FIIR-TCM - 56. Robot Work Cell Calibration and Error Compensation

56. Robot Work Cell Calibration and Error Compensation

Pey Yuen Tao: Industrial Robotics Team, Mechatronics Group, Singapore Institute of Manufacturing Technology, Singapore
Shabbir Kurbanhusen Mustafa: Industrial Robotics Team, Mechatronics Group, Singapore Institute of Manufacturing Technology, A*STAR, Singapore
Guilin Yang: Institute of Advanced Manufacturing, Ningbo Institute of Materials Technology and Engineering of the Chinese Academy of Sciences, Zhenhai District, Ningbo, Zhejiang, People’s Republic of China
Masayoshi Tomizuka: Department of Mechanical Engineering, University of California, Berkeley, CA, USA

56.1 Introduction

56.2 Literature Survey and Commercial Solutions

56.3 Mathematical Preliminaries
Transformation Matrices and Twist Representation
Adjoint Representation for Lie Group and Lie Algebra
Matrix Exponentials
Product of Exponentials

56.4 Calibration
Methodology
Implementation and Experimentation

56.5 Compensation
Methodology
Point Cloud Registration
Defining Error Functions
Solving the Least Square Problem
Data Preprocessing
Simulation

56.6 Calibration with Compensation Framework
System Overview
Simulation

56.7 Summary, References

Abstract

Industrial robotic manipulators have excellent repeatability. However, accuracy is significantly poorer due to the numerous error sources in the robot work cell. A literature survey on recent calibration and compensation methods is presented as well as an overview of existing commercially available solutions. Subsequently, two methods to improve robot work cell accuracy are proposed to illustrate the concepts behind calibration and error compensation. The first method is a model-based calibration approach, where the end-effector poses and corresponding joint angles are measured and used to improve the kinematic model of the robot. The second method is a non-model-based compensation approach where sensor information is used to establish the relative pose of the work object and tool frames at discrete locations. Following this, a robot accuracy enhancement framework is proposed in which both techniques are integrated for an industrial robotic work cell where the strengths of one method are used to compensate for the inherent weaknesses of the other. Specifically, the non-model-based compensation approach improves the robot accuracy locally at the point of compensation, accounting for unmodeled effects which cannot be compensated by the model-based calibration approach, while calibration improves the nominal robot kinematic model extending the compensation effects to a larger working envelope.