Robot Vision: Strategies, Algorithms and Motion Planning

489 Pages·2009·1.98 MB·English

by  Daiki Ito

R V : S ,
OBOT ISION TRATEGIES
ALGORITHMS AND MOTION PLANNING

No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.

R V : S ,
OBOT ISION TRATEGIES
ALGORITHMS AND MOTION PLANNING

DAIKI ITÔ
EDITOR

Nova Science Publishers, Inc.
New York
Copyright © 2009 by Nova Science Publishers, Inc.

All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.

For permission to use material from this book please contact us:
Telephone [phone]; Fax [phone]
Web Site: http://www.novapublishers.com

NOTICE TO THE READER
The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or
implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.

Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.

This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

Robot vision : strategies, algorithms, and motion planning / Daiki Itô, Editor.
p. cm.
ISBN 978-1-61668-981-0 (E-Book)
1. Robot vision. 2. Robotics. I. Itô, Daiki.

TJ211.3.R53 2009
629.8'92637--dc22
2008037508

Published by Nova Science Publishers, Inc. (cid:212) New York
CONTENTS

Preface vii
Short Communication
A Building Algorithm Based on Grid-Topological Map for
Mobile Robot Navigation 1
Bing Xu and Nan-Feng Xiao
Chapter 1 Mobile Robot Navigation Strategies Oriented to Goal
Achievement with Local Monocular Data 9
Lluís Pacheco, Ningsu Luo and Xevi Cufí
Chapter 2 Latest Developments in Feature Based Mapping and Navigation
for Indoor Service Robots 123
Diego Rodriguez-Losada, Luis Pedraza, Fernando Matia and
Pablo San Segundo
Chapter 3 Robot Navigation Using Parameterized Growing Neural Gas
that Learns Topologic Maps 171
José García-Rodríguez, Francisco Flórez-Revuelta and
Juan Manuel García-Chamizo
Chapter 4 Topological Knowledge Extraction Using Self-Organising
Neural Networks 185
P. Baldassarri, A. Montesanto and P. Puliti
Chapter 5 The SWAN, Submarine Wobble-free Autonomous Navigator,
Project: The Compensated Vectored-Thrust Patroller 211
R. C. Michelini, E. Cavallo and R. P. Razzoli
Chapter 6 Autonomous Robot Navigation Using Different Features and
Hierarchical Discriminant Regression 279
Xiaodong Gu
Chapter 7 Advances in Mapless and Map-Based Strategies for Robot
Navigation -- A Review 301
Leena Vachhani , Panakala Rajesh Kumar and K. Sridharan
vi Contents
Chapter 8 Dynamics, Motion Planning and Control of Wheel-Legged
Robots for Superior Field Navigations 325
S. Ali A. Moosavian and Asghar Mozdbaran
Chapter 9 Artificial Neural Networks for Learning and Decision Making
in Multi-agent Robot Soccer Systems 355
K.G. Jolly, R. Sreerama Kumar and R. Vijayakumar
Chapter 10 Accurate Camera Calibration for 3d Data
Acquisition: A Comparative Study 383
Junjie Liu, Ningfeng Wei and Yonghuai Liu
Chapter 11 Real-Time Map Building and Active Exploration for
Autonomous Robot in Unknown Environments 421
Meng Wang and James N. K. Liu
Chapter 12 Teleoperation of Mobile Robots via Internet:
Guidelines of Design 445
Emanuel Slawiñski and Vicente A. Mut
Index 461
PREFACE

Short Communication - To efficiently carry out complex missions in an indoor
environment, an autonomous mobile robot must be able to acquire and maintain models of the
environment. This paper presents a building algorithm that integrated both a grid-based
algorithm and a topological map algorithm. Grid-Topological maps are learned by using a BP
neural network, the control system of the mobile robot adopts a hybrid architecture. On the
high level, a FSM-based behavior selection method realizes behavior planning of the mobile
robot; on the low level, a behavior-based technique controls actions of the mobile robot. The
mobile robot can coordinate multi-behavior and react rapidly. Simulation results show the
effectiveness of the algorithm.
Chapter 1 - The research presented here integrates control science and robot vision
knowledge in a computer science environment. The perception of the navigation environment
is based on sensor systems that provide distance measurements in the vicinity of the robot.
This essential task can be accomplished by different range systems such as ultrasonic sensors,
laser rangefinders and vision-based systems. Among them, the computer vision-based system
is one of the most attractive and useful sensing method. It presents some interesting aspects
such like its falling down price and capacity of providing richer information than other
traditional ranging devices. The monocular techniques considered are Depth from Focus
(DFF) and Optical Flow (OF); the methodology presented exploits benefits from both
approaches. Moreover odometer system information is locally considered in order to correlate
the acquired frames. Among many objectives in the mobile robot navigation, it is very
important to achieve the feasible and accurate trajectory planning. The navigation control
signals should include the policy of obstacle avoidance, as well as the final desired coordinate
achievement. These mobile robot navigation approaches propose the use of potential fields,
and the use of model predictive control (MPC) on a short prediction horizon. Real Time MPC
control laws are analyzed by considering several constraints. In this work, trajectory tracking
is pursued by using real time mobile robot MPC implementations with monocular local vision
sensing. This article will be organized as follows. First, it is presented the useful
methodologies relevant to the present research work, especially on the mobile robot
navigation based on computer vision, monocular 3D machine vision techniques and model-
based control strategies. Experimental indoor results obtained with the available lab platform,
consisting in a differential-driven robot with a free rotating wheel, are depicted. Then, it is
presented the research developed and results obtained on the predictive model-based control
viii Daiki Itô
by using an occupancy grid framework obtained by monocular perception techniques and
odometer system data. Finally, conclusions are drawn and further research works are
commented.
Chapter 2 - Mobile service robots typically require an internal representation of its
environment to successfully accomplish its tasks. Such an internal representation (map) can
be provided to the robot in some scenarios where such a priori man made map exists.
Nevertheless there are many cases where such a map is not known in advance, so it must be
built by the robot itself while exploring the environment, which is known as the Simultaneous
Localization and Mapping (SLAM) problem. In our case, such necessity aroused several
years ago in the deployment of our interactive tour guide robot in new exhibits, museums or
fairs, where the maps provided by the organizers were completely useless for navigation. The
authors describe in this work our steps towards the current navigation system of our robot
Urbano, including the feature based mapping algorithm based on the SPMap concepts, our
solutions to the inconsistency and computational complexity of SLAM based on an Extended
Kalman Filter (EKF), and our recent research to perform advanced modeling of curved
environments using BSplines. The authors also describe our navigation system based on the
resulting maps: path planning and execution, reactive control and feature based localization.
Chapter 3 - Self-organising neural networks try to preserve the topology of an input space
by means of their competitive learning. This capacity has been used, among others, for the
representation of objects and their motion. In this work the authors use a kind of self-
organising network, the Growing Neural Gas, to represent different objects shape. As a result
of an adaptive process the objects are represented by a topology representing graph that
constitutes an induced Delaunay triangulation of their shapes. This feature can be used to
learn and represent topologic maps that mobile devices use to navigate in different
environments.
Chapter 4 - This paper proposes a method for self-localising a mobile agent, using images
supplied by a single monodirectional camera. The images acquired by the camera may be
viewed as an implicit topological representation of the environment. This last is a priori
unknown and its topological representation is derived by an unsupervised neural architecture.
The creation of a right topologically map allows overcoming the “perceptual aliasing”
problem: for which different locations in the environment can look identical, because they
produce similar or the same sensor readings. In fact when there is uncertainty, the agent is
able to identify its current position knowing the last visited node. The self-localization
process in conjunction with the built topological map allow to quite correctly establish the
agent position. To compare the performances of two self-organising networks, a given
architecture has been realized considering both Self-Organising Maps (SOM) and Growing
Neural Gas (GNG). Extensive simulations are provided to compare the effectiveness of the
GNG and SOM models in recognition speed, classification tasks and topology preserving.
The sperimental evidences show that both SOM and GNG have been particularly successful
in data recognition tasks, although the GNG model assures the best results and in a shortest
time. The GNG representation is able to make explicit the important topological relations in a
given distribution of input signals. So, the system not only optimally and quickly recognises
the images in each region, but it is able to spontaneously reconstruct the topological map of
the environments.
Chapter 5 - The chapter deals with developing small and cheap autonomous underwater
vehicles, AUV, entrusted of extended manoeuvrability for surveying and docking missions,