Autonomous Robotic Systems

Studies in Fuzziness and Soft Computing

Editor-in-chief Prof. Janusz Kacprzyk Systems Research Institute Polish Academy of Sciences ul. Newelska 6 01-447 Warsaw, Poland E-mail: kacprzyk@ibspan.waw.pl http://www.springer.de/cgi-bin/search_book.pl ?series= 2941

Further volumes of this series can be found at our homepage.

Vol. 95. T. Y. Lin, Y. Y. Yao and L.A. Zadeh (Eds.) Data Mining, Rough Sets and Granular Computing, 2002 ISBN 3-7908-1461-X

Vol. 96. M. Schmitt, H.-N. Teodorescu, A. Jain, A. Jain, S. Jain and L. C. Jain (Eds.) Computational Intelligence Processing in Medical Diagnosis, 2002 ISBN 3-7908-1463-6

Vol. 97. T. Calvo, G. Mayor and R. Mesiar (Eds.) Aggregation Operators, 2002 ISBN 3-7908-1468-7

Vol. 98. L. C. Jain, Z. Chen and N. lchalkaranje (Eds.) Intelligent Agents and Their Applications, 2002 ISBN 3-7908-1469-5

Vol. 99. C. Huang and Y. Shi Towards Efficient Fuzzy Information Processing, 2002 ISBN 3-7908-1475-X

Vol. 100. S.-H. Chen (Ed.) Evolutionary Computation in Economics and Finance, 2002 ISBN 3-7908-1476-8

Vol. 10 I. S. J. Ovaska and L. M. Sztandera (Eds.) Soft Computing in Industrial Electronics, 2002 ISBN 3-7908-1477-6

Vol. 102. B. Liu Theory and Practice of Uncertain Programming, 2002 ISBN 3-7908-1490-3

Vol. 103. N. Barnes and Z.-Q. Liu Knowledge-Based Vision-Guided Robots, 2002 ISBN 3-7908-1494-6

Vol. I 04. F. Rothlauf Representations for Genetic and Evolutionary Algorithms, 2002 ISBN 3-7908-1496-2

Vol. I 05. J. Segovia, P. S. Szczepaniak and M. Niedzwiedzinski (Eds.) £-Commerce and Intelligent Methods, 2002 ISBN 3-7908-1499-7

Vol. 106. P. Matsakis and L. M. Sztandera (Eds.) Applying Soft Computing in Defining Spatial Relations, 2002 ISBN 3-7908-1504-7

Vol. 107. V. Dimitrov and B. Hodge Social Fuzziology, 2002 ISBN 3-7908-1506-3

Vol. 108. L. M. Sztandera and C. Pastore (Eds.) Soft Computing in Textile Sciences, 2003 ISBN 3-7908-1512-8

Vol. 109. R.J. Duro, J. Santos and M. Grafia (Eds.) Biologically Inspired Robot Behavior Engineering, 2003 ISBN 3-7908-1513-6

Vol. 110. E. Fink Changes of Problem Representation, 2003 ISBN 3-7908-1523-3

Vol. Ill. P. S. Szczepaniak, J. Segovia, J. Kacprzyk and L.A. Zadeh (Eds.) Intelligent Exploration of the Web, 2003 ISBN 3-7908-1529-2

Vol. 112. Y. Jin Advanced Fuzzy Systems Design and Applications, 2003 ISBN 3-7908-1537-3

Vol. 113. A. Abraham, L.C. Jain and J. Kacprzyk (Eds.) Recent Advances in Intelligent Paradigms and Applications, 2003 ISBN 3-7908-1538-1

Vol. 114. M. Fitting and E. Orlowska (Eds.) Beyond Two: Theory and Applications of Multiple Valued Logic, 2003 ISBN 3-7908-1541-1

Vol. 115. J.J. Buckley Fuzzy Probabilities, 2003 ISBN 3-7908-1542-X

Changjiu Zhou Darfo Maravall Da Ruan Editors

Autonomous Robotic Systems Soft Computing and Hard Computing Methodologies and Applications

With 263 Figures and 21 Tables

Springer-Verlag Berlin Heidelberg GmbH

A Springer-Verlag Company

Prof. Dr. Changjiu Zhou Singapore Polytechnic School of Electrica! and Electronic Engineering 500 Dover Road Singapore 139651 Republic of Singapore zhoucj@ sp.edu.sg

Prof. Dr. Darfo Maravall Universidad Politecnica de Madrid Department of Artificial Intelligence Faculty of Computer Science 28660 Boadilla de! Monte, Madrid Spain dmaravall @fi. upm.es

Dr. Dr. h.c. Da Ruan Belgian Nuclear Research Centre (SCK-CEN) Boeretang 200 2400 Moi Belgium druan@ sckcen.be

ISSN 1434-9922 ISBN 978-3-7908-2523-7 ISBN 978-3-7908-1767-6 (eBook) DOI 10.1007/978-3-7908-1767-6 Cataloging-in-Puhlication Data applied for A catalog record for this book is available from the Library of Congress. Bibliographic information published by Die Deutsche Bibliothek Die Dcutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the lntemet at <http://dnb.ddb.de>.

This work is subject to copyright. AII rights are rcserved, whether the whole or part of the material is con­cerned. specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, repro­duction on microfilm or in any other way, and storagc in data banb. Duplication of lhis publication or parts thereof is permilleu only under the provisions of the German Copyright Law of Septcmbcr 9. 1965, in its CUJTent vers ion, and pennission for usc must always be obtained from Physica-Yerlag. Yiolations are liable for prosecution under the German Copyright Law.

© Springer-Verlag Berlin Heidelberg 2003 Originally published by Physica-Verlag Heidelberg in 2003.

Softcover reprint of the hardcover 1st edition 2003

The use of general descriptive names. registered names. trademarks, etc. in this publication does not imply. even in the absence of a specific statemenl, thal 'uch nume' are exempt from the relevant protectivc laws and regulations and thereforc free for general use.

Foreword

The key words soft computing, hard computing and autonomous robotics used in the title of this edited volume Autonomous Robotic Systems: Soft Computing and Hard Computing Methodologies and Applications, reflect upon the development of robotic systems and some human-like attributes such as perception and cogni­tion.

It was in 1965 when our friend and mentor, Professor Lotfi A. Zadeh, first in­troduced the notion of fuzzy sets. In 1991 he founded the Berkeley Initiative in Soft Computing (BISC).

The principle constituent methodologies of soft computing employ a number of emerging theoretical tools: fuzzy logic (FL), neural computing (NC), fuzzy-neural computing (FNC), genetic algorithms (GAs), fuzzy-genetic algorithms (FGAs), neuro-genetic algorithms (NGAs), fuzzy neural-genetic algorithms (FNGAs), evo­lutionary computing (EC), probabilistic computing (PC) and components of ma­chine learning theory (MLT).

Traditionally, theoretical tools in science have led to a better understanding of the world in which we live. In the development of these theoretical tools, we have employed many mathematical tools and concepts inherent in the natural sciences. But as we move further into the area of autonomous robotic systems, a major area of our research is the understanding of robust tasks performed by humans. This in­cludes the tasks of vision, perception, cognition, thinking, reasoning, speech un­derstanding, pattern recognition, decision-making, and reasoning and control, amongst others.

We must congratulate the editors of this volume, Dr. Changjiu Zhou, Dr. Darfo Mara vall, and Dr. Da Ruan, for bringing together this collection of research papers in this important field of Autonomous Robotic Systems. This volume contains eighteen invited chapters co-authored by 45 international researchers from nine different countries.*

These eighteen chapters are divided into four parts. In Part 1 (2 chapters), the authors deal with the development of some of the basic principles and methodolo­gies of soft computing for intelligent robotic systems. In Part 2 (6 chapters), the authors introduce some basic cognitive aspects in path planning and navigation, which lead to the design of intelligent sensory and control mechanisms. In Part 3 (7 chapters), the authors describe their work on learning, adaptation and control

*Canada (4 authors, 1 chapter); Germany (2 authors, 1 chapter); Italy (1 author, 1 chapter); Korea (1 author, 1 chapter); Singapore (8 authors, 4 chapters); Spain (16 authors, 8 chap­ters); Sweden (2 authors, 1 chapter); UK (6 authors, 2 chapters); and USA (5 authors, 1 chapter).

VI

mechanisms in an uncertain mobile environment. Finally, in Part 4 (3 chapters), the authors introduce several provocative thoughts on vision and perception - the basic sensory elements in humans. Such theoretical notions and mathematical tools may contribute significantly to the further development of robust robotic sys­tems.

With the rapidly growing research interest in the theoretical aspects of intelli­gent systems, and the increasing fields of applications of intelligent robots (aero­space, process control, ocean exploration, manufacturing and resource based in­dustry, etc.), there is a need for books that deal with their theoretical foundations, implementations and applications. I am pleased to see that the editors of this vol­ume conceived these ideas by their interactions with the intelligent systems com­munity and invited researchers in this field.

It is gratifying to acknowledge the devotion of the many researchers that has helped the exploration of new theoretical approaches, the stimulation of exchanges of scientific information, and the reinforcement of international cooperation in this important field.

It is this vision that underlines the authoritative up-to-date and reader-friendly exposition of Autonomous Robotic Systems in this book. It is a must reading for students and researchers interested in exploring the potentials of the fascinating field that will form the basis for the design of the intelligent machines of the fu­ture.

Madan M. Gupta Professor and Director

Intelligent Systems Research Laboratory College of Engineering

University of Saskatchewan Saskatoon, Saskatchewan, Canada

S7N-5A9 guptam@ sask.usaask.ca

July 2002

Preface

Autonomous Robotic Systems aim at building physical systems that can accom­plish useful tasks without human intervention and perform in the unmodified real­world situations that usually involve unstructured environments and large uncer­tainties. Hence, ideal autonomous robots should be capable of determining all the possible actions in an unpredictable dynamic environment using information from various sensors such as computer vision, tactile sensing, ultrasonic and sonar sen­sors, and other smart sensors. There are several methodologies that are capable of solving existing problems in the field of autonomous robotic systems. The well­developed field of conventional hard computing (HC) techniques offers efficient solutions to a wide variety of existing applications of robotics. However, HC tech­niques are model-based schemes that in most cases are synthesized using incom­plete information and partially known or inaccurately defined parameters. They are extremely sensitive to the lack of sensor information and to unplanned events and unfamiliar situations in the working environment. It does not seem that such techniques by themselves can cope very well with uncertain and unpredictable en­vironments all the time. On the other hand, the advent of soft computing (SC) techniques does provide us with powerful tools to solve demanding real-world problems with uncertain and unpredictable environments.

As indicated by Professor Lotfi A. Zadeh, soft computing may be viewed in two related perspectives. In one view, SC - in contrast with HC - is aimed at an accommodation with pervasive impression of the real world, exploiting the toler­ance for impression, uncertainty, and partial truth to achieve tractability, robust­ness, low solution cost, and better rapport with reality. In another view, SC is a coalition or consortium of methodologies that share this objective. At present, the principal members of the coalition are: fuzzy logic; neurocomputing; evolutionary computing; probabilistic computing; chaotic computing; and machine learning. Thanks to their strong leaning and cognitive ability and good tolerance of uncer­tainty and impression, the SC techniques have shown their great potential to solve demanding problems in the field of autonomous robotic systems. The emerging SC and conventional HC should not be viewed as competing with each other but rather as complementary. From the existing literature and successful applications, it can be concluded that SC methodologies can enhance and extend traditional HC methods, and the fusion of SC and HC techniques has provided innovative solu­tions for autonomous robotic systems. Therefore, it is interesting to gather current trends and provide a high-quality volume for scientists and engineers working in SC, HC and autonomous robotic systems areas. As we are entering a new infor­mation technological era, the fusion of HC and SC techniques will certainly play a significant role in the field of autonomous robotic systems.

VIII

This volume is in part based on the recent invited sessions at the two interna­tional conferences. One is on "Autonomous Mobile Robotics" at the 6th Interna­tional Work-Conference on Artificial and Natural Neural Networks, June 13-15, 2001, Granada, Spain. Another is on "Soft Computing and AI Methods for Autonomous Robotic systems" at the International Conference on Computational Intelligence, Robotics and Autonomous Systems, November 28-30, 2001, Singa­pore. We have also invited some known authors as well as announced a formal Call for Papers to several research groups related to SC and robotics to contribute the latest progress and recent trends and research results in this field. The primary aim of this volume is to provide researchers and engineers from both academic and industry with up-to-date coverage of new results and trend toward mobility, intelligence and autonomy in an unstructured world.

The volume is divided into four logical parts containing eighteen chapters writ­ten by some of the world's leading experts in the field of autonomous robotic sys­tems in conjunction with SC and HC methodologies.

Part 1 on Basic Principles and Methodologies contains two chapters that con­tribute to a deeper understanding of the methodologies. In the first chapter, Mira and Delgado address some methodological issues on symbolic and connectionist perspectives of Al and answer a very important question to the researchers work­ing in the area of robotics - where is knowledge in robotics? The second chapter by Oussalah explores how the fusion methodology can be decomposed into a set of primary subtasks where the elicitation and the architecture play a central role in the fusion process. A mobile robot application is given to show how the different steps of the fusion architecture have been handled.

In the second part on Planning and Navigation, various algorithms for planning and navigation as well as methods for mapping the environment of a robot are dis­cussed. In the third chapter, Tunstel et al. address computing strategies designed to enable field mobile robots to execute tasks requiring effective autonomous tra­versal of natural outdoor terrain. The primary focus is on computer vision-based perception and autonomous control. HC methods are combined with applied SC strategies in the context of three case studies associated with real-world robotics tasks including planetary surface exploration and land survey/reconnaissance. The fourth chapter by De Lope and Maravall describes a hybrid autonomous naviga­tion system for mobile robots. The control architecture proposed is highly modular and is based on the concept of behavior. In the fifth chapter, Peters et al. present a rough neurocomputing approach for line-crawling robot (LCR) navigation. The sixth and seventh chapters address some conventional HC techniques on naviga­tion and planning. Urdiales et al. propose a hybrid layered architecture, which is used to navigate in totally or partially explored environments using sonar sensors. The main advantage of the proposed scheme is that it can operate in both known and unknown environments rapidly and efficiently. The seventh chapter by Paz et al. presents an analytical method for decomposing the external environment repre­sentation task for a robot with restricted sensory information. In the eighth chap­ter, Vadakkepat et al. discuss the application of the evolutionary artificial potential

IX

field (EAPF) in mobile robot path planning. The parameters of the EAPF are op­timized with the multi-objective evolutionary algorithm.

Seven chapters on Learning, Adaptation, and Control are presented in Part 3. In the ninth chapter, Saffiotti and Wasik demonstrate the use of hierarchical fuzzy behaviors to implement a set of navigation and ball control behaviors for a Sony four-legged robot operating in the RoboCup domain. They also show that the logi­cal structure of the rules and the hierarchical decomposition simplify the design of very complex behaviors, like the "GoalKeeper" behavior. In the tenth chapter, Maravall and De Lope present a robotic mechanism, which aims at navigating in unconventional environments. The proposed method follows the perception­reason-action paradigm and is based on a reinforcement learning process guided by perceptual feedback, which can be considered as biologically inspired at the functional level. It can be straightforwardly applied to real-time collision avoid­ance for articulated mechanisms, including conventional manipulator arms. In the eleventh chapter, Hagras et al. show how intelligent embedded agents situated in an intelligent domestic environment can perform learning and adaptation. In the twelfth chapter, Wong and Ang provide a survey on the different uses of SC methods in the different aspects of legged robotics. In the thirteenth chapter, Zhang and Rossler propose a self-valuing learning system based on continuous B­spline model, which is capable of learning how to grasp unfamiliar objects and how to generalize the learned abilities. The fourteenth chapter by Er and Gao pre­sents a robust Adaptive Fuzzy Neural Controller that is suitable for identification and control of a class of uncertain Multi-Input-Multi-Output (MIMO) nonlinear systems. In the fifteenth chapter, Er and Sun propose a new approach towards op­timal design of a hybrid fuzzy proportional-integral-derivative (PID) controller for robotics systems using the genetic algorithm.

In the last part on Vision and Perception for autonomous robotic systems, the sixteenth chapter by Balsi and Vilasfs-Cardona shows how Cellular Neural Net­works (CNNs) can provide the necessary image processing to guide an autono­mous mobile robot in a maze made of black lines on a light surface. The system consists of a fuzzy controller performing the elementary navigation tasks fed by the result of processing the image only by CNN techniques. In the following chap­ter by Camacho et al., the authors address multiresolution vision in autonomous systems. Multiresolution systems are one alternative to cover wide fields of view without involving high data volumes and, therefore, considerably reduce the con­straints imposed by off-the-shelf uniresolution vision systems. In the last chapter, Buenaposada and Baumela focus on the real-time location and tracking of human faces in video sequences. The tracking is based on the cooperation of two low­level trackers from colour and template information. As a result of the co­ordination of these two trackers, it emerges a robust real-time tracker that accu­rately computes face position and orientation in varying environmental conditions.

This volume highlights the advantages of fusion of SC and HC methodologies and applications to autonomous robotic systems. Each chapter is self-contained and also indicates the future research direction on the topic of autonomous robotic systems.

X

We would like to thank Professor Madan Gupta, University of Saskatchewan, Canada, for his willingness to write a foreword for this volume; to Professor Janusz Kacprzyk, Editor-in-Chief of the Book Series "Studies in Fuzziness of Soft Computing", for his kind acceptance to publish this volume; to all the contributors for their kind cooperation to this book; to Hnin Wai Yin and Phyu Phyu Khing, fi­nal-year-project students of Singapore Polytechnic, for their editorial assistance, and to Katharina Wetzel-Vandai and Judith Kripp ofPhysica-Verlag for their ad­vice and help during the production phases of this book.

Changjiu Zhou, Singapore Polytechnic, Singapore Dario Maravall, Universidad Politecnica de Madrid, Spain

Da Ruan, The Belgian Nuclear Research Centre (SCK•CEN), Belgium

July 2002

Contents

Foreword M.M. Gupta

Preface C. Zhou, D. Maravall and D. Ruan

v

vii

Part 1: Basic Principles and Methodologies 1

Where is Knowledge in Robotics? Some Methodological Issues on Symbolic and Connectionist Perspectives of AI 3

J. Mira and A. E. Delgado

Introduction to Fusion Based Systems - Contributions of Soft Computing Techniques and Application to Robotics 35

M. Oussalah

Part 2: Planning and Navigation 73

Applied Soft Computing Strategies for Autonomous Field Robotics 75 E. Tunstel, A. Howard, T. Huntsberger, A. Trebi-Ollennu and J. M. Dolan

Integration of Reactive Utilitarian Navigation and Topological Modeling 103

]. de Lope and D. Maravall

Line-Crawling Robot Navigation: A Rough Neurocomputing Approach 141

J.F. Peters, T.C. Ahn, M. Borkowski, V. Degtyaryov and S. Ramanna

XII

Hierarchical Planning in a Mobile Robot for Map Learning and Navigation 165

C. Urdiales, A. Bandera, E. Perez, A. Poncela and F. Sandoval

An Analytical Method for Decomposing the External Environment Representation Task for a Robot with Restricted Sensory Information 189

F. de la Paz, J.R. Alvarez and J. Mira

Evolutionary Artificial Potential Field - Applications to Mobile Robot Path Planning 217

P. Vadakkepat, T.H. Lee and L. Xin

Part 3: Learning, Adaptation and Control 233

Using Hierarchical Fuzzy Behaviors in the RoboCup Domain 235 A. Saffiotti and Z. Wasik

A Bio-Inspired Robotic Mechanism for Autonomous Locomotion in Unconventional Environments 263

D. Maravall and J. de Lope

Online Learning and Adaptation for Intelligent Embedded Agents Operating in Domestic Environments 293

H. Hagras, V. Callaghan, M. Colley, G. Clarke and H. Duman

Integration of Soft Computing Towards Autonomous Legged Robots 323 A. Wong and M.H. Ang Jr

Grasp Learning by Active Experimentation Using Continuous B-Spline Model 353

J. Zhang and B. Rossler

Online Adaptive Fuzzy Neural Identification and Control of Nonlinear Dynamic Systems 373

M.J. Er and Y. Gao

Hybrid Fuzzy Proportional-Integral plus Conventional Derivative Control of Robotics Systems 403

M.J. Er and Y. L. Sun

Part 4: Vision and Perception

Robot Vision Using Cellular Neural Networks M. Balsi and X. Vilas{s-Cardona

Multiresolution Vision in Autonomous Systems P. Camacho, F. Arrebola and F. Sandoval

A Computer Vision Based Human-Robot Interface J. M. Buenaposada and L. Baumela

Subject Index

Contributors

XIII

429

431

451

471

493

497