This edition first published 2017
© 2017 John Wiley & Sons, Inc.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions.
The right of Aleksandar Vakanski and Farrokh Janabi‐Sharifi to be identified as the author(s) of this work has been asserted in accordance with law.
Registered Offices
John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA
Editorial Office
111 River Street, Hoboken, NJ 07030, USA
For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com.
Wiley also publishes its books in a variety of electronic formats and by print‐on‐demand. Some content that appears in standard print versions of this book may not be available in other formats.
Limit of Liability/Disclaimer of Warranty
The publisher and the authors make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties; including without limitation any implied warranties of fitness for a particular purpose. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for every situation. In view of on‐going research, equipment modifications, changes in governmental regulations, and the constant flow of information relating to the use of experimental reagents, equipment, and devices, the reader is urged to review and evaluate the information provided in the package insert or instructions for each chemical, piece of equipment, reagent, or device for, among other things, any changes in the instructions or indication of usage and for added warnings and precautions. The fact that an organization or website is referred to in this work as a citation and/or potential source of further information does not mean that the author or the publisher endorses the information the organization or website may provide or recommendations it may make. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this works was written and when it is read. No warranty may be created or extended by any promotional statements for this work. Neither the publisher nor the author shall be liable for any damages arising here from.
Library of Congress Cataloguing‐in‐Publication Data
Names: Vakanski, Aleksandar, author. | Janabi-Sharifi, Farrokh, 1959– author.
Title: Robot learning by visual observation / Aleksandar Vakanski, Farrokh Janabi‐Sharifi.
Description: Hoboken, NJ, USA : John Wiley & Sons, Inc., [2017] | Includes bibliographical references and index.
Identifiers: LCCN 2016041712| ISBN 9781119091806 (cloth) | ISBN 9781119091783 | ISBN 9781119091998 (epub) | ISBN 9781119091783 (epdf)
Subjects: LCSH: Robot vision. | Machine learning. | Robots--Control systems. | Automatic programming (Computer science)
Classification: LCC TJ211.3 .V35 2017 | DDC 629.8/92631-dc23
LC record available at https://lccn.loc.gov/2016041712
Cover image: fandijki/Gettyimages
Cover design by Wiley
To our families
The ability to transfer knowledge has played a quintessential role in the advancement of our species. Several evolutionary innovations have significantly leveraged the knowledge transfer. One example is rewiring of the neuronal networks in primates’ brains to form the so‐called mirror neuron systems, so that when we observe tasks performed by others, a section of the brain that is responsible for observation and a section that is responsible for motor control are concurrently active. Through this, when observing actions, the brain is attempting at the same time to learn how to reproduce these actions. The mirror neuron system represents an especially important learning mechanism among toddlers and young kids, stimulating them to acquire skills by imitating the actions of adults around them. However, the evolutionary processes and modifications are very slow and prodigal, and as we further developed, we tended to rely on employing our creativity in innovating novel means for transferring knowledge. By inventing writing and alphabets as language complements, we were able to record, share, and communicate knowledge at an accelerated rate. Other innovations that followed, such as the printing press, typing machine, television, personal computers, and World Wide Web, each have revolutionized our ability to share knowledge and redefined the foundations for our current level of technological advancement.
As our tools and machines have grown more advanced and sophisticated, society recognized a need to transfer knowledge to the tools in order to improve efficiency and productivity, or to reduce efforts or costs. For instance, in the manufacturing industry, robotic technology has emerged as a principal means in addressing the increased demand for accuracy, speed, and repeatability. However, despite the continuous growth of the number of robotic applications across various domains, the lack of interfaces for quick transfer of knowledge in combination with the lack of intelligence and reasoning abilities has practically limited operations of robots to preprogrammed repetitive tasks performed in structured environments. Robot programming by demonstration (PbD) is a promising form for transferring new skills to robots from observation of skill examples performed by a demonstrator. Borrowed from the observational imitation learning mechanisms among humans, PbD has a potential to reduce the costs for the development of robotic applications in the industry. The intuitive programming style of PbD can allow robot programming by end‐users who are experts in performing an industrial task but may not necessarily have programming or technical skills. From a broader perspective, another important motivation for the development of robot PbD systems is the old dream of humankind about robotic assistance in performing everyday domestic tasks. Future advancements in PbD would allow the general population to program domestic and service robots in a natural way by demonstrating the required task in front of a robot learner.
Arguably, robot PbD is currently facing various challenges, and its progress is dependent on the advancements in several other research disciplines. On the other hand, the strong demand for new robotic applications across a wide range of domains, combined with the reduced cost of actuators, sensors, and processing memory, is amounting for unprecedented progress in the field of robotics. Consequently, a major motivation for writing this book is our hope that the next advancements in PbD can further increase the number of robotic applications in the industry and can speed up the advent of robots into our homes and offices for assistance in performing daily tasks.
The book attempts to summarize the recent progress in the robot PbD field. The emphasis is on the approaches for probabilistic learning of tasks at a trajectory level of abstraction. The probabilistic representation of human motions provides a basis for encapsulating relevant information from multiple demonstrated examples of a task. The book presents examples of learning industrial tasks of painting and shot peening by employing hidden Markov models (HMMs) and conditional random fields (CRFs) to probabilistically encode the tasks. Another aspect of robot PbD covered in depth is the integration of vision‐based control in PbD systems. The presented methodology for visual learning performs all the steps of a PbD process in the image space of a vision camera. The advantage of such learning approach is the enhanced robustness to modeling and measurement errors.
The book is written at a level that requires a background in robotics and artificial intelligence. Targeted audience consists of researchers and educators in the field, graduate students, undergraduate students with technical knowledge, companies that develop robotic applications, and enthusiasts interested in expanding their knowledge on the topic of robot learning. The reader can benefit from the book by grasping the fundamentals of vision‐based learning for robot programming and use the ideas in research and development or educational activities related to robotic technology.
We would like to acknowledge the help of several collaborators and researchers who made the publication of the book possible. We would like to thank Dr. Iraj Mantegh from National Research Council (NRC)—Aerospace Manufacturing Technology Centre (AMTC) in Montréal, Canada, for his valuable contributions toward the presented approaches for robotic learning of industrial tasks using HMMs and CRFs. We are also thankful to Andrew Irish for his collaboration on the aforementioned projects conducted at NRC‐Canada. We acknowledge the support from Ryerson University for access to pertinent resources and facilities, and Natural Sciences and Engineering Research Council of Canada (NSERC) for supporting the research presented in the book. We also thank the members of the Robotics, Mechatronics and Automation Laboratory at Ryerson University for their help and support. Particular thanks go to both Dr. Abdul Afram and Dr. Shahir Hasanzadeh who provided useful comments for improving the readability of the book. Last, we would like to express our gratitude to our families for their love, motivation, and encouragement in preparing the book.