ROBOT TRANSFER MOVEMENTS WITH VISUAL INPUT
Abstract
Robot transfer movements involve the transferring of 3-D objects from one place to another in a robot workspace, avoiding obstacles in between. Three problems crucial to the planning of robot transfer movements incorporating visual input are addressed here. (1) The first is a graph-theoretic approach to 3-D object recognition and estimation of position and orientation. In this approach, the graphs constructed from the 2-D projections of a 3-D object are modeled as subgraph isomorphisms of the model-graph, constructed from the 3-D model of the 3-D object. Recognition is by subgraph searching. Techniques are described for the estimation of position and orientation of a 3-D object based on this approach. Experiments using a TV camera are conducted to evaluate the performance. (2) The second problem is a hierarchical-orthogonal-space approach to the 3-D path planning of a moving object among obstacles. Collison checking is done in the three orthogonal 2-D images of a 3-D environment. A hierarchical path search method is used to speed up the search process. Computer simulation experiments are presented. (3) The third problem is the sequential updating of a robot workspace model using a camera situated at a sequence of different viewing directions. An octree model is used to represent the workspace and existing information in the octree is used to calculate the next best viewing direction. Computer simulation results using the proposed algorithm are presented.
Degree
Ph.D.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.