La lecture en ligne est gratuite
Le téléchargement nécessite un accès à la bibliothèque YouScribe
Tout savoir sur nos offres
Télécharger Lire

Interactive acquisition of spatial representations with mobile robots [Elektronische Ressource] / Fang Yuan. Technische Fakultät - AG Angewandte Informatik

De
164 pages
Publié par :
Ajouté le : 01 janvier 2011
Lecture(s) : 11
Signaler un abus

Interactive Acquisition of Spatial
Representations with Mobile Robots
Fang YuanM.Sc. ComputerSc. Fang Yuan
AG Angewandte Informatik
Technische Fakultät
Universität Bielefeld
email: meekfang@hotmail.com
Abdruck der genehmigten Dissertation zur Erlangung
des akademischen Grades Doktor-Ingenieur (Dr.-Ing.).
Der Technischen Fakultät der Universität Bielefeld
am 12.04.2011 vorgelegt von Fang Yuan,
am 16.09.2011 verteidigt und genehmigt.
Gutachter:
Dr.-Ing. Marc Hanheide, Universität Bielefeld
Assistant Prof. Patric Jensfelt, Kungliga Tekniska Högskolan
Prof. Dr.-Ing. Franz Kummert, Universität Bielefeld
Prüfungsausschuss:
Prof. Dr. Barbara Hammer, Universität Bielefeld
Dr.-Ing. Marc Hanheide,
Prof. Dr.-Ing. Franz Kummert, Universität Bielefeld
Dr. Christina Unger, Universität Bielefeld
Gedruckt auf alterungsbeständigem Papier nach ISO 9706InteractiveAcquisitionofSpatial
RepresentationswithMobileRobots
DerTechnischenFakultätderUniversitätBielefeld
zurErlangungdesGrades
Doktor-Ingenieur
vorgelegtvon
FangYuan
Bielefeld–April2011iii
Abstract
Robots that are able to provide services to human users in personal context are known as
personal service robots. They are attracting more and more attention, as enriching and
easing people’s lives are their general missions. Acquisition of spatial representations is
crucial for such a robot to move around in a human-shared environment, provide services
and interact with the user about the environment. With methods developed in the field of
autonomous exploration it is possible for a mobile robot to build a metric representation
from sensor data, while exploring an initially unknown environment. However, due to the
limited perceptual abilities of the robot and dynamic changes in the environment over time,
challenges still pose for robots. Besides, in order to facilitate the communication about the
environment, a mapping between the spatial understanding of the user and the robot has
to be built in a service context.
To acquire a spatial representation, exploration strategies are necessary for the robot to
move around in the environment and gain information from the surroundings. The term
Human-Robot Joint Spatial Exploration is used for a framework that allows the human user
to influence the robot’s exploration of the environment and the environment representation
built by the robot. By interactively gathering information, individual spatial knowledge
from the user’s point of view is expected to be integrated into the built environment repre-
sentation. Therefore, how to make the robot interactively learn from the human user and
build a spatial representation incorporating the gathered information is the central problem
of this thesis. A Home-Tour is assumed as the initial scenario for information gathering. In
the interactive guided tour the user presents the environment to the robot, while the robot
learns various models of the environment with the help of the user. A link between the
spatial representation built by the robot and human understanding of the environment is
able to be created.
The major contribution of this thesis is a framework that enables a mobile robot to inter-
actively acquire an environment representation. We first present the interactive Home-Tour
scenario assumed as the starting point of the integrated solution to the central problem.
Benefits and challenges of this solution are discussed as well. Then, a hybrid person-
following behavior that facilitates the joint exploration of the environment for the robot
and the guide-person is described. The robot is able to choose an appropriate behavior ac-
cording to strategies designed on the basis of spatial context and relative person’s positions
with respect to the robot. Afterwards, a metric map of the surroundings can be built by the
robot during the joint exploration. Because of the guide-person perceived by the robot in
the Home-Tour, the violation of the static world assumption used in the mapping process
has to be pointed out. Consequently, we provide a technique to reduce the negative im-
pact on the mapping process, resulting from the guide-person. Finally, a topological graph
built upon the metric representation is discussed. Human navigation strategies and spatial
concepts achieved by the robot in the interactive learning process are incorporated in the
graph-based environment representation.
Based on the idea of interactive information gathering, an integrated system is imple-
mented. Experimental results achieved from individual modules validate their effective-
ness respectively. By means of a user study, the robustness and practicality of the system
consisting of these modules are verified as a whole in real world scenarios.
Bielefeld Universityv
Acknowledgments
It is a pleasure to thank those who made this thesis possible. First of all, I would like
to thank Gerhard Sagerer and Franz Kummert who gave me a chance to show that a for-
eign student from a university of applied sciences in Germany is competent to complete a
doctoral thesis. Their trust is a great encouragement for me throughout these years.
Looking back my work as a whole I owe my deepest gratitude to my supervisor Marc Han-
heide who was always able to help me out of a difficult position. With his comprehensive
expertise, enthusiasm, and patience he inspired and influenced my work at the most. When
I was frustrated by the work, he tried to help me stand up again. I would never forget your
help and attention when I fell ill, your encouragement when my paper submission failed at
the first time, long time discussion before my presentation in Osaka, working together in
Graz. I am heartily thankful to you, Marc, my supervisor, my friend and my brother! I am
also grateful that Patric Jensfelt would like to agreed to review the thesis.
Furthermore, I want to thank all my colleagues in the Applied Informatics. The harmonious
working atmosphere and your friendly supports make the last years in Bielefeld a really
good experience. Especially I would like to thank Thorsten Peter Spexard who gave me the
best overview of the robot platform, Jan Moringen who would like to discussed with me for
trivial problems that were actually far from his own work and Lars Schillingmann who can
give me the best technical support.
Of course, I need to thank my bachelor students Lukas Twardon and Anton Helwart, as well
as my diploma student Orhan Engin. It is my pleasure to supervise and benefit from their
work. In particular I want to thank again Lukas Twardon who accompanied my research
as a student assistant as well. For grammar correction of my thesis I would like to thank
Sengupta, Debanjan.
Above all I thank my parents, who have always supported me and helped me to find my way
anytime and anywhere. I owe my deepest gratitude to my wife Xixi. You have been always
considerate to me, tried your best to support me and exclude the difficulty and anxiety for
me. Of course, I thank Meng, our lovable daughter. You have always brought happiness
and hope to us.
Bielefeld UniversityContents vii
Contents
1. Introduction 1
1.1. Home-Tour with a Mobile Robot . . . . . . . . . . . . . . . . . . . . . . . 3
1.2. Background Knowledge and Related Work . . . . . . . . . . . . . . . . . 7
1.3. Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4. Organization of this Thesis . . . . . . . . . . . . . . . . . . . . . . . . . 15
2. Person-Following 17
2.1. Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2. Person Tracking and Motion Control . . . . . . . . . . . . . . . . . . . . . 19
2.2.1. Multi-Modal Person Tracking . . . . . . . . . . . . . . . . . . . . . 20
2.2.2. Motion Control . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3. Adaptive Context-Aware Following Behavior . . . . . . . . . . . . . . . . 22
2.3.1. Core Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4. Elementary Person-Following Behaviors . . . . . . . . . . . . . . . . . . . 26
2.4.1. General Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.2. Direction-Following . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4.3. Path-Following . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.4. Parallel-F . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3. Map Building in a Home-Tour 41
3.1. Introduction of SLAM with Rao-Blackwellisation . . . . . . . . . . . . . . 41
3.2. Mapping during the Home-Tour . . . . . . . . . . . . . . . . . . . . . . . 45
3.2.1. Basic Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2.2. Evaluation on a Simulation Platform . . . . . . . . . . . . . . . . . 47
3.3. Summary and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4. Path Planning Adopting Human Strategies 59
4.1. Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2. Path Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.1. Graph Creation . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.2. Re-planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.3. Empirical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3. Multisensor-based Motion Generation . . . . . . . . . . . . . . . . . . . . 73
4.3.1. Motion Generation . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.3.2. Data Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.3.3. Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.4. Summary and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5. Evaluation 89
5.1. Experimental Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.1.1. Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.1.2. System Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 93
Bielefeld Universityviii Contents
5.2. Assessment of Individual Modules in the Integrated System . . . . . . . . . 99
5.2.1. Person-Following . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.2.2. Map Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.2.3. Path Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.3. Observations from a Subject’s Perspective . . . . . . . . . . . . . . . . . . 123
5.4. Insights and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6. Conclusion and Future Considerations 129
6.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.2. Concluding Discussion and Future Considerations . . . . . . . . . . . . . . 130
A. A Brief Introduction to Particle Filters 135
Bibliography 139
Fang Yuan

Un pour Un
Permettre à tous d'accéder à la lecture
Pour chaque accès à la bibliothèque, YouScribe donne un accès à une personne dans le besoin