations in order to determine both the linear model and the parameter variation limits. Next, it studies the off-line linearization of the robot dynamics and the tessellation of the working space, in order to construct tables for local models of the robot. It then studies the ghost method in relation to collision avoidance. Lastly, it includes a survey of further research which may be of interest to others.