Automated Construction of Robotic Manipulation Programs

Abstract

Society is becoming more automated with robots beginning to perform most tasks in factories<br>and starting to help out in home and office environments. One of the most important<br>functions of robots is the ability to manipulate objects in their environment. Because the<br>space of possible robot designs, sensor modalities, and target tasks is huge, researchers end<br>up having to manually create many models, databases, and programs for their speci c task,<br>an e ort that is repeated whenever the task changes. Given a speci cation for a robot and<br>a task, the presented framework automatically constructs the necessary databases and programs<br>required for the robot to reliably execute manipulation tasks. It includes contributions<br>in three major components that are critical for manipulation tasks.<br>The rst is a geometric-based planning system that analyzes all necessary modalities of<br>manipulation planning and o ers efficient algorithms to formulate and solve them. This<br>allows identi cation of the necessary information needed from the task and robot speci cations.<br>Using this set of analyses, we build a planning knowledge-base that allows informative<br>geometric reasoning about the structure of the scene and the robot's goals. We show how<br>to efficiently generate and query the information for planners.<br>The second is a set of efficient algorithms considering the visibility of objects in cameras<br>when choosing manipulation goals. We show results with several robot platforms using<br>grippers cameras to boost accuracy of the detected objects and to reliably complete the<br>tasks. Furthermore, we use the presented planning and visibility infrastructure to develop<br>a completely automated extrinsic camera calibration method and a method for detecting<br>insufficient calibration data. <br>The third is a vision-centric database that can analyze a rigid object's surface for stable<br>and discriminable features to be used in pose extraction programs. Furthermore, we show<br>work towards a new voting-based object pose extraction algorithm that does not rely on<br>2D/3D feature correspondences and thus reduces the early-commitment problem plaguing<br>the generality of traditional vision-based pose extraction algorithms.<br>In order to reinforce our theoric contributions with a solid implementation basis, we discuss<br>the open-source planning environment OpenRAVE, which began and evolved as a result of<br>the work done in this thesis. We present an analysis of its architecture and provide insight<br>for successful robotics software environments.