Robotic Manipulation Using an Open-Architecture Industrial Arm: A Pedagogical Overview

摘要

Robotics education at the undergraduate level is most effective as a coupling between theoretical concepts and tangible experiments. Making this connection effective requires a pragmatic way of applying the traditional robotic material to exciting laboratory exercises. A course recently offered in Harvard’s School of Engineering and Applied Sciences, simply titled ‘‘Introduction to Robotics,’’ utilizes an open-architecture robotic arm to give students handson experience with topics that they encounter in lecture. None of the experiments conducted in this course are wholly novel; however, the use of an open-architecture hardware or software system enables the instructor to rapidly prototype lab exercises with minimal effort. This column will give an overview of the apparatus and experiments used for this course. Apparatus Overview The majority of existing industrial arms are not conducive to education: the user interface (software or teach pendant) is typically oriented to repetition of precise tasks. Although the physical instantiation of the arm is not a primary concern, the software interface to the arm is of quintessential importance. Students should not spend an inordinate amount of time learning a proprietary motion description language specific to any given manufacturer. Instead, we settled on the six degrees of freedom (DoF) open-architecture robot from Quanser. This system consists of a 5 DoF CRS CataLyst-5 from Thermo Electron Corporation mounted to a linear track (for the sixth axis). The existing CRS controller is supplemented with a Quanser control board, allowing the user to switch between the industrial controller and an open-architecture controller in which the user has access to everything from high-level commands to individual joint signals. The open-architecture configuration uses a Matlab or Simulink interface that includes libraries for common functions such as kinematics and control. At the base of the workspace, a peg board was installed, which enabled the lab instructors to interchangeably place objects and obstacles for the latter labs. Additionally, an overhead camera, with its primary axis anti-parallel with the inertial z-axis, is used for vision. Lab Overview Prior to each lab, students write Matlab functions to solve tasks as prelab exercises. Each successive lab builds on tools that students developed for the previous exercise while maintaining a close connection to the material presented in class. Lab 1: Forward Kinematics Given the Denavit-Hartenberg convention and the geometry of the arm (taken from data sheets), the students first write Matlab functions for the homogeneous transformations and a script to calculate position and orientation of the tool frame. During lab time, the students input various joint angles into both the arm controller and their script. They are then required to physically measure the location of the tool frame and compare to their predictions while using observations of the arm to debug any discrepancies. It is important that the scripts consider joint limitations, and thus some of the joint angles given to the students are outside the physical limits, so as to test the robustness of their code. Furthermore, the students use this script to evaluate the extent of the workspace by varying the joint angles through the configuration space.