Concept of Intelligent Mechanical Design

for Autonomous Mobile Robots

Amir Ali Forough Nassiraei

Supervisor Prof. Kazuo Ishii

Department of Brain Science and Engineering

Graduate School of Life Science and System Engineering
Kyushu Institute of Technology

March 2007

ii
Acknowledgments

Many people have helped me along the way. Their guidance, good humor, advice
and inspiration sustained me trough the years of work. First of all, I’d like to thank
all of them. I wish to express my sincerest appreciation to Prof. Kazuo ISHII who
provided the chance to study in Japan, for his outstanding and endless support during
the many years of research. Professor ISHII’s invaluable comments, suggestion, and
encouragement have been of the greatest help in this research.
I wish to express my deepest sense of indebtedness to the members of the examina-
tion committee, Prof. Masumi ISHIKAWA, Prof. Iwan GODLER, and Prof. Hiroyuki
MIYAMOTO for their valuable comments.
My thanks also go to the FAIS-Robotics Development Support Office and the for-
mer organization, GMD-Japan research institute for their deep insight and unfailing
enthusiasm and constant support.
I gratefully acknowledge the Small & Medium Enterprises and Regional Innovation,
JAPAN (SMRJ) and New Energy and Industrial Technology Development Organiza-
tion (NEDO) for providing the international funding and facility which enabled me to
pursue this research. I thankfully acknowledge the cooperation with project partners:
Yaskawa Electric Co., Kyushu Keisokuki Co., Nihon Tecmo Ltd., Mitsuwa Ltd., and
the Kitakyushu Foundation for the Advancement of Industry Science and Technology
(FAIS).
Furthermore, I wish to thanks my colleagues in ISHI laboratory especially Ohata
Satomi and Robocop team members for their unlimited help during various stages of
the research work.
Expression of graduate and apology are directed to my family whose encourage-
ments have been the most influence support during the whole course of my education.

i
Abstract

During the 21st century, it is expected that the robots with different degrees of
autonomy and mobility will play an increasingly important role in all side of human
life. Thus these kinds of robots will become much more complex than today, and
the development of such robots present a great challenge for researchers. However,
drawbacks of robot complexity, necessity of more complex hardware, software and
mechanical structure may lead to low reliability and increasing of size, weight, cost,
power consumption and motion limitation. In order to avoid the problems mentioned
above, the simplification of robots mechatronics will be a critical point in their design.
In this thesis, the concept of “Intelligent Mechanical Design” will be presented and
how a mechanical structure can be designed to have an effect to the robot controlla-
bility, simplification and its tasks performance is discussed. The description of this
concept will lead us to establish the landmarks of the territory of mechanical designing
in the form of seven design principles. The design principles, named “Mecha-telligence
principles”, provide a guidance on how to design autonomous mobile robots mechan-
ics. These principles guide us in asking the right questions when investigating issues
concerning a self-controllable, reliable, realizable, and compatible mechanics for au-
tonomous mobile robots.
To show how the “Mcaha-telligence principles” can be applied on the processes of
the design of robots mechanics, we proposed a novel methodology, named "Mecha-
telligence methodology”. Mechanical design in the proposed methodology is performed
based on preference of classification of the robot specification described by interaction
of the robot with its environment and, also, the physical parameters of the robot
mechatronics itself. In this approach the robot specification is classified to a set of
high-level- and low-level-specification which can be expressed as tasks and physical
parameters of the robot, respectively. A main goal in these analyzing process of the
robot specification, is that, figuring out to a Mono-spec layer (the last layer of low-level-
specification) including a simple basic function or selecting a sensor for a single task
or behavior. The process of mechanical design will be started based on an important
defined robot mono-spec, and then will be proceeded to realize the other mono-specs

ii
by adding the minimum actuator(s) and sensor(s). In each step of the mechanical
design in this methodology, the proper and sufficient mechanics will be achieved by
considering the robot tasks, behaviors, morphology, and by applying the description of
environmental and physical-morphological constraints on the design process.
To prove the validity of the proposed method we concentrate on developing of the
mechanical structures and devices of three types of autonomous mobile robots used
in the three different educational, entertainment and industrial applications. The first
project is on developing an autonomous mobile soccer robot, “Musashi robot”, which
has a fully mechatronics modular architecture including a strong ball-kicking device
with capability of lifting a ball and a ball-holding mechanism, aiming at getting cham-
pion in RoboCup Midsize League. The second challenge is on the design, modeling,
simulation and implementation of a series of entertainment robots, named "Jumping
Joe"s, with capabilities of the fast waking up, jumping, and somersault actions, ex-
hibited in Aichi Exposition 2005. In third project we propose a new approach to the
sewer pipe inspection by developing a fully autonomous mobile robot “KANTARO”.
Cooperation in this national funding project, is our third challenge item in mechan-
ical design to compromise how a passive-active intelligent moving mechanism can be
employed as a robot platform for inspecting the real sewer pipe network. In these
three projects we describe, not only, the mechanical design process based on “mecha-
telligence design” methodology, but also, the detail design which is considered to lead
readers to get familiar as much as possible to our vision relative to the presented design
methodology.

iii
Contents

1 Introduction 2
1.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Overview of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Mechanics & Intelligence 8

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 What is Intelligence? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Definitions of The Necessary Selected Terms . . . . . . . . . . . . . 10
2.2.2 Definition of Task and Desired Behaviors . . . . . . . . . . . . . . 13
2.2.3 Design principles for autonomous agents . . . . . . . . . . . . . . 14
2.3 Intelligent Mechanical Design . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Mechanical Design Principles: “Mecha-telligence Principles” . . . . . . . . 20
2.5 Mechanical Design Process: “Mecha-Telligence Methodology” . . . . . . . 23

3 Soccer Robot 27
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Summary of design based on “Mecha-telligence methodology” . . . . . . . 28
3.2.1 Description of the basic terms . . . . . . . . . . . . . . . . . . . . . 28
3.2.2 Simplification process of design functions . . . . . . . . . . . . . . 30
3.2.3 Design process: Applying the “mecha-telligence principle” . . . . . 33
3.3 Concept of Designing “Musashi” Robot . . . . . . . . . . . . . . . . . . . . 36
3.4 “Musashi” robot architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4.1 “Omni-directional” concept . . . . . . . . . . . . . . . . . . . . . . 38
3.4.2 “Modularity” concept . . . . . . . . . . . . . . . . . . . . . . . . . . 39

137

iv
CONTENTS 138

3.5 Strong novel kicking device . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.5.1 Overview of the Different Types of Kicking Device . . . . . . . . . 44
3.5.2 Design of New Kicking Device . . . . . . . . . . . . . . . . . . . . . 49
3.6 Conclusion and Experimental Result . . . . . . . . . . . . . . . . . . . . . 56

4 Artistic Robot 59
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2 Mecha-telligence methodology . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.1 Description of the basic terms . . . . . . . . . . . . . . . . . . . . . 60
4.2.2 Simplification process of design functions . . . . . . . . . . . . . . 62
4.2.3 Design process: Applying the “Mecha-telligence principle” . . . . . 64
4.3 Mechanical Design of “Jumping Joe”s . . . . . . . . . . . . . . . . . . . . . 66
4.3.1 Inertia Actuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3.2 Hip Joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3 Knee Joint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.4 Jumping Foot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4 Electric Circuit Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.4.1 Total System Architecture . . . . . . . . . . . . . . . . . . . . . . . 78
4.4.2 Control Board of Each Actuator . . . . . . . . . . . . . . . . . . . 79
4.5 Modular Architecture of “Jumping Joe”s . . . . . . . . . . . . . . . . . . . 79
4.6 Simulation and Experimental Results . . . . . . . . . . . . . . . . . . . . . 81
4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5 Sewer Pipe Inspection Robot 85

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1.1 State-of-The-Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.1.2 Qualitative Degrees of Autonomy in Sewer Robots . . . . . . . . . 87
5.1.3 Sewer Inspection Robots . . . . . . . . . . . . . . . . . . . . . . . . 88
5.2 Summary of design based on “Mecha-telligence methodology” . . . . . . . 95
5.2.1 Simplification process of design functions . . . . . . . . . . . . . . 98
5.2.2 Design process: Applying the “Mecha-telligence principle” . . . . . 101
5.3 A New Approach to the Autonomous Sewer Pipe Inspection . . . . . . . 105

v
CONTENTS 139

5.3.1 An Overview to the Sewer Pipe System in Japan . . . . . . . . . . 105

5.3.2 Basic Concepts of Design . . . . . . . . . . . . . . . . . . . . . . . 106
5.4 Novel Sewer Pipe Platform: naSIR Platform . . . . . . . . . . . . . . . . 107
5.4.1 Drawbacks of Current Commercial Sewer Inspection Robots . . . . 107
5.4.2 Novel Moving Mechanism (naSIR Mechanism) . . . . . . . . . . . . 112
5.4.3 naSIR Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.5 Autonomous Sewer Inspection Robot: KANTARO . . . . . . . . . . . . . 117
5.5.1 KANTARO Architecture . . . . . . . . . . . . . . . . . . . . . . . . 118
5.5.2 KANTARO Hardware . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.5.3 KANTARO Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.5.4 KANTARO as A Fully Autonomous Sewer Inspection System . . . 122
5.6 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

6 Conclusion 129

vi
List of Figures

2.1 Mechanical design of an autonomous mobile robot and related factors. . . 8

2.2 New approach to mechanical design of an autonomous mobile robot. . . . 9
2.3 A sample of a complete agent, Garbage robot. . . . . . . . . . . . . . . . . 11
2.4 A sample of a complete agent, Garbage robot. . . . . . . . . . . . . . . . . 11
2.5 Ping pong collecting robot. . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.6 Overview of design principles for autonomous agents. . . . . . . . . . . . . 14
2.7 First example of garbage-collector robot. . . . . . . . . . . . . . . . . . . 16
2.8 Second example of garbage-collector robot. Cost limitation is assumed. . 17
2.9 Third example of garbage-collector robot. Size limitation is assumed. . . . 18
2.10 Forth example of garbage-collecting robot. . . . . . . . . . . . . . . . . . 19
2.11 Overview of mechanical design principles for autonomous mobile robots. . 21
2.12 Overview of ”Mecha-telligence methodology”. . . . . . . . . . . . . . . . . 24

3.1 The definitions of environmental niche (Soccer robot). . . . . . . . . . . . 29

3.2 The selected type of soccer robot. . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 The high-level specification layers (Soccer robot). . . . . . . . . . . . . . . 30
3.4 Layer 1 generated from the sub-layer (Soccer robot). . . . . . . . . . . . . 31
3.5 Layer 2 generated from the layer 1 (Soccer robot). . . . . . . . . . . . . . 32
3.6 Mono-spec layer generated from the layer 2 (Soccer robot). . . . . . . . . . 32
3.7 The function design priority (Soccer robot). . . . . . . . . . . . . . . . . . 33
3.8 The definitions of environmental niche (Kicking device). . . . . . . . . . . 34
3.9 The results of design (Soccer robot). . . . . . . . . . . . . . . . . . . . . . 36
3.10 The first version of “Hibikino-Musashi” team robot. . . . . . . . . . . . . . 37
3.11 “Musashi” robot, designed by Autodesk Inventor 3D-CAD. . . . . . . . . . 38

140

vii
LIST OF FIGURES

3.12 Flowchart of “Musashi” robot power system. . . . . . . . . . . . . . . . . . 40

3.13 System architecture of “Musashi” robot. . . . . . . . . . . . . . . . . . . . 41
3.14 Motor driver module and USB module. . . . . . . . . . . . . . . . . . . . . 42
3.15 Modules of “Musashi” robot hardware. . . . . . . . . . . . . . . . . . . . . 43
3.16 Modules of “Musashi” robot. . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.17 Schema of a solenoid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.18 Three alternatives for making a solenoid kicking device. . . . . . . . . . . 45
3.19 Two alternatives for making a pneumatic kicking device. . . . . . . . . . . 46
3.20 Basic operation of an elastic kicking device. . . . . . . . . . . . . . . . . . 47
3.21 Schema of Philips kicking device. . . . . . . . . . . . . . . . . . . . . . . . 48
3.22 Amount of applied force by the plate of Philips kicking device. . . . . . . 48
3.23 Kicking device of KIRC team (KIT, Izuka, Japan) . . . . . . . . . . . . . 49
3.24 “Musashi” robot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.25 New kicking device space. . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.26 Three possibility to design a new kicking device. . . . . . . . . . . . . . . 51
3.27 Basic concept of cam charger mechanism. . . . . . . . . . . . . . . . . . . 52
3.28 The coordinate and relation between the cam and the torsion spring. . . . 53
3.29 Motor torque τC (blue graph) and spring torque (orange graph). . . . . . 53
3.30 The sequence of charging of springs. . . . . . . . . . . . . . . . . . . . . . 54
3.31 Ball-lifting mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.32 Ball-holding mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.33 Control sequences of the ball-lifting mechanism. . . . . . . . . . . . . . . . 56

4.1 Design of full function Jumping Joe. . . . . . . . . . . . . . . . . . . . . . 60

4.2 The definitions of environmental niche (Artistic robot). . . . . . . . . . . . 61
4.3 The selected type of artistic robot. . . . . . . . . . . . . . . . . . . . . . . 62
4.4 The high-level specification layers (Artistic robot). . . . . . . . . . . . . . 62
4.5 The layer 1 generated from the sub-layer (Artistic robot). . . . . . . . . . 63
4.6 The mono-spec layer generated from the layer 1(Artistic robot). . . . . . . 63
4.7 The function design priority (Artistic robot). . . . . . . . . . . . . . . . . 64
4.8 Two alternatives for fast wake-up action. . . . . . . . . . . . . . . . . . . . 65
4.9 The results of design (Artistic robot). . . . . . . . . . . . . . . . . . . . . . 65

viii
LIST OF FIGURES

4.10 Inertia Actuator architecture. . . . . . . . . . . . . . . . . . . . . . . . . . 66

4.11 The brake mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.12 Necessary geometric parameters of the brake mechanism. . . . . . . . . . . 67
4.13 The experimental results of the Inertia Actuator. . . . . . . . . . . . . . . 68
4.14 Braking time for the different rotation speed of the rotor. . . . . . . . . . 69
4.15 The developed 2 DOF hip joint. . . . . . . . . . . . . . . . . . . . . . . . . 70
4.16 The developed hip joint consists of 3 serial-link arms. . . . . . . . . . . . 70
4.17 The coordinate system and parameters of developed hip joint. . . . . . . 71
4.18 The necessary geometric parameters of Lever-rank mechanism. . . . . . . 72
4.19 The developed knee joint using lever-crank mechanism. . . . . . . . . . . 73
4.20 The relation between the angle of knee joint and the obtained torque. . . 74
4.21 Basic concept of Cam Charger mechanism. . . . . . . . . . . . . . . . . . 74
4.22 The coordinate and relation between the cam and the torsion spring. . . 75
4.23 Motor torque τC (red graph) and spring torque (blue graph). . . . . . . . 76
4.24 Cam Charger mechanism was designed to fit to the foot shape. . . . . . . 77
4.25 The cam design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.26 The experimental results of the Cam Charger jumping. . . . . . . . . . . 78
4.27 Electric system architecture of Jumping Joe. . . . . . . . . . . . . . . . . 79
4.28 The developed control boards. . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.29 “Jumping Joe" robots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.30 Upper body module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.31 Simulation and experimental results of Junior 1 (rapid wake up). . . . . . 81
4.32 Simulation and experimental results of Junior 1 (samersault). . . . . . . . 82
4.33 Simulation and experimental results of Junior 2 (samersault). . . . . . . . 82

5.1 Current conventional sewer inpection method. . . . . . . . . . . . . . . . . 86

5.2 KURT2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.3 MAKRO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.4 KARO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.5 PIRAT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.6 PIPER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.7 PipeEye. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

ix
LIST OF FIGURES

5.8 The definitions of robot task. . . . . . . . . . . . . . . . . . . . . . . . . . 96

5.9 The definitions of robot task. . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.10 The definitions of environmental niche (Sewer robot). . . . . . . . . . . . . 97
5.11 The selected type of sewer robot. . . . . . . . . . . . . . . . . . . . . . . . 98
5.12 The high-level specification layers (Sewer robot). . . . . . . . . . . . . . . 98
5.13 Layer 1 generated from the sub-layer (Sewer project). . . . . . . . . . . . . 99
5.14 Layer 2 generated from the layer 1 (Sewer project). . . . . . . . . . . . . . 100
5.15 Mono-spec layer generated from the layer 2 (Sewer project). . . . . . . . . 100
5.16 The function priority of design (Sewer project). . . . . . . . . . . . . . . . 101
5.17 Two possible model of the water flowing inside of the pipe. . . . . . . . . . 102
5.18 Design extension of “naSIR mechanism”. . . . . . . . . . . . . . . . . . . . 103
5.19 The robot tilt angle adjusting. . . . . . . . . . . . . . . . . . . . . . . . . 104
5.20 Robot navigation by passive encoder. . . . . . . . . . . . . . . . . . . . . . 104
5.21 The results of design (Sewer robot). . . . . . . . . . . . . . . . . . . . . . . 105
5.22 An example of a network of sewage pipes, in Japan. . . . . . . . . . . . . . 106
5.23 Three samples of current commercial sewer inspection robots. . . . . . . . 108
5.24 Car like robot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.25 IThree different starting point for the small car-like robot. . . . . . . . . . 109
5.26 Car like robot motion in a curve (1). . . . . . . . . . . . . . . . . . . . . . 109
5.27 Car like robot motion in a curve (2). . . . . . . . . . . . . . . . . . . . . . 110
5.28 Car like robot motion in a Y-Junction. . . . . . . . . . . . . . . . . . . . . 110
5.29 Car like robot motion in a T-Junction. . . . . . . . . . . . . . . . . . . . . 111
5.30 naSIR mechanism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.31 naSIR mechanism motion during passing a 90 degree curve. . . . . . . . . 113
5.32 naSIR mechanism motion during passing a Y-Junctionn. . . . . . . . . . . 113
5.33 naSIR mechanism motion during passing a T-Junction. . . . . . . . . . . . 114
5.34 Motion of naSIR mechanism from a big diameter to small diameter pipe. . 114
5.35 Second version of naSIR mechanism. . . . . . . . . . . . . . . . . . . . 115
5.36 Prototype of naSIR mechanism. . . . . . . . . . . . . . . . . . . . . . . . . 115
5.37 naSIR platform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.38 A sample of sewer network. . . . . . . . . . . . . . . . . . . . . . . . . . . 118

x
LIST OF FIGURES

5.39 Developed KANTAROs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

5.40 KANTARO modular architecture. . . . . . . . . . . . . . . . . . . . . . . . 119
5.41 The newest version of KANTARO electronic boards. . . . . . . . . . . . . 120
5.42 First and second version of KANTARO. . . . . . . . . . . . . . . . . . . . 121
5.43 The newest laser scanner. . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.44 The sequence of on-board inspection process. . . . . . . . . . . . . . . . . 123
5.45 Nine different types of fault. . . . . . . . . . . . . . . . . . . . . . . . . . . 124
5.46 The sequence of off-line inspection process. . . . . . . . . . . . . . . . . . . 124
5.47 KANTARO with the life-optic-untethered cable. . . . . . . . . . . . . . . . 125
5.48 Our sewer test field including all kind of pipe bends at RRI laboratory. . 126
5.49 KANTARO in real world sewage pipe network. . . . . . . . . . . . . . . . 126

xi
List of Tables

3.1 Specification of “Musashi” robot . . . . . . . . . . . . . . . . . . . . . . . . 39

5.1 The practical result of naSIR mechanism movement with the different
wheels angle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.2 The practical result of naSIR mechanism motion with the different wheels
in shape and material. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.3 KANTARO electronics boards specification . . . . . . . . . . . . . . . . . 120

145

xii
 Chapter 1
Introduction
Chapter 1

Introduction

1.1 Objectives
Early development of robotics research in the 1960s and 1970s was focused on industrial
robots/manipulators for the automation of industrial processes. Mechanical manipula-
tors resemble human arms are deployed in the factories for various automation tasks.
In the 1980s, robots started walking out of the manufacturing floors in the form of
wheeled or legged mobile mechatronic systems and underwater autonomous vehicles.
The roles of robots are no longer limited to automated factory workers but are chang-
ing into explorers for hazardous, human-unfriendly, and extreme environments, and
servants to provide surveillance, security, and cleaning tasks. Ingenious autonomous
robotic systems that equipped with artificial intelligence capability resemble biological
counterparts were emerging in the late 1990s, such as Sonys Aibo robotic dog, and
Hondas humanoid robots from P2, P3, to Asimo. These systems not only walked out
of factories and service sector but also walked into our everyday life and households.
Eventually these robots are going to co-habit with humans to provide assistance and
cares.
Thereby, it is expected that the robots with different degrees of autonomy and mo-
bility will play an increasingly important role in all side of human life, thus these kinds
of robots will become much more complex than today, and the development of such
robots presents a great challenge for researchers. However, drawbacks of robot com-
plexity, necessity of more complex hardware, software and mechanical structure may
lead to low reliability and increasing of size, weight, cost, power consumption and mo-
tion limitation. In order to avoid the mentioned problems, the simplification of robots
mechatronics will be a critical point in their design. Although every working in robotics
soon confronts the reality that the design of the mechanical structure greatly affects
its performance and controllability, the theoretical and practical investigations of the
relationship (between the robot mechanical structure and its resulting controllability

2
CHAPTER 1. INTRODUCTION 3

and reliability) have largely been ignored. This is not simply a peripheral oversight,
but a fundamental gap in field of robotics.
The main question to be answered in this context is: “what is the unified approach
or methodology to actually design a proper mechanics for autonomous mobile robots
to provide a high degree of performance, functionality, simplicity and reliability”. In
this direction, there is no concrete method or approach to actually design a system, in
so much as there does not exist a unified approach to creativity. Given a particular
need, each individual designer would probably design something different[1]. In this
thesis the concept of “intelligent mechanical design” will be presented. The definition
of this concept will lead us to provide a design methodology, called “Mecha-telligence
methodology”, for design of autonomous mobile robots mechanics. In this methodology
the proper and sufficient mechanics will be achieved by considering the robot tasks,
behaviors, morphology, and applying the description of environmental and physical-
morphological constraints on the design process. To prove the validity of proposed
method we concentrate on developing of the mechanical structure and devices of three
types of autonomous mobile robots used in the three different educational, entertain-
ment and industrial applications. In these three projects we describe, not only, the
mechanical design process based on “mecha-telligence design” method, also, the detail
design which is considered to lead readers to get familiar as much as possible to our
vision relative to the presented design methodology.

1.2 Overview of the Thesis

In “chapter 2: Mechanics & Intelligence” the concept of “intelligent mechanical design”
will be presented to realize how a mechanical structure can be designed to have an
effect to the robot controllability, simplification and its tasks performance. The de-
scription of this concept will lead us to establish the landmarks of the territory of
mechanical designing in the form of seven design principles. The design principles,
named “Mecha-telligence principles”, provide a guidance on how to design autonomous
mobile robots mechanics. These principles guide us in asking the right questions when
investigating issues concerning a self-controllable, reliable, realizable, and compatible
mechanics for autonomous mobile robots. To show how the “Mecha-telligence princi-
ples” can be applied on the processes of the design of robots mechanics, we proposed a
novel methodology, named "Mecha-telligence methodology”. Mechanical design in the
proposed methodology is performed based on preference of classification of the robot
specification described by interaction of the robot with its environment and, also, the
physical parameters of the robot mechatronics itself. In this approach the robot spec-
ification is classified to a set of high-level- and low-level-specifications which can be
CHAPTER 1. INTRODUCTION 4

expressed as tasks and physical parameters of the robot, respectively. The high-level-
specification, including the two layers, are described by considering the robot tasks and
its interaction with the environment where it is used. A low-level-specification may,
consecutively, consists of a set of layers describing the physical setup of the robot, its
body, sensory, and motor systems. A main goal in these analyzing process of the robot
specification, is that, figuring out to a Mono-spec layer (the last layer of low-level-
specification) including a simple basic function or selecting a sensor for a single task
or behavior.
In “Mecha-telligence design” methodology the mechanical design will be started
based on an important defined robot mono-spec, and it will be extended to realize the
other mono-specs by adding the minimum actuators and sensors to the design. In each
step of the design process, also, we are trying to come up to the proper solution by
considering the robot morphology, design the suitable actuators and employing passive
mechanisms which have inherent intelligence characteristics. To prove the validity of
proposed method we concentrate on developing of the mechanical structure and devices
of three types of mobile robots used in the three different educational, entertainment
and industrial applications.
“Chapter 3: Soccer robot” is our first mechanical design challenge on developing
an autonomous mobile soccer robot which has a fully mechatronics modular archi-
tecture including a strong ball-kicking device with capability of lifting the ball and a
ball-holding mechanism, aiming at getting champion in RoboCup Midsize League. We
proposed a mechatronics modular platform which consists of an omni-directional mov-
ing mechanism, an omni-vision and a novel ball-kicking-lifting-holding device. In this
approach we show that selecting a proper mechanical moving mechanism and a suit-
able vision system can lead to realize a reliable, simple, and low cost robot comparing
with the first version of our car-like soccer robot which includes many different kinds of
sensors and a complex design structure. The main focus in this theses, with regard to
this project, is on detail design and developing process of a strong kicking device with
capability of shooting (up to 5 [m/s]), lifting (up to 120 [cm]), and holding the ball. The
ball-kicking is accomplished by design of an unique spring charging mechanism called
“Cam Charger“. The key idea is to charge a series of strong torsion springs by using a
cam with special shape. One of the specific features of the Cam Charger mechanism is
that charging, keeping and releasing of the springs energy are done only by employing a
simple DC motor-gearhead and a limit switch. Another special point is that the design
of the kicking mechanism can be extended to hold the ball by adding a simple linkage
mechanism to the Cam Charger structure. The same motor and the limit switch, used
for the ball-kicking, can be employed to activate and control the linkage mechanism of
ball-holding. This kicking device, also, has an inherent design characteristic to lift the
CHAPTER 1. INTRODUCTION 5

ball by using the different shape of plates installed on the kicker.

Next chapter “chapter 4: Artistic Robot” is our second challenge on the design,
modeling, simulation and implementation of a series of entertainment robots named
“Jumping Joe”, exhibited in Aichi Exposition 2005. Jumping Joe robots, artistic and
agile robots, can perform several rapid movements such as fast wake up, jumping and
somersault. In order to realize acrobatic movements, four different actuators which
can create the high-speed movements are developed: Inertia Actuator, Hip joint, Knee
joint and Cam Charger. Inertia Actuator consists of a H-shape cylindrical rotor , a DC
motor, and a brake mechanism, is an inertial energy-storage device. Inertia Actuator
absorbs mechanical energy by increasing its rotor angular velocity and delivers energy
by decreasing its rotor velocity. Inertia Actuator can generate a small internal torque
by changing the speed of the rotor and a big internal torque in short time by using a
brake to stop the high speed rotor in a moment. Hip joint and knee joint are designed
based on a parallel and Lever-Crank mechanisms, respectively. Parallel mechanism is
utilized, because the output torque of this kind of mechanisms can be increased without
a large ratio gear so that, a speedy and powerful motion can be expected. In order to
realize the jumping capability, “Cam Charger” mechanism presented as a kicking device
for RoboCup is redesigned to fit to the foot shape. In this case we designed a unique
and compact cam including a releasing part to avoid the distribution of the force at
the moment of springs energy releasing to the foot structure. To achieve the reliability
and easy to assemble and maintenance features, “Jumping Joe”s are designed to have
a modular architecture in their mechanical parts and eclectic circuits. “Jumping Joe”s
are made of combination of four main modules: Upper body, Hip joint, Knee joint and
Jumping foot. By using the modular architecture for “Jumping Joe”s, we can realize
the different kinds of artistic robots, only, by plugging the different modules to each
other.
“Chapter 5: Sewer Pipe Inspection Robot” will describe a new approach to the
sewer pipe inspection by developing a fully autonomous mobile robot “KANTARO”.
Cooperation in this national funding project, is our third challenge item in mechanical
design to show how a passive-active intelligent moving mechanism can be employed as
a robot platform for inspecting the real sewer pipe network. With regard to the current
sewer pipe inspection technology, all commercial robots are completely tele-operated,
usually via a tether cable, by a human operator. In addition, current sewer inspection
robots have a poor mobility function to pass any kind of pipe-bends such as curves
and junctions so that those robots are only capable to move into the straight pipes.
Inspecting the sewage pipes using the state of the arts inspection methods by the cur-
rent robots is costly, mostly human cost, and not fast enough to check and inspect
the amount of sewage pipes will grow stronger than it has actually happened, specially
CHAPTER 1. INTRODUCTION 6

in Japan. In this part of thesis we propose an innovative, fast and robust sewer in-
spection method by using a passive-active intelligent, fully autonomous, un-tethered
robot called KANTARO which fits to the pipes within a diameter range of 200-300
millimeters. KANTARO prototype robot, including a novel passive-active intelligent
moving mechanism (the main focus in this theses, with regard to this project), can
move into the straight pipe and pass different kinds of pipe bends without need to any
intelligence of the controller or sensor reading. In order to realize a fully autonomous
inspection robot, we also developed a small and intelligent 2D laser scanner for detect-
ing of the navigational landmarks such as manholes and pipe joints independently with
main computer system, and fusion with a fish eye camera to assist the pipe state and
fault detection. Consequently, we propose a fully autonomous sewer inspection system
but with a possibility to interfere the KANTARO control as it drives through the pipe
either via an optical underground wireless communication module, developed in this
project, or via a life-optic-untethered cable.
 Chapter 2
Mechanics & Intelligence
Chapter 2

Mechanics & Intelligence

2.1 Introduction
One of the basic, fundamental, and significant question in the field of mechanics related
to design an autonomous mobile robot is that:

“What is the ideal mechanical design for an autonomous mobile robot

which can perform the defined desired behavior (s) in an environment in
which it will be used?”.

By thinking deeply about this basic question and trying to disintegrate it in detail,
we can realize two main objections: first, the exact meaning of the term of “ideal”
regard to the amount of terms (such as Reliable, Optimal, Robust, Simple, proper,
and etc.) that can be replaced to it and second, difficulties and lack of exact definition
and description of these terms related to design a mechanic of an autonomous mobile
robot (see Fig. 2.1). In this thesis we figure out the mentioned problems by proposing
to integrate all possible terms at one, which is called “Intelligent mechanical design”
(Fig. 2.2). Note that the meaning of “Intelligent mechanical design”, presented in this
thesis, is not same as definition of the common phrases such as “Intelligent design”,

Figure 2.1: Mechanical design of an autonomous mobile robot and related factors.

8
CHAPTER 2. MECHANICS & INTELLIGENCE 9

Figure 2.2: New approach to mechanical design of an autonomous mobile robot and related
factors.

“Intelligent mechanics”, “Mechanical intelligence”, and “Intelligent designer”. In fact, as

it will be described in this chapter, “Intelligent mechanical design” is a primary concept
for realization of a methodology applying for mechanical design of autonomous mobile
robots. Using this methodology, called “Mecha-telligence methodology”, can lead the
designer to create and design the simple, proper and sufficient mechanical solution for
complex problem. Output of a mechanical design based on proposed methodology may
include intelligent mechanics, passive mechanism, mechanics related to body morphol-
ogy and materials and also mechanical solution exploited by the robot interaction with
its environment. Since “Mecha-telligence methodology” can be recognized as a guid-
ance method, the results of design ,also, can be affected by the designer expertness and
his/her intelligence and punctilious. For all of these, we selected the word of “Intelli-
gent” as a first step of our approach in realization of a methodology to design the ideal
mechanics for robots with the different degrees of autonomy and mobility.
Using this methodology is not only for autonomous mobile robots, also can be
extended and applied for any kinds of systems which request mechanical structures.
In this thesis we focus on mechanical developing process of autonomous mobile robots,
because we believe that these types of robots will play an increasingly important role
in all side of human life and the development of such autonomous robots presents a
great challenge for researchers. In this direction, exact understanding of definition of
“Intelligence” (subsection 2.2.1), “Autonomous robot”, “Mobile robot”, “task and desired
behavior” (subsection 2.2.2), and “Robot environment” are key points to describe the
concept of “Intelligent mechanical design” and also the definition and implementation
of “Mecha-telligence methodology”.

2.2 What is Intelligence?

“Intelligence is a large field and the human being is not yet ready to even understand its
fundamentals. Maybe he will never ever understand what it’s all about. We are living
CHAPTER 2. MECHANICS & INTELLIGENCE 10

in a huge complex environment and nature is just too complex to explain in three to
four sentences. Intelligence is something that was created by nature and for sure not by
us humans ourselves”[2]. There are the attempts of many leading experts in the field
to describe intelligence. The results were all kind of different definitions none of those
experts seemed to be sure about their own definition. One of the wonderful approach,
not only, describe the meaning of “intelligence”, also use this item to show the design
process of an autonomous agent is that a book with the title “UNDERSTANDING
INTELLIGENCE” written by Rolf Pfeifer and Christian Scheier. Here we will briefly
present a summery of their approach1 . In the next subsection, we figure out, not only,
meaning of the term of “Intelligence, but also consideration of the definitions of the
other necessary terms related to description of “Intelligence”, presented in this book, in
detail. Because complete understanding about their approach to describe the term of
“intelligence” and also design process of autonomous agents can lead us to explain, our
concept of “Intelligent Mechanical Design” for autonomous mobile robots presented in
this thesis in proper and easy way, Some important terms are described in the next
sections.

2.2.1 Definitions of The Necessary Selected Terms

Agent

“The term is used in many ways. (1) As an umbrella term if no distinction between
humans, animals, and robots is intended. (2) To designate an animated creature in
cyberspace. (3) To distinguish a certain type of simulation model (agent simulation)
from others. In agent simulations, agent and environment are modeled separately and
have independent dynamics. Agents acquire information about the environment only
through their (simulated) sensory systems. (4) In the context of Internet, to describe
the programs (software agents) that perform a certain service for a user, typically
information retrieval.”

Intelligence

“No generally accepted definition exists. The term is used to describe complete agents
(agents that are autonomous, self-sufficient, embodied, and situated) that resolve the
diversity-compliance trade-off in interesting way. Intelligence must always be seen
with respect to a particular ecological niche.” Figure 2.3 and 2.4 show two sample of a
complete agent, Todas Fungus Eater and a garbage-collecting robot.
“Todas Fungus Eater is operating on a distant planet. its task is to collect uranium
1
Most of the materials presented in this section (2.2) have been selected from “UNDERSTANDING IN-
TELLIGENCE” book.[3] The sentences placed in “this symbol” are selected material.
CHAPTER 2. MECHANICS & INTELLIGENCE 11

Figure 2.3: A sample of a complete agent, Garbage robot.

Figure 2.4: A sample of a complete agent, Garbage robot.

CHAPTER 2. MECHANICS & INTELLIGENCE 12

ore. It feeds on a certain type of fungs. It is autonomous (too far away for remove
control), self sufficient (it must take care of its own energy supply which, in this case,
is a particular type of funguse that grows on this planet, thus the name Fungus Eater),
embodied (it exist as a physical system), and situated (its knowledge about the en-
vironment is acquired through its own sensory system). In the Fig. 2.3, it is in the
process of devouring fungus.
“The task of the garbage-collecting robot is to collect the objects in its environment
and bring them to the home base. At the same time, it has to sustain itself by regularly
visiting the charging station, which is equipped with a light source. The robot uses this
light to find the charging station and the home base with its ambient light sensors, that
is, by performing phototaxis (i.e., moving toward a light source) and antiphototaxis
(moving away from the light source), respectively”.

Autonomous Agent

“An agent that has a certain independence of external control.”

Self-sufficient Agent

“An agent is said self sufficient if it is cable of sustaining itself over extended periods
of time.”

Embodiment/Embodied Agent

“A term used to refer to the fact that intelligence can not merely exist in the form of an
abstract algorithm but requires a physical instantiation, a body. In artificial systems,
the term refers to the fact that a particular agent is realized as a physical robot or as
a simulated agent. Embodied agents must be realized as physical systems capable of
acting in the real world.”

Situated Agent

“An agent is said to be situated if it acquires information about its environment solely
through its sensors in interaction with the environment. A situated agent interacts
with the world on its own, without an intervening human. It has the potential to
acquire its own history, if equipped with appropriate learning mechanisms.”

Diversity-compliance Trade-off

“Generating of diversity while complying with the givens of the system. Represents
a compromise between, on the one hand, generating new behavior, and on the other,
conforming to existing conditions.”
CHAPTER 2. MECHANICS & INTELLIGENCE 13

Ecological Niche

“The ecological niche for an animal is the range in each variable in its environment,
such as temperature, humidity, and food items, within which a species can exist and
reproduce. Niche occupancy usually implies competition (when animals of different
species use the same resources).”

Behavior

“What an autonomous agent is observed doing. Always the result of an interaction of

an agent with its environment.”

Task

“As used in a design context, the designer’s perspective on what the agent should
accomplish. A task is accomplished by a set of behaviors.”

2.2.2 Definition of Task and Desired Behaviors

In this subsection we distinguished between tasks and desired behaviors. “Both robots
in Fig. 2.5 achieve the task of collecting the ping pong balls, but they do so using
behaviors that are very different. So the task is consented with the effect of behaviors
rather than the behaviors themselves. Often the separation is not so clear.” For
example mowing robot. “The robot’s task is to mow the lawn. Mowing is in fact the
desired behavior to achieve the task of keeping the grass short. A garbage-collecting
robot obviously has the task of collecting garbage, or rather of eliminating garbage from
the streets. Its designer has to map this task onto desired behaviors like searching for
garbage, picking it up, bringing it back to garbage truck, ad so forth. Alternatively
the garbage could be burned on the spot, requiring no collecting. Note that, strictly
speaking, the definition of the task is independent of the robot itself. The designer

Figure 2.5: Ping pong collecting robot (printed from [3] (P:301))
CHAPTER 2. MECHANICS & INTELLIGENCE 14

decides what the tasks of the robot are to be and designs the robot in such a way
that it can accomplished them. This does not mean that there must be an explicit
representation of the task within the robot.2 “

2.2.3 Design principles for autonomous agents

Figure 2.6 shows Rolf Pfeifer and Christian Scheier’s approach to establish the land-
marks of the territory of designing an autonomous agent in the form of eight design
principles. “The set of design principles consists of two parts: a meta principle (prin-

Figure 2.6: Overview of design principles for autonomous agents [3] (P:303).

ciple 1) that tells us the essential constituents of the design process, and a number of
principles that concern the agent itself, its morphology, its sensors and effectors, and its
control architecture, that is, its internal mechanisms (Principles 2 to 8)”. The definition
of robot environment and robot behaviors and tasks are highlight as a meta principle
to show the significant effect of these terms in the design process of an autonomous
agent. Also, note that using the term of “Intelligence” for describing a complete agent,
is landmark as second principle. we refer the readers, in case of interested, about more
detail of the principles to “UNDERSTANDING INTELLIGENCE” book, chapter 10.
2
The terms of robot, typed in format of Italic in the two last sentences, in original reference are “Agent”
that they are changed to the term of “Robot” regards to our concept in this thesis.
CHAPTER 2. MECHANICS & INTELLIGENCE 15

2.3 Intelligent Mechanical Design

In this thesis the concept of "Intelligent mechanical design" is presented to show how
a mechanical structure can be designed to have an effect to the robot controllability,
simplification and its tasks performance. The description of this concept will lead
us to establish the landmarks of the territory of mechanical designing in the form of
seven design principles. The same as Rolf Pfeifer approach in describing the term
of “intelligence” we will try to come up to an accepted definition about “Intelligent
mechanical design” and consecutively, we will describe the necessary terms related to
the definition of the presented concept.

Intelligent mechanical design

The term can be used to describe complete mechanical designs (designs that are self-
controllable, reliable, realizable, and compatible) that resolve the functionality-usability
trade-off in optimal way. Mechanical design must always be considered with respect
to a particular environmental niche.

Self-controllable

A mechanical design is self-controllable, if it can be controlled by a set of certain

internal hardware, independence of external motive forces. Self-controllability must
always be considered with respect to a specific system.

Reliable

A mechanical design is reliable, if the mechanical design parameters specification can

be realized during requested system life-time. In this approach the number of actuators
and sensors used for a system with certain desired tasks can be recognized as a remark-
able mechanical design parameter that it has direct affect to the system reliability.

Realizable

A mechanical design is realizable, if it can be realized as a real mechanical structure

using the current manufacturing methods and the commercial materials exist in the
market. Realizable mechanical design can be affected by a number of user requests
such as cost, size, weight, water proof, dust proof, etc.

Compatible

A mechanical design is compatible, if the design of the sensory-motor architecture has

compatibility with the size, weight and shape of the hardware. Compatibility must
CHAPTER 2. MECHANICS & INTELLIGENCE 16

always be considered with respect to a specific system.

The functionality-usability trade-off

Represents a compromise between, on the one hand, adding new functionality, and on
the other, preserving usability condition [4].

Environmental niche

Environmental niche for a robot is described by the range in each variable in its envi-
ronment, within which a mechanical design as a part of a system should be realized.
We used the term of “environment” here to mean essentially two things. First, the
direct-physical meaning defined by the area in which the robot is used, including the
natural or artificial objects. Second, the indirect-physical meaning expressed by the
possible existing rules or user requested in case of considering the robot usability. This
requires some explanation that we explain here in form of examples. let’s call the ex-
ample of garbage-collecting robot which we introduced earlier in this chapter. Figures
2.7 to 2.10 show four examples of garbage-collecting robot different in the description
of environmental niche and related constraints. In the first example, Fig. 2.7, robot
should collect the garbage in the corner of the room. The environmental niche and
constraints can be defined as follows:
Environmental niche:

• A flat surface (Direct-physical meaning)

• Indoor environment (Direct-physical meaning)

Figure 2.7: First example of garbage-collector robot.

CHAPTER 2. MECHANICS & INTELLIGENCE 17

• A set of cylindrical garbage (Direct-physical meaning)

• A light (Direct-physical meaning)

Environmental niche constraint:

• Garbage size (D) (Direct-physical meaning)

• Light height (H1) (Direct-physical meaning)

In the second example, compare with the first example, a garbage room is located
in a corner of the room (Fig. 2.8). Also the limitation cost for making the robot in
this example has been assumed. By changing the environment of the robot, a garbage
room and the cost limitation, the environmental niche and relative constraints will be
changed as follows:
Environmental niche:

• A flat surface (Direct-physical meaning)

• Indoor environment (Direct-physical meaning)

• A set of cylindrical garbage (Direct-physical meaning)

• A light (Direct-physical meaning)

• A garbage room (Direct-physical meaning)

• Cost limitation (it may be considered as a user request) (Indirect-physical meaning)

Figure 2.8: Second example of garbage-collector robot. Cost limitation is assumed.

CHAPTER 2. MECHANICS & INTELLIGENCE 18

Environmental niche constraint:

• Garbage size (D) (Direct-physical meaning)

• Light height (H1) (Direct-physical meaning)

• Garbage room size (L2, W2, H2) (Direct-physical meaning)

• Garbage room door size (W3, H3) (Direct-physical meaning)

• Cost (cost < X $) (Indirect-physical meaning)

In this example “cost limitation” illustrates the second aspect of environmental niche
and relative constraint meaning, presented in this thesis. Third example indicates the
second aspect of environmental niche in form of game rules (Fig. 2.9). we assumed
that the robot will be designed based on the rules to participate in garbage-collecting
competition. With regard to the rules, there are limitations in size of the robot so that
the robot should be fit in a box with specific dimensions. In this case, the environmental
niche and relative constraint can be itemized as follows:
Environmental niche:

• A flat surface (Direct-physical meaning)

• Indoor environment (Direct-physical meaning)

• A set of cylindrical garbage (Direct-physical meaning)

• A light (Direct-physical meaning)

Figure 2.9: Third example of garbage-collector robot. Size limitation is assumed.

CHAPTER 2. MECHANICS & INTELLIGENCE 19

• A garbage room (Direct-physical meaning)

• Size limitation (it is considered based on the game rules) (Indirect-physical meaning)

Environmental niche constraint:

• Garbage size (D) (Direct-physical meaning)

• Light height (H1: if H4>H1) (Direct-physical meaning)

• Garbage room size (L2, W2, H2) (Direct-physical meaning)

• Garbage room door size (W3: if W3<W4, H3: if H3<H4) (Direct-physical meaning)

• Robot size (L4, W4, H4) (Indirect-physical meaning)

Note, because of the robot size limitation, some environmental niche constraint should
be considered in form of conditional constraints. As an example, the height of the light
(H1) is a constraint, if it is lower than the maximum height of the robot (H4>H1).
In fourth example (Fig. 2.10) it is targeted to design a camera mechanism (as an
example) for existing garbage-collecting robot. This example described here to show
that our aim in this thesis is not ,only, limited to mechanical design of whole robot,
but also it can be applied to design necessary parts which should be installed to the
existing robot. The environmental niche and relative constraint, for this example can
be itemized as follows:
Environmental niche:

• A flat surface (Direct-physical meaning)

Figure 2.10: Forth example of garbage-collecting robot. It is targeted to design a camera

mechanism (as an example) for existing robot.
CHAPTER 2. MECHANICS & INTELLIGENCE 20

• Indoor environment (Direct-physical meaning)

• A set of cylindrical garbage (Direct-physical meaning)

• A light (Direct-physical meaning)

• Size limitation (it is considered based on the game rules) (Indirect-physical meaning)

Environmental niche constraint:

• Garbage size (D) (Direct-physical meaning)

• Light height (H1: if H2+H4<H1) (Direct-physical meaning)

• Robot size (L2, W2, H2) (Indirect-physical meaning)

• Camera part size (L3, W3, H3) (Indirect-physical meaning)

In this case the “existing robot” is considered as a environmental niche, because the
size of robot has direct effect to the design of the camera mechanism. Also limitation
in the existing hardware to accept the additional sensors or motors can be recognized
as constraints of environmental niche.
In this approach the all design parameters that they affect a mechanical design,
can be considered as a environmental niche and relative constraint, and applied to the
process of design through the fourth Mechanical Design Principle presented in next
section.

2.4 Mechanical Design Principles: “Mecha-telligence Princi-

ples”
In this thesis the concept of "intelligent mechanical design" is presented to show how
a mechanical structure can be design to have an effect to the robot controllability,
simplification and its tasks performance. The description of this concept will lead
us to establish the landmarks of the territory of mechanical designing in the form of
seven design principles. The design principles, named “Mecha-telligence principles”,
provide a guidance on how to design autonomous mobile robots mechanics. They
incorporate the insights gained in this large design field in a compact and coherent form.
Macha-telligence principles guide us in asking the right questions when investigating
issues concerning a self-controllable, reliable, realizable, and compatible mechanic for
autonomous mobile robots (see fig. 2.11).
CHAPTER 2. MECHANICS & INTELLIGENCE 21

Figure 2.11: Overview of design principles of mechanical design of autonomous mobile robots.

Principle 1: The three-constituents principle3

Designing mechanical parts of autonomous mobile robots always involves three con-
stituents: (1) definition of environmental niche, (2) definition of desired behaviors and
tasks, and (3) design of robot mechanics. Constituents (3), design of robot mechanics,
has been split into design principle 2 through 7. A mechanical design problem can
be started as follows: Given intended environmental niche and the desired behaviors,
how do we design the mechanical parts of robot? Alternatively it can be proceeded
as follows: we already have a robot with a particular architecture. Also we have a
particular environmental niche. For a new desired behaviors or task which mechanism
or mechanical part is necessary to add to the robot and how do we design them?

Principle 2: The complete-mechanical design principle

The complete-mechanical design principle states that intelligent mechanical designs

are complete. Recall from section 2.3 that complete mechanical designs are self-
controllable, reliable, realizable, and compatible.
3
This principle is described based on the “principle 1” presented in UNDERSTANDING INTELLI-
GENCE book (P:302), mentioned in subsection 2.2.3. Here we selected some part of their description with
minor change regards to our concept in this thesis.
CHAPTER 2. MECHANICS & INTELLIGENCE 22

Principle 3: The principle of mechanical balance

The mechanical balance principle has two aspects: First it states that the complexity
of the mechanical design has to match the complexity of the robot desired behaviors.
Second it states that a robot mechanics should be designed as a part of a system not
arbitrarily. It must be done when the sensory system and also hardware architecture
are considered.

Principle 4: The environmental principle4

The environmental principle states essentially two things: “First, it implies exploiting
the physics of the system-environment interaction. Second, it means exploiting the
constraints of the environmental niche”. The environmental principle is one of the
significant principles that has direct effect to the mechanical design simplicity and
reliability. In many different cases (Specially “competition” and “ordering” cases) the
mechanics of the robot have to be designed in such a way to satisfy all environmental
niche constraints regard to the game rules or user requested, when the robot usability
is considered. Also, in our experiences in different projects, we realized that any kinds
of limitation in design parameters (even user requests such as cost and operating life-
time) can play a very important role in designer’s perspective and creativity. Therefore,
for all of these, we integrated all possible design parameters in form of “environmental
niche” and “environmental niche constraints”, explained in detail earlier in section 2.3.
In this approach the all design parameters that they affect a mechanical design, can be
considered as a environmental niche and relative constraint, and applied to the process
of design through “the environmental principle”.

Principle 5: The sensory-actuator principle

The sensory-actuator principle states that the type and position of the sensors and
actuators in a system can have direct effect to the robot functionality, performance,
simplicity, and reliability. In different cases by designing or selecting the proper and
sufficient sensors and actuators, also, their position in the system, the performance
and functionality of the robot can be higher than a case that the design is performed
without considering this principle. Also the number of actuators and sensors used in
the system can be affected by applying this principle in the mechanical design process
(contributing with the principle of sharing sensory-actuator system).
4
This principle is described based on the principle of “Cheap design” presented in UNDERSTANDING
INTELLIGENCE book, mentioned in subsection 2.2.3. In definition of the environmental principle we
selected two of three aspects of principle of “cheap design” regards to our concept in this thesis. Third aspect
of cheap design principle, designing parsimoniously , is used in description of the principle of sharing sensory-
actuator system but different meaning and approach compare with the concept mentioned in this book.
CHAPTER 2. MECHANICS & INTELLIGENCE 23

Principle 6: The principle of sharing sensory-actuator system

The principle of sharing sensory-actuator system means designing parsimoniously in

point of view of actuators and sensors number. For detail understanding of this prin-
ciple we refer to the three examples described in this thesis (chapter 3, 4, and 5)

Principle 7: The mechatronics modularity principle

The mechatronics modularity principle states that the robot should have a modular
architecture in its electronics and mechanics. This principle can be significant role
in the mechanical design reliability mentioned in the second principle, “the complete
mechanical design principle” and, also, easy maintenance. Modularization process al-
ways involves three steps: First: description of the robot system architecture. In this
step designer needs to have or make the architecture of the robot system in form of
flowchart. Visualization can help the designer’s perspective to realize what the robot
is includes and how can work as a system. Second, defining the possible module (single
module). Based on the flowchart provided in previous step, the possible module should
be defined by considering the similar hardware structures and the similar mechanical
connections. Third, merging the single modules to a merged module. The basic idea
underlying in this step is to extend the “modularity” concept by merging the “single
modules” into the “merged modules” aiming at the decreasing of the number of wires.

2.5 Mechanical Design Process: “Mecha-Telligence Methodol-

ogy”
To show how the Mecha-telligence principles can be applied on the processes of the
design of autonomous mobile robots mechanics, we proposed a novel methodology,
"Mecha-telligence" methodology (see Fig. 2.12). Mechanical design in the proposed
methodology is done based on preference of classification of the robot specification
described by interaction of the robot with its environment and, also, the physical
parameters of the robot mechatronics itself (principle 1). In this approach the robot
specification is classified to a set of high-level- and low-level-specifications which can be
expressed as tasks (desired behavior) and physical parameters of the robot, respectively.
In fact the high-level-specification, including the two layers: main and sub layers, is
described by considering the robot tasks and its interaction with the environment in
which it is used. A low-level-specification may, consecutively, consist of a set of layers
explaining the physical setup of the robot, its body, sensory, and motor systems. A
main goal in these analyzing process of the robot specification is that figuring out to
a Mono-spec layer (the last layer of low-level-specification) including a simple basic
 CHAPTER 2. MECHANICS & INTELLIGENCE

Figure 2.12: Overview of ”Mecha-telligence methodology” for mechanical design of autonomous mobile robots with defined task “X”.
24
CHAPTER 2. MECHANICS & INTELLIGENCE 25

function or selecting a sensor for a single task or behavior. The mechanical design
will be started based on an important defined robot mono-spec, and then, mono-spec
design will be extended to the other mono-specs by adding the minimum actuators
and sensors to the design (principle 6). In each step of the design process we are
trying to come up to a proper solution by considering the robot morphology, design
of the suitable actuators, by applying the description of environmental and physical-
morphological constraints, and by employing passive mechanisms which have inherent
intelligence characteristics (principle 3 and 4) . In this methodology the mechanical
design will be finalized by considering the proper and sufficient positioning of sensory-
motor system (principle 5) and concept of modularity in robot mechatronics (principle
7).
 Chapter 3
Soccer Robot
Chapter 3

Soccer Robot

3.1 Introduction
ROBOCUP is an international joint project to promote AI, robotics, and related field.
It is an attempt to foster AI and Intelligent robotics research by providing a standard
problem where wide range of technologies can be integrated and examined. RoboCup
chose to use soccer game as a central topic of research, aiming at innovations to be ap-
plied for socially significant problems and industries. The ultimate goal of the RoboCup
project is by 2050, develop a team of fully autonomous humanoid robots that can win
against the human world champion team in soccer [5]. “Hibikino-Musashi” is a joint
middle-size league RoboCup [6, 7] soccer team. Members of the team are from three
different research and educational organizations1 , all located in the Kitakyushu Science
and Research Park, Kitakyushu, Japan.
In this chapter we introduce our approach to realize a simple, robust, and valu-
able platform for “Musashi robots”. “Musashi” robots, presented in this thesis, are
series of autonomous mobile soccer robots which have fully mechatronics modular ar-
chitecture including a strong ball-kicking device with capability of lifting the ball and
a ball-holding mechanism, aiming at getting champion in RoboCup Midsize League.
“Musashi” robot, including an omni-directional moving mechanism, an omni-vision and
a novel ball-kicking-lifting-holding device, can be presented as a reliable and robust
soccer robot with high degree of simplicity, mobility, and maneuverability. In this ap-
proach we show that selecting a proper mechanical moving mechanism and a suitable
vision system can lead to realize a reliable, simple, and low cost robot comparing with
the first version of our car-like soccer robot included many different kinds of sensors
and a complex design structure. The main focus in this theses, with regard to this
project, is on detail design and developing process of a strong kicking device with ca-
1
The three organizations are: Kyushu Institute of Technology, The University of Kitakyushu, and Ki-
takyushu Foundation for the Advancement of Industry Science and Technology.

27
CHAPTER 3. SOCCER ROBOT 28

pability of shooting (up to 5 [m/s]), lifting (up to 120 [cm]), and holding the ball. The
ball-kicking is accomplished by designing an unique spring charging mechanism called
“Cam Charger“.

3.2 Summary of design based on “Mecha-telligence methodol-

ogy”
In this section the mecha-telligence methodology will be described for design and im-
plementation of an autonomous mobile soccer robot.

3.2.1 Description of the basic terms

The aim of this project is to participate in the international robot competition “RoboCup
Midsize League”. The first step in “Mecha-telligence methodology” as shown in Fig.
2.12 is to describe the tasks, desired behaviors, environmental niche, and environmental
niche constraints. Let’s start from a basic question: “what is ROBOCUP?”
RoboCup is an international joint project to promote AI, robotics, and related field.
It is an attempt to foster AI and Intelligent robotics research by providing a standard
problem where wide range of technologies can be integrated and examined. RoboCup
chose to use soccer game as a central topic of research, aiming at innovations to be
applied for socially significant problems and industries [5].

Robot task

By considering the aim of RoboCup, “playing soccer” can be recognized as the robot’s
task.

Desired behavior

Based on “RoboCup Midsize League” rules and common sense about soccer, the desired
behaviors can be described to three following items:

• Put the robot in the field

• Robot gets the desired commands (such as start the game, corner kick, throw in,
and etc) sent by a referee from a host computer via wireless communication.

• Robot moves, searches for the ball, carries a ball, avoid to crash to other robots
and kicking the ball to the opponent goal autonomously.
CHAPTER 3. SOCCER ROBOT 29

Environmental niche and relative constraints

Alternatively, regard to “RoboCup Midsize League” rules, the environmental niche and
the relative constraints can be defined and itemized as follows: (Fig. 3.1)

Figure 3.1: The definitions of environmental niche and the relative constraints (Soccer robot).

Environmental niche:

• A flat surface (Direct-physical meaning)

• Indoor environment (Direct-physical meaning)

• RoboCup field (a flat surface) (Direct-physical meaning)

• RoboCup rules (Indirect-physical meaning)

Environmental niche constraint:

• Two goals (yellow and blue) (Direct-physical meaning)

• Four poles (Mix of yellow and blue) (Direct-physical meaning)

• One ball (red) (Direct-physical meaning)

• Robot size (less than: L=50, w=50, H=80 [cm]) (Indirect-physical meaning)

• Robot weight (Less than 80 [kg]) (Indirect-physical meaning)

• Robot color (black) (Indirect-physical meaning)

Degree of autonomy and mobility

Robots should perform all actions necessary to play soccer, completely, autonomously
without human intermediary. Only the start, stop, and desired commands will be sent
via wireless communication to the robots by a side-referee. Robots have to be designed
as a wheel type-untethered robot (Fig. 3.2).
CHAPTER 3. SOCCER ROBOT 30

Figure 3.2: The selected type of soccer robot.

3.2.2 Simplification process of design functions

High-level specification layers (Main layer and Sub layer)

Regard to the description of four above terms, the high-level specification layers in-
cluding main-layer and sub-layer can be illustrated as Fig. 3.3.

Figure 3.3: The high-level specification layers including main-layer and sub-layer (Soccer
robot).
CHAPTER 3. SOCCER ROBOT 31

Low-level specification layers (Layer 1, layer 2, and mono-spec layer)

Figures 3.4, 3.5, and 3.6 show the generation of layer 1, layer 2, and mono-spec layer,
respectively. In this three figures the red color shows the terms which are on the
process of simplification, white color illustrates the terms that are not important for
our design purpose, and yellow color indicates the terms which recognized as a simple
basic function or selecting a sensor for a single task or behavior.

Functions priority of design

Figure 3.7 shows the result of our simplification process from “main layer” to “mono-
spec layer”. Also, this figure illustrates the function design priority in point of view of
mechanical parts design.

Figure 3.4: Layer 1 generated from the sub-layer (Soccer robot).

CHAPTER 3. SOCCER ROBOT 32

Figure 3.5: Layer 2 generated from the layer 1 (Soccer robot).

Figure 3.6: Mono-spec layer generated from the layer 2 (Soccer robot).
CHAPTER 3. SOCCER ROBOT 33

Figure 3.7: Results of the mono-spec layer generation and the function design priority (Soccer
robot).

3.2.3 Design process: Applying the “mecha-telligence principle”

This sub-section provides, only, an overview to our robot design and its architecture.
Also it shows how the “mecha-telligence principles” have been applied in the process of
the design. The whole detail designs and all definition and description of the related
concepts are explained in the next sections. To deep understanding about our approach
and detail of the robot design, reading the these sections is recommended.

Designing function 1

By surveying on the proper existing mechanisms and considering the principle 3 (The
CHAPTER 3. SOCCER ROBOT 34

Figure 3.8: The definitions of environmental niche and the relative constraints (Kicking de-
vice).

principle of mechanical balance), we designed an omni-directional platform for our

robot, called “Musashi robot”. The dynamical and kinematic characteristics of the de-
sign allows for a high maneuverability in the field, easy controllability for programing,
and expected reliability and compatibility (Principle 1: The complete-mechanical de-
sign principle) for the robot. Robots have been designed in such a way to satisfy the
three environmental constraints, size, weight and color (Principle 4: the environmental
principle).

Designing function 2 to 6

For the design of the kicking device, we must assume that we have a robot with the
particular architecture. We can now ask what kinds of kicking device should be designed
to fit in the robot and satisfy the our requirements. Therefore the “environmental niche”
and “environmental niche constraints” are different and can be defined as follows: (Fig.
3.8)
Environmental niche:

• Indoor environment (Direct-physical meaning)

• A flat surface (Direct-physical meaning)

• RoboCup rules (Indirect-physical meaning)

CHAPTER 3. SOCCER ROBOT 35

• A robot (Indirect-physical meaning)

Environmental niche constraint:

• Robot size (less than: L=50, w=50, H=80 [cm]) (Indirect-physical meaning)

• Maximum size of kicking device: L=310, w=93, H=188 [mm] (Indirect-physical

meaning)

By considering the second constraint of the environmental niche (Applying principle 4 )

and implementation of principle 3 , we developed a strong kicking device with capability
of shooting up to 5 [m/s]. The ball-kicking is accomplished by design of an unique
spring charging mechanism called “Cam Charger“. In this novel kicking device, “Cam
Charger” mechanism has been designed instead of three mechanisms to save, keep and
release the energy of a series of torsion springs just using a motor and a limit switch
(Applying principle 5 ). The functionality of this ball-kicking mechanism is extended to
lift the ball up to 120 [cm] by adding an edge to the kicker plate (Applying principle 6 )
and hold the ball by installing a simple linkage mechanism inside of the Cam Charger
mechanism. This linkage mechanism is actuated and controlled by the same motor and
limit switch used for ball-kicking (Applying principle 5 & 6). (For detail description,
please refer to the next sections)

Designing other functions

Object detection, collision avoidance, and self localization were performed by selecting
an omni-directional camera as a robot’s vision sensor, installed in the top of the robot
(Applying principle 5). An on-board wireless network device installed on the robot’s
PC is used for robot communication with host computer and also other robots.
For easy understanding of what we did and what we designed, the results are visu-
alized in Fig. 3.9. The black color shows our solutions for the listed problems in the
mono-spec layer. Note that we could realize all functions, only, by selecting a proper
moving mechanism (Omni-directional mechanism), a suitable vision system (Omni vi-
sion), and designing a novel charging spring mechanism (Cam Charger).
CHAPTER 3. SOCCER ROBOT 36

Figure 3.9: The results of design (Soccer robot). The black color shows our solutions for the
listed problems in the mono-spec layer.

3.3 Concept of Designing “Musashi” Robot

Figure 3.10 shows the first version of our robot, which is purchased from “Frounhofer
AIS (Institute Autonomous Intelligent System)”. This robot has a car-like locomotion
mechanism including two active wheels and two castor wheels in back and front, re-
spectively. A digital camera attached 70 [deg] wide-angle lens is mounted the top of the
robot that can rotate 360 [deg] around it’s axis, actuated a DC motor and controlled
using an absolute encoder. Because, the robot is used in the dynamic environment, for
obstacle avoidance, the robot is also equipped with different kinds of sensors such as
two IR sensors, two distance sensors, and touch sensors as shown in fig. 3.10. Also, a
pneumatic kicker is installed on the robot which is supplied by two small air pressure
tanks. Regard to above explanation about the mechatronics design of the robot, the
existing problems can be itemized as follows:

• (a) Poor mobility functions to perform the requested motion such as rotate around
the ball or lateral movements.

• (b) Complex data processing and sensors reading regard to sensor fusion.
CHAPTER 3. SOCCER ROBOT 37

Figure 3.10: The first version of “Hibikino-Musashi” team robot equipped with eleven sensors

• (c) Complexity of the camera motion includes designing hardware, software, and
its control.

• (d) Low reliability and not enough robust for a dynamic environment such as
RoboCup.

• (e) Complex mechatronics system in consideration of difficulties of assembling,

maintaining, extendability, and transporting.

• (f) Low speed ball kick, comparing with the other team’s robots which can shoot
the ball in high speed up to 6.0 [m/s] [8, 9].

In order to solve above mention problems and achieve the required characteristics
for the RoboCup scenario and develop generic hardware platform for mobile robotic
research and teaching, a new model of a mobile robot, named “Musashi” robot, is com-
pletely designed and constructed by “Hibikino-Musashi” team members. In this design
approach we show that selecting a proper mechanical moving mechanism, a suitable
vision system, and mechatronics modular architecture design can lead to realize a re-
liable, simple, and low cost robot comparing with the previous car-like robot included
many kinds of sensors and a complex design structure in its hardware and software.

3.4 “Musashi” robot architecture

“Musashi” robot is designed based on two significant and fundamental concepts: “omni-
CHAPTER 3. SOCCER ROBOT 38

directional” and ”modularity” concepts (Applying principle 7: The mechatronic mod-

ularity principle).

3.4.1 “Omni-directional” concept

“Musashi” robot includes an omni-directional platform, an omni-vision (Fig. 3.11). The

dynamical and kinematic characteristics of the design allow for a high maneuverability
in the field. The 3D mechanical CAD drawings were done with Autodesk Inventor
and the mechanical parts are completely manufactured by “Hibikino-Musashi” team
members. Each robot is equipped with 3 omni-wheels, each of them driven by a 70 [W]
DC motor. Gearboxes with reduction rations of 18:1 are used to reduce the high angular
speeds of the motors (7000 rpm) and to amplify the wheel’s mechanical torques. The
velocity feedback is done by using 540 [ppr] digital incremental encoders. The velocities
of the wheels are controlled by three Faulhaber motor drivers (MCBL 2805) that each
one has a RS232 communication port. The controllers read the pulse trains from the
motor encoders and produce amplified PWM output voltages for the motors based on
a PID algorithm. The result is a mobile robot with maximum linear speed of 1.9 [m/s]
and acceleration of 2.5 [m/s2 ]. In this approach, we could figure out the first problem
(a) mentioned in section 3.3.
The problems (b) and (c) are considered, by changing the mono-directional vision
(the vision system of the first version of our robot) to the omni-directional vision

Figure 3.11: “Musashi” robot includes an omni-directional platform, an omni-vision, and a

strong novel ball-kicking device, designed by Autodesk Inventor 3D-CAD.
CHAPTER 3. SOCCER ROBOT 39

Table 3.1: Specification of “Musashi” robot

system consisting of a digital camera (IEEE 1394) and a hyperbolic mirror. The main
parameters which should be considered and, carefully, chosen are the height of the
mirror (H) and distance of between mirror and the camera (h) as shown in a Fig. 3.11.
They also determine the viewing angle of the lens for a full-size image of the mirror
[10]. In this approach, the omni-camera can be used, not only, for object detection and
localization, but also for fully collision avoidance, comparing with our first robot that
is requested the different types of sensors to avoid in contact with the other robots.
As shown in Table 3.1, only sensors using in ’Musashi” robot are an omni-directional
camera and three DC motor encoders comparing with the first version of our robot
that was equipped with 11 sensors (two IR sensors, two distance sensors, a camera,
two DC motor encoders, touch sensors, two limit switches for robot fingers, and an
absolute encoder for camera motion). Figure 3.12 shows flowchart of “Musashi” robot
power system including a main Li-Polymer battery (25.9 [V]) and an extra Li-Polymer
battery (7.2 [V]) using for high acceleration and speed during catching and carrying
the ball. The necessary voltage for the camera and the micro computer power supply
are produced by converting 25.9 [V] to 12.0 [V] and 5.0 [V], respectively. The power
consumption of the robot is approximately 40 [W], and, the operation duration of the
robot is estimated to be 0.5 [h].

3.4.2 “Modularity” concept

1. First step: Description of the Robot System Architecture

To describe and emphasize the “modularity” concept, in the first step, it is necessary
CHAPTER 3. SOCCER ROBOT 40

Figure 3.12: Flowchart of “Musashi” robot power system.

to have a short overview to the robot system architecture design. “Musashi” robot is
equipped with a laptop on which the image processing, the control and the communica-
tion and data exchange are performed. Behavior commands such as start, stop, corner
kick, and etc. are received by a referee box PC, located out of the field, via on-board
wireless LAN. To achieve a safe, simple and robust system, the robot is adopted using
three different transport protocols: IEEE 1394, USB, and RS232. The communication
between omni-directional camera and laptop PC, installed on the robot, is performed
using IEEE 1394 (FireWire). The laptop PC sends the motor control commands (tar-
get velocity) to the motor drivers via USB protocol. However, the motor drivers have
only RS232 protocol port, we use three USB/RS232 converters between PC and motor
drivers (Fig. 3.13-a, note: Considering with no color). Additionally, a USB/RS232
converter is used between PC and the circuit of kicking device.

2. Second step: Defining the possible modules (single module)

Regard to the flowchart of the robot architecture, the possible modules should be
defended by considering the similar hardware structures or similar mechanical con-
nections. Figure 3.13-a shows the six possible modules, illustrated with six different
colors, which can be defined for “Musashi” robot hardware. For examples, USB hub
and USB/RS232 converters can be selected to be a module (USB module), because
they have a close hardware structure (Fig. 3.13-b). Also, because there is a mechanical
CHAPTER 3. SOCCER ROBOT 41

connection between each omni-wheel and each shaft of the motor, MW modules can
be defined. One of the significant considerations in the process of a simple module
design is that: “The module should be designed to have connector/s as an
I/O port/s”. The connector/s should be fixed to the body of the module except of
special cases that fixing the connector is not possible or it requests the complex design

Figure 3.13: System architecture of “Musashi” robot. (a) Shows the possible modules. (b)
Shows the result of single modules.
CHAPTER 3. SOCCER ROBOT 42

Figure 3.14: (a): Motor driver module (MD module) including a connector as I/O port. (b):
USB module consists of two connectors for input and output port. Output connector is not
visible in the figure.

or expensive manufacturing. Figure 3.14-a,b show the developed motor driver module
(MD module) and USB module, respectively. USB module (Fig. 3.14-b) has an output
connector same as connector of the MD module in its bottom side that it is not visible
in the figure.

3. Third step: Merging the single modules (merged module)

The “modularity” concept can be extended by merging the single modules into each
other (called merged modules), aimed to decrease the number of the wires that they
should be used for connection between the single modules. A “merged module” can be
defined by considering the flow chart connections of the “single” modules and can be
accomplished by design and implementation of the “back plane” concept. A back plane
can be functioned as a “single” modules communication port for a “merged” module.
For example, by considering the connection between the USB, KC and MD modules,
illustrated in Fig. 3.13-b, a back plane can be designed to merge the five modules
(an USB, a KC, and three MD modules) and solve the problem of the complex wiring
connections. The new “merged” module is named “central controller module” (Fig.
3.15).
As a summary of “Musashi’ robot modular architecture: We designed the robot
in two main modules, Bottom module and Upper module (Fig. 3.16), considering the
easy assembling, maintaining, and transporting. Consequently, bottom module consists
of six “single modules” (a switch, a battery, a kicking device, and three Motor-wheel
modules) and one “merged module” (Central Control module). In this approach, we
CHAPTER 3. SOCCER ROBOT 43

Figure 3.15: Modules of “Musashi” robot hardware.

Figure 3.16: Modules of “Musashi” robot.

CHAPTER 3. SOCCER ROBOT 44

could solve two problems (d) and (e) mentioned in the section 3.3.

3.5 Strong novel kicking device

The first official RoboCup games and conference was held in 1997. In total 40 teams
participated on the competitions, seven of them participated in the RoboCup Mid-size
league (MSL). In that year the robots had no kicking device and they were, only, able
to carry the ball to the inside of the goal. One year later (the second RoboCup game,
1998), Freiburg team had made the first kicking device in the history of the RoboCup
and became champion. After this year, developing the kicking device became a key
point for all the participated teams, so that, many researchers put their efforts to
develop kicking devices with stronger than as possible. For example, Freiburg team
made a kicking device, in 2000, which worked with spring, much faster than their
first solenoid type of shooting device, or in 2002, Philips team developed a spring
type kicking device with high speed (about 6 [m/s]). In that time, GMD-MUSASHI
team (our team) had a robot including a pneumatic kicking device with capability of
shooting the ball about 2 [m/s] (Fig. 3.10). The new developed kicking device has
high performance, so that, the kicking device can speed the ball up to 6 [m/s] and lift
the ball about 1 [m].

3.5.1 Overview of the Different Types of Kicking Device

In general, the kicking devices can be classified in four different types in dependence
of the activated force:

• The Solenoid kicking device

• The Pneumatic Kicking Device

• The Elastic Kicking Device

• The Spring Kicking Device

In this section, we only make an over view about advantage, disadvantage and basic
operation of each type of kicking devices.

The Solenoid kicking device

As shown in Fig. 3.17, a solenoid consists of a coil and iron rode. If a current is applied,
a magnitude flux will be produced which can pull the iron rode into the coil (PULE
type) or push it out of the coil (PUSH type). This movement of the iron rode generates
a force (F ) which can be used as an motive force to activate the kicking device.
CHAPTER 3. SOCCER ROBOT 45

Figure 3.17: Schema of a solenoid.

Basic operation

Three methods are common in making a solenoid kicking device (see Fig. 3.18). The
difference is dependence of, only, the position that the force of solenoid is applied.

Advantage

• High speed: In general solenoid can generate a very fast motion because of the
characteristic of magnetic flux.

• Easy control: The control of a solenoid can be achieved by on/off a normal switch,
located between a battery and the solenoid.

Disadvantage

• High current:Force in a solenoid and the speed of the iron rode depends on several
parameters such as the material of the rode, the gap between the coil and the
rode, and the number of turns of wire on the coil.

• High weight: Solenoid with capability of high force generation requested a heavy
iron rode also many number of turns of wire, that totally make heavy device.

Figure 3.18: Three methods are common in making a solenoid kicking device. The difference
is dependence of, only, the position that the force (F ) of solenoid is applied.
CHAPTER 3. SOCCER ROBOT 46

The Pneumatic Kicking Device

Basic operation

Two methods are common in making a pneumatic kicking device (see Fig. 3.19). The
difference is dependence of, only, the position that the force of solenoid is applied.

Advantage

• Low weight: With recent improvements in the pneumatic technology, new cylin-
ders are produced to have light wight with high performance.

• Easy control: The control of a pneumatic cylinder is done by the solenoid. then,
as mentioned before, the control of a solenoid can be achieved by on/off a normal
switch, located between a battery and the solenoid.

Disadvantage

• High current: Force in a solenoid and the speed of the iron rode depends on
several parameters such as the material of the rode, the gap between the coil and
the rode, and the number of turns of wire on the coil.

• High weight: Solenoid with capability of high force generation requested a heavy
iron rode also many number of turns of wire, that totally make heavy device.

Figure 3.19: Two methods are common in making a pneumatic kicking device. The difference
is dependence of, only, the position that the force (F ) of the pneumatic device is applied.
CHAPTER 3. SOCCER ROBOT 47

Figure 3.20: Basic operation of a elastic kicking device.

The Elastic Kicking Device

Basic operation

Figure 3.20 shows the basic operation of a elastic kicking device.

Advantage

• High speed: Since stored energy in a elastic can be released in a very short time,
it is able to give a high speed to the ball.

Disadvantage

• Needs large space: Elastic kicking device requests a big area for operation and
control.

The Spring Kicking Device

Basic operation

It depends on the designer can be completely different in basic operations. In fact,

there is no concrete classification in these types of kicking devices. As an example,
Philips team made a kicking device using a car-pressure spring (see Fig. 3.21) that it
is able to shoot the ball more than 6 [m/s]. Amount of applied force by the plate of
Philips kicking device, in the moment of hitting the ball, shown in Fig. 3.22. Figure
3.23 shows the spring mechanism of KIRC team from Japan in which is designed base
on torsion springs.
CHAPTER 3. SOCCER ROBOT 48

Figure 3.21: Schema of Philips kicking device [8].

Figure 3.22: Amount of applied force by the plate of Philips kicking device, in the moment
of hitting the ball.

Advantage

• High speed: Since stored energy in a spring can be released in a very short time,
it is able to give a high speed to the ball.

Disadvantage

• Needs large space: Spring kicking device requests a big area for operation and
control.
CHAPTER 3. SOCCER ROBOT 49

Figure 3.23: Kicking device of KIRC team (KIT, Izuka, Japan)

3.5.2 Design of New Kicking Device

Basic Concepts of Design

The new kicking device is designed based on two main functions and two environmental
constraints:

• Function to shoot the ball with high speed as much as possible (Target about 5
or 6 [m/s])

• Function to lift the ball

• First environmental constraint: Very simple control (we had limitation in our
current robot hardware. using one motor without encoder and a few limit switches
are permitted)

• Second environmental constraint: Small size (We have limitation for space of
kicking device in our current robot (see Fig. 3.24). As shown in Fig. 3.25 this
space, roughly, are L=310 [mm], W=93 [mm], and H=188 [mm].

Regards to second environmental constraint and also robot shape, there is three possi-
bilities to design a kicking device as shown in Fig. 3.26. We selected the third possibility
because of limitation in space.

Principle of Design

In general three mechanisms are necessary to use energy of a spring:

• Mechanism to charge the spring (Saving energy)

CHAPTER 3. SOCCER ROBOT 50

Figure 3.24: “Musashi” robot.

• Mechanism to lock the spring (Keeping energy)

• Mechanism to release the spring (Releasing energy)

In the developed novel kicking device presented in this section, we designed a mech-
anism instead of three mechanisms to save, keep and release the energy of a series of
torsion springs just using a motor and a limit switch.

Saving energy

A cam mechanism use to charge the torsion springs and save the energy. The basic
concept of this mechanism shown in the Fig. 3.27. With regard to the eqs. (3.1) and
(3.2), the motive force (F ) to charge the torsion spring is minimized while the force (F )
is perpendicular to the axes (m) (β =90◦ ) where the torque τS takes a certain value.

τS
F = (3.1)
X · sinβ

τS
β = 90◦ =⇒ Fmin = (3.2)
X
The main function of the special designed cam is keeping the angle β 90◦ during the
spring charging process (Fig. 3.28). Then the torque of motor τC , needed to charge
the spring, will be minimized as following equation:
CHAPTER 3. SOCCER ROBOT 51

Figure 3.25: New kicking device should be designed to fit to the size of L=150 [mm], W=150
[mm], and H=230 [mm].

τC (min) = Fmin · d (3.3)

where d is the perpendicular distance between the direction of contact force F (line
N) and the center of the cam C. The distance d can be calculated with the following
equations:

Line M : y = x · tanθ (3.4)

Line N : x · cosθ + y · sinθ + l1 = 0 (3.5)

−l1 cosθ − l2 sinθ + l1

d= √ = −l1 cosθ − l2 sinθ + l1 (3.6)
cos2 θ + sin2 θ
Here, l1 is the length of OG, and the angle θ is the angle between the line OG and the
line M. And the force F and τC can be obtained as following:
CHAPTER 3. SOCCER ROBOT 52

Figure 3.26: Three possibility to design a new kicking device.

τS nkα
F = = (3.7)
l1 l1

τC = F · d = nkα(1 − cosθ − l2 sinθ/l1 ) (3.8)

Where n is the number of springs, k is the stiffness coefficient of springs and α is the
charging angle of springs. In this mechanism, during the charging time, the perpendic-
ular distance between the direction of contact force F and the center of the cam (d in

Figure 3.27: Basic concept of cam charger mechanism. Joint A is fixed and the torsion springs,
as a resistance torque, are mounted in this joint. τS = spring torque (resistant torque), F =
Motive force , X= distance between joint A and the point where force F is applied, and β=
angle between force F and the direction of n
CHAPTER 3. SOCCER ROBOT 53

Figure 3.28: The coordinate and relation between the cam and the torsion spring. The main
function of the cam is keeping the β angle 90 degree in during the spring charging process.

Fig.3.28) decreases when the spring torque (τS ) and the contact force (F ) increase. The
red line in Fig.3.29 shows the derivation of motor torque relative to the charged angle
of springs. The important point is that when the spring is charged more and more,

Figure 3.29: The derivation of motor torque τC (blue graph) and spring torque (orange graph)
respect to the charged angle of spring α where n= 2, k=3.0 [Nm/deg], α = maximum 22 [deg],
l1 =50 [mm] and l2 =50 [mm].
CHAPTER 3. SOCCER ROBOT 54

Figure 3.30: Shows the sequence of charging of springs in begging, middle, and end of the
charging process. In the end of process of charging because of the direction of force F passes
the center of the cam, the mechanism has inherent characteristics to lock.

the required motor torque is getting less and less in the second half of the graph in
Fig.3.29. This mechanism makes charging very strong series of torsion springs possible
with a small motor. As an example in case of a cam charger with specification n= 2,
k=3.0 [Nm/deg], α = maximum 22 [deg], l1 =50 [mm] and l2 =50[mm], for charging the
125 [N.m] resistance torque of two springs, we only need a motor-gearhead which can
produce almost 9 [N.m] torque as a peak.

Keeping energy

In this mechanism, when spring is charged full, because of the direction of the force
F passes the center of the cam, if the motor stops, the mechanism will lock. Then we
can keep the energy of springs (See Fig. 3.30).

Releasing energy

It is clear when the motor continues to rotation (see Fig. 3.30) after the locking process;
the spring will be released.

Ball-lifting Mechanism

Figure 3.31 indicates our simple solution to lift the ball (Applying the principles 4 and
6 ). By changing the width of the edge-plate we can adjust the parameter of lifting the
ball up to 120 [cm].

Ball-holding Mechanism

Figure 3.32 illustrates our simple linkage mechanism, installed inside of the cam charger
CHAPTER 3. SOCCER ROBOT 55

Figure 3.31: Ball-lifting mechanism.

mechanism to hold the ball (Applying the principle 6 ). This linkage mechanism is
activated and controlled by the same motor and limit switch used for lifting the ball as
shown in Fig. 3.33. The key point is using the death area used for lucking the “Cam
Charger” mechanism. In lucking area, if the motor is rotating the kicking plate does
not move but the linkage mechanism will be active by pushing the small cam fixed to
the big cam.

Figure 3.32: Ball-holding mechanism.

CHAPTER 3. SOCCER ROBOT 56

Figure 3.33: Control sequences of the ball-lifting mechanism.

3.6 Conclusion and Experimental Result

In this chapter we presented an autonomous mobile soccer robot which has a fully
mechatronics modular architecture including a strong ball-kicking device with capa-
bility of lifting the ball and a ball-holding mechanism, aiming at getting champion in
RoboCup Mid size League. We proposed a mechatronics modular platform consists of
an omni-directional moving mechanism, an omni-vision and a novel ball-kicking-lifting-
holding device. In this approach we show that selecting a proper mechanical moving
mechanism and a suitable vision system can lead to realize a reliable, simple, and low
cost robot comparing with the first version of our car-like soccer robot included many
different kinds of sensors and a complex design structure. We described, in detail, the
design and developing process of a strong kicking device with capability of shooting
(up to 5 [m/s]), lifting (up to 120 [cm]), and holding the ball. The ball-kicking is ac-
complished by design of an unique spring charging mechanism called “Cam Charger“.
The key idea is to charge a series of strong torsion springs by using a special design
of a cam shape. One of the specific features of the Cam Charger mechanism, is that,
charging, keeping and releasing of the springs energy are done only employ a simple
DC motor-gearhead and a limit switch. Another special point is that the design of the
kicking mechanism can be extended to hold the ball by adding a simple linkage mecha-
nism to the Cam Charger structure. The same motor and the limit switch, used for the
ball-kicking, employed to activate and control the linkage mechanism of ball-holding.
CHAPTER 3. SOCCER ROBOT 57

This kicking device, also, has an inherent design characteristic to lift the ball by using
the different shape of plates installed on the kicker. As a result, Hibikino-Musashi got
the first place at RoboCup Japan Open 2006 and was ranked among the best 8 teams
at the RoboCup 2006 world championships in Bremen.
 Chapter 4
Artistic Robot
Chapter 4

Artistic Robot

4.1 Introduction
Human beings and animals have remarkable abilities to walk, run and jump over a wide
variety of hazard terrains. These actions in general are dynamic activities in which
inertial forces are significant and balance is achieved by dynamic responses. There
have been a number of efforts to build robots with running and jumping capability. An
early jumping robot was Matsuoka’s planar one-legged hopper in 1980 that had short
stance time with thrust provided by a high-force electric solenoid.[3] Marc Raibert
developed one-, two-, and four-leg that can balance in the plane and in 3D in the 1980s
[12, 13]. These robots had telescoping legs with internal air spring for compliance,
and hydraulic actuators. More recently, the Jet Propulsion Laboratory developed a
several jumping robots [14] and Sandia National Laboratories created a robot capable
of jumping thousands of times [15]. S.A. Stoeter and N. Papanikolopoulos proposed a
small cylindrical robot, Scout, capable of rolling and jumping [16, 17]. “Jumping Joe"s
are artistic and agile robots that perform acrobatic movements such as fast wake up,
jumping and somersault. The main objective of these robots are the development of
new behavior like jumping, rolling and their combinations. The full function Jumping
Joe design is shown in Fig.4.1. In order to realize acrobatic movements, like rapid
getting-up, walking, jumping and somersault, four kinds of actuators which have been
designed to create high-speed movements are installed. Inertia actuator, installed in
the upper body, is a torque generator that consists of a rotor and a brake mechanism.
Each of hip joints consists of a 2 DOF actuator developed on a basis of parallel link
mechanisms for achieving light weight and high torque. The knee joints have 1 DOF,
high torque mechanisms employing a lever-crank mechanism. Each jumping foot has 1
DOF and is designed based on a special mechanism named “Cam Charger" consists of a
special designed cam and torsion springs. Each actuator is developed as an independent
module, with the movements being controlled by microcomputer PICs. The central

59
CHAPTER 4. ARTISTIC ROBOT 60

Figure 4.1: Design of full function Jumping Joe. Full function Jumping Joe has eight degrees
of freedom, two DOF in each hip joint and one DOF in each knee joint and each foot.

IC is a rewritable FPGA that controls each microcomputer and communicates wireless

with a host computer via Bluetooth.

4.2 Mecha-telligence methodology

In this section the mecha-telligence methodology will be described for design and im-
plementation of an artistic robot.

4.2.1 Description of the basic terms

The aim of this project was to exhibit an artistic robot in Aichi Exposition 2005. We
proposed a proposal to the NEDO (New Energy and Industrial Technology Develop-
ment Organization) for design, simulation and implementation of a legged robot with
no hand that it can perform the rapid movements such as fast wake up, jumping, and
somersault.

Robot task

By considering the aim of what we would like to exhibit, “Artistic actions” can be
recognized as the robot’s task.
CHAPTER 4. ARTISTIC ROBOT 61

Desired behavior

Based on the presented proposal the desired behavior can be described in two following
sentences:

• The artistic robot is standing or laying down on a flat surface.

• The robot, suddenly, performs different rapid movements such as fast wake up,
jumping, and somersault.

Environmental niche and relative constraints

Alternatively, regard to the presented proposal and exhibition area, the environmental
niche and the relative constraints can be defined and itemized as follows: (Fig. 4.2)
Environmental niche:

• Indoor environment (Direct-physical meaning)

• A flat surface (Direct-physical meaning)

Environmental niche constraint:

• Robot has two leg (Indirect-physical meaning)

• Robot has no hand (Indirect-physical meaning)

Figure 4.2: The definitions of environmental niche and the relative constraints (Artistic robot).
CHAPTER 4. ARTISTIC ROBOT 62

Figure 4.3: The selected type of artistic robot.

Degree of autonomy and mobility

Robot should perform all actions, completely, autonomously without human intermedi-
ary. Only the start, stop, and desired actions can be sent via wireless communication to
the robot by a human operator. Robot has to be designed as a legged type-untethered
robot (Fig. 4.3).

4.2.2 Simplification process of design functions

High-level specification layers (Main layer and Sub layer)

Regard to the description of four above terms, the high-level specification layers in-
cluding main-layer and sub-layer can be illustrated as Fig. 4.4.

Figure 4.4: The high-level specification layers including main-layer and sub-layer (Artistic
robot).
CHAPTER 4. ARTISTIC ROBOT 63

Low-level specification layers (Layer 1, layer 2, and mono-spec layer)

Figures 4.5 and 4.6 show the generation of layer 1 and mono-spec layer, respectively.
In this three figures the red color shows the terms which are on the process of simplifi-
cation, white color illustrates the terms that are not important for our design purpose,
and yellow color indicates the terms which recognized as a simple basic function or
selecting a sensor for a single task or behavior.

Figure 4.5: The layer 1 generated from the sub-layer (Artistic robot).

Figure 4.6: The mono-spec layer generated from the layer 1(Artistic robot).
CHAPTER 4. ARTISTIC ROBOT 64

Figure 4.7: Results of the mono-spec layer generation and the function design priority (Artistic
robot).

Functions priority of design

Figure 4.7 shows the result of our simplification process from “main layer” to “mono-
spec layer”. Also, this figure illustrates the function design priority in point of view of
mechanical parts design.

4.2.3 Design process: Applying the “Mecha-telligence principle”

This sub-section provides, only, an overview to our robot design and its architecture.
Also it shows how the “mecha-telligence principles” have been applied in the process of
the design. The whole detail designs and all definition and description of the related
concepts are explained in the next sections. To deep understanding about our approach
and detail of the robot design, reading the these sections is recommended.

Designing function 1

Realization of the fast wake up can be achieved in two basic ways illustrated in Fig. 4.8.
By considering the environmental niche constraint and applying principle 4 , for a legged
robot with no hand the second alternative can be expected. To generate an internal
torque, we designed and developed a new type of actuator, called “Inertia actuator”
(Applying principle 5 ). Inertia Actuator absorbs mechanical energy by increasing its
rotor angular velocity and delivers energy by decreasing its rotor velocity. Inertia
Actuator can generate a small internal torque by changing the speed of the rotor and
a big internal torque in short time by using a brake to stop the rotor at high speed.
CHAPTER 4. ARTISTIC ROBOT 65

Figure 4.8: Two alternatives for fast wake-up action.

Designing function 2 and 3

In order to realize the jumping capability “Cam Charger” mechanism, presented as a

kicking device for RoboCup, is redesigned to fit to the foot shape (Applying principle
5 ). In this case we designed a unique and compact cam including a releasing part to
avoid the distribution of the force at the moment of springs energy releasing to the
foot structure. The action of somersault is performed by combination of jumping and
rotating using “Cam Charger” mechanism and “Inertia Actuator” (Applying principle
6 ).
For easy to follow what we did and what we designed, the results are visualized in
Fig. 4.9. The black color shows our solutions for the listed problems, related to the
mechanical design, in the mono-spec layer.

Figure 4.9: The results of design (Artistic robot). The black color shows our solutions for the
listed mechanical problems in the mono-spec layer.
CHAPTER 4. ARTISTIC ROBOT 66

4.3 Mechanical Design of “Jumping Joe”s

4.3.1 Inertia Actuator

Inertia Actuator consists of a H-shape cylindrical rotor (maximum diameter : 96 [mm],

inner diameter: 75[mm], weight: 1.8[kg]), a DC motor and a brake mechanism (see
Fig.4.10) is an inertial energy-storage device. Inertia Actuator absorbs mechanical
energy by increasing its rotor angular velocity and delivers energy by decreasing its
rotor velocity. In fact, the energy of Inertia Actuator is accumulated as the kinetic
energy of the rotor rotation. Inertia Actuator can generate a small internal torque by
changing the speed of the rotor and a big internal torque in short time by using a brake
to stop the rotor at high speed.[18, 19] The rotor is accelerated up to 6000 [rpm] within
20 [sec], and stopped suddenly using the brake mechanism in approximately 0.15 [sec].
The maximum generated torque is about 12 [Nm] where the moment of inertia of rotor
is 3.1 × 10−3 [kg.m2 ]. The brake mechanism is designed based on a rim type brake
with external contraction shoes, as illustrated in Fig.4.11. The rotor is arranged in the
center of the shoe brakes whose diameter including lining (brake material) is almost
same with the rotor diameter. When the force F is given to the point P (see Fig.4.12),
the force FC to the point C and FC to the point D are :

FD = F L1 /L3 (4.1)

Figure 4.10: Inertia Actuator architecture. Inertia Actuator consists of a H-shape cylindrical
rotor and a brake mechanism can generate a small internal torque by changing the speed of
the rotor and a big internal torque in short time by using the brake to stop the rotor at high
speed.
CHAPTER 4. ARTISTIC ROBOT 67

Figure 4.11: The brake mechanism designed based on a rim type brake with external contrac-
tion shoes. It stops the Inertia Actuator rotor, accelerated up to 6000 [rpm], in approximately
0.15 [sec].

Figure 4.12: Necessary geometric parameters of the brake mechanism for calculation of the
motive force F.

p
FC = F 1 + (L1 /L3 )2 (4.2)

Where, Li is the length of each link (L1 = 57.1, L2 = 13.0, L3 = 14.3, L4 = 44.8,
L5 = 8.0 [mm]). The length R (= 87.55 [mm] ) in Fig.4.12 is given as follows:

FC · R = FC (L2 − L5 )sinθ − 2FC L4 cosθ (4.3)

q
R = (L3 (L2 − L5 ) − 2L1 L4 )/ L21 + L23 (4.4)
CHAPTER 4. ARTISTIC ROBOT 68

Figure 4.13: The experimental result of the obtained torque from the developed Inertia Ac-
tuator. The black line means that the rotor speed is 2000 [rpm], the red line 4000 [rpm] and
the blue line 6000 [rpm].

Where, θ is the angle of the force FC to the link C-P, the relation between the
braking force F and braking momentum can be represented in the following equations
[20].
Z β2
Pa bra
MN = sinβ(r − a · cosβ)dβ (4.5)
sinβa β3

Z β2
f Pa bra
Mf = sin2 βdβ (4.6)
sinβa β3

FC = (MN − Mf )/R (If the rotor rotates clockwise) (4.7)

FD = (MN + Mf )/(2L4 + L3 ) (If the rotor rotates counterclockwise) (4.8)

Here, Pa (=542 [kpa]) is the maximum pressure of the brake material, b (=15 [mm])
is the width of the brake material, f (=0.32) is the friction coefficient between the
rotor and the brake material , r (=40 [mm]) is the distance between the center of rotor
and the brake shoe, a (=46 [mm]) is the distance between the center of rotor and the
joint B, and βa = β2 if β2 ≤ 90◦ otherwise β2 = 90◦ . The lining (brake material) is
the “J-2 frictional material" of STARLITE Co.,Ltd.. having a coefficient of friction (f )
of 1.8 at 25 [C]. The brake is actuated by a simple linkage mechanism connected to a
servo motor, shown in Fig.4.13 with 3 [Nm] torque as a peak. The selected servo motor
can create the required force F (=64.55 [N]) calculated in eq.(4.2) where β1 = 15◦
and β2 = 145◦ . The experimental results are shown in Fig.4.14. The rotor in the
CHAPTER 4. ARTISTIC ROBOT 69

Figure 4.14: This graph shows the braking time for the different rotation speed of the Inertia
Actuator rotor.

experiments is a cylinder with the outer diameter: 80 [mm], the inner diameter: 55
[mm], weight: 2.0 [kg] and the inertia moment: 2.3 × 10−3 [kgm2 ]). The obtained
torque increases according that the rotation speed is going up or the braking time is
going down. The maximum torque at 6000 [rpm] is approximately 11.5 [Nm]. The
braking time to stop the rotor with different rotation speed is shown in Fig.4.14. At
the operation velocity 2000 to 6000 [rpm], it takes between 0.15 and 0.2 [sec]. The
Inertia Actuator can generate big torque in a short time and the gyro effect to keep
the robot stable is also expected.

4.3.2 Hip Joint

A robot which performs acrobatic behaviors such as jump, waking-up, somersault, etc.,
needs very powerful , quick and small actuators. In this research, we utilized a parallel
mechanism [21] to the hip joints. The parallel mechanism can distribute external forces
given to the end-effector into each link and all output forces from all active joints are
synthesized and become an output of the end-effector. The output torque of parallel
mechanisms can be increased without a large ratio gear so that, a speedy and powerful
motion can be expected. On the other hand, parallel mechanisms has the disadvan-
tages such that the workspace is smaller than that of serial link manipulators, and the
kinematics and dynamics are complicated so that the computational power is needed
in order to solve these equations of motion in real-time. Most of applications of the
parallel mechanism into walking robots employ planar mechanisms such as pantograph
mechanism. Recently, the parallel mechanism with special movements is also applied
as the actuators of walking robots. Para-walker developed by Ota et. al. [22], has
the serial/parallel hybrid mechanism to climb over hazardous terrain, and WL_16 de-
CHAPTER 4. ARTISTIC ROBOT 70

Figure 4.15: The developed 2 DOF hip joint. Each arm has a servo motor in the middle joint
and the motion is restricted by the support bar in the center.

veloped by Sugawara et. al., is a biped robot on which an operator can get on. The
parallel mechanism of Steward platform is introduced into each leg [23].
The developed hip joint (See Fig. 4.15) consists of a base link, 3 serial-link arms,
3 servo motors (3 [Nm]), an end effector and a protector. Each arm is composed of 5
passive joints and 1 active joint as shown in Fig. 4.16. In the most applications, the
active joint is arranged at the end of serial-link arm. One of the specific feature of
the developed hip join is that the active joint locates in the middle of serial-link arm
in order to keep a larger workspace. The another special point is that the motion of
hip joint is restricted to 2 DOF motion, pitching and rolling motion by the support
bar in the center of hip joint in spite that the number of motors is 3 (Fig. 4.15).
The hip joint can output 10 [Nm] at maximum, and the ranges of pitching and rolling

Figure 4.16: The developed hip joint consists of 3 serial-link arms, and the each arm composes
of five passive free joints (joint 1,2,4 and 5) and one active joint (joint 3).
CHAPTER 4. ARTISTIC ROBOT 71

Figure 4.17: The coordinate system and parameters of developed hip joint.The developed 2
DOF hip joint.

angles are 110 [deg] and 45 [deg] and the weight is 850 [g]. The parallel mechanism has
characteristic singular point [21, 24]. If the posture is in the singular point, it’ll become
uncontrollable and becomes freely movable at specific directions. The developed hip
joint has no singular point within its working space.
The dynamics and kinematics of parallel mechanism are complex because of a lot
of active and passive joints, however, inverse kinematics can be solved easily for some
cases. Figure 4.17 shows the coordinate system and parameters of the developed par-
allel mechanism. The coordinate systems of the base link and the end-effector are
P P P
denoted by b and e , respectively. The attitude angle of the end-effector in b
is given by θe = [0, θey , θez ]T , where the angle along x axis is zero because the the
rotation around x-axis is restricted by the support bar. Let the position of the joint on
the end-effector Ji3 (i: index of arm) in e is e ei = [e eix , e eiy , e eiz ]T , and the position
P

of the joint on the base link Ji1 in b is bi = [bix , biy , biz ]T , respectively. The vector
P

from Ji1 to Ji3 , xei = [xeix , xeiy , xeiz ]T in b is expressed by eq. (4.9).
P

xei = B RE e ei − bi (4.9)

The rotation angle θi3 is expressed by following eq. (4.10).

| xei |2 − | xi1 |2 − | xi2 |2

θi3 = ±acos( ) (4.10)
2 | xi1 || xi2 |
CHAPTER 4. ARTISTIC ROBOT 72

4.3.3 Knee Joint

The knee joint is designed using Lever-Crank mechanism. The required range of motion
is 110 [deg] where the rotation angle of input joint is 180 [deg], and the required torque is
more than 8 [Nm] which is 1.3 times larger than input torque 6 [Nm]. Figure 4.18 shows
the geometric parameters of Lever-Crank mechanism. The input joint is the point A
and the output is D. By the way of Shirakawa’s triangular method, the relations of
angles and torques between A and D can be obtained with following equations. Where
the input angle is α , the length of BD is given as:

l52 = l12 + l42 − 2l1 l4 cosα (4.11)

then, the angle δ is:

l42 + l52 − l12

ε = cos−1 ( ) (4.12)
2l4 l5
l32 + l52 − l22
ζ = cos−1 ( ) (4.13)
2l3 l5

δ =ε+ζ (4.14)

Where the input torqueTa is given, the output torque Td can be calculated as follows
[25]:

Figure 4.18: The necessary geometric parameters of Lever-rank mechanism.

CHAPTER 4. ARTISTIC ROBOT 73

Figure 4.19: The developed knee joint using lever-crank mechanism. The points A, B,C, and
D correspond to those points of fig. 4.18. The joint angle is set to 0 in the right-top figure,
and approximately 110 degree in the right-bottom.

F1 = Ta /l1 (4.15)

Fbc = F1 /sinβ (4.16)

F2 = Fbc sinγ (4.17)

l4 sinγ
Td = F2 l4 = · Ta (4.18)
l1 sinβ
Where, the each length of links li is selected as 27[mm], 40 [mm], 40 [mm] and 30 [mm],
respectively. The overview of the developed knee joint is illustrated in Fig. 4.19. The
points A to D correspond to the those points of Fig. 4.18 and the joint angle δ is set
to 0 [deg] in the right-top and approximately 110 [deg] in the right-bottom of Fig 4.19.
The relation between the angle δ and the torque Td is shown in Fig. 4.20. The knee
joint takes a minimum torque 8.1 [Nm] where the angle δ is about 70 to 80 [deg].
CHAPTER 4. ARTISTIC ROBOT 74

Figure 4.20: The relation between the angle of knee joint and the obtained torque. The input
torque and range of knee joint angle are 6 [Nm] and 180 [deg], respectively.

4.3.4 Jumping Foot

In order to realize the jumping capability, Cam Charger mechanism [26] is introduced
to fit to the foot shape. The key idea is to charge a series of strong torsion spring by
using a special design of a cam shape. The basic concept of this mechanism is shown
in Fig. 4.21. With regard to the eqs. (4.19) and (4.20), the motive force (F ) to charge
the torsion spring is minimized while the force (F ) is perpendicular to the axes (m) (β
=90◦ ) where the torque τS takes a certain value.

τS
F = (4.19)
X · sinβ

Figure 4.21: Basic concept of cam charger mechanism. Joint A is fixed and the springs, as
a resistance torque, are mounted in this joint. τS = spring torque (resistant torque), F =
Motive force, X= distance between joint (O) and the point where force (F ) is applied, β =
angle between force (F ) and the direction of (m).
CHAPTER 4. ARTISTIC ROBOT 75

Figure 4.22: The coordinate and relation between the cam and the torsion spring. The main
function of the cam is keeping the β angle 90 degree in during the spring charging process.

τS
β = 90◦ =⇒ Fmin = (4.20)
X
The main function of the special designed cam is keeping the angle β 90◦ during the
spring charging process (Fig. 4.22). Then the torque of motor τC , needed to charge
the spring, will be minimized as following equation:

τC (min) = Fmin · d (4.21)

where d is the perpendicular distance between the direction of contact force F (line
N) and the center of the cam C. The distance d can be calculated with the following
equations:

Line M : y = x · tanθ (4.22)

Line N : x · cosθ + y · sinθ + l1 = 0 (4.23)

−l1 cosθ − l2 sinθ + l1

d= √ = −l1 cosθ − l2 sinθ + l1 (4.24)
cos2 θ + sin2 θ
Here, l1 is the length of OG, and the angle θ is the angle between the line OG and the
line M. And the force F and τC can be obtained as following:
CHAPTER 4. ARTISTIC ROBOT 76

Figure 4.23: The derivation of motor torque τC (red graph) and spring torque (blue graph)
respect to the charged angle of spring α where n= 2, k=3.0 [Nm/deg], α = maximum 15
[deg], l1 =150 [mm] and l2 =64.7[mm].

τS nkα
F = = (4.25)
l1 l1

τC = F · d = nkα(1 − cosθ − l2 sinθ/l1 ) (4.26)

Where n is the number of springs, k is the stiffness coefficient of springs and α is the
charging angle of springs. In this mechanism, during the charging time, the perpendic-
ular distance between the direction of contact force F and the center of the cam (d in
Fig.4.22) decreases when the spring torque (τS ) and the contact force (F ) increase. The
red line in Fig.4.23 shows the derivation of motor torque relative to the charged angle
of springs. The important point is that when the spring is charged more and more,
the required motor torque is getting less and less in the second half of the graph in
Fig.4.23. This mechanism makes charging very strong series of torsion springs possible
with a small motor. As an example in case of a cam charger with specification n= 2,
k=3.0 [Nm/deg], α = maximum 15 [deg], l1 =150 [mm] and l2 =64.7[mm], for charging
the 90 [N.m] resistance torque of two springs, we only need a motor-gearhead which
can produce almost 2.2 [N.m] torque as a peak.
In order to realize the jumping capability, the cam charger mechanism is designed so
as to fit to the foot shape. The developed jumping foot, illustrated in Fig. 4.24, consists
of (1) two torsion spring, (2)a special designed cam, (3)a needle bearing, (4)a releasing
CHAPTER 4. ARTISTIC ROBOT 77

Figure 4.24: (A): To realize the jumping capability, Cam Charger mechanism was designed
to fit to the foot shape: (1)torsion spring(s), (2)special designed cam, (3)needle bearing,
(4)releasing part, (5) motor and (6) worm gear box. (B): This figure illustrates the developed
jumping foot with capability of jumping.

part, (5)a motor and (6)a worm gear box.The cam is designed as the combination of
successive circular arc as shown in Fig.4.25-A by using 3D CAD software Auto-desk
Inventor. Figure 4.25-B shows the final design of cam where α = 15 [deg], l1 =150 [mm],
and l2 =64.7[mm]. As mentioned in above, τc takes the maximum value around middle
of charging angle and becomes 0 at the end of charging, because the direction of F is
toward the center of cam and d is 0. The diameter of needle bearing is 9 [mm] and
the maximum radius of cam is 60.2 [mm] in the case that the l1 is set to 150 [mm].
In this case, Cam Charger can charge approximately 41 times strength torsion spring
of the maximum torque of the motor. The experimental results of the developed Cam
Charger are shown in Fig.4.26. A Cam Charger can make the 5 [kg] weight (including

Figure 4.25: (A): The cam is designed as the combination of successive circular arc by using
Auto-desk Inventor software. (B): The cam is finalized where α = 15 [deg], l1 =150 [mm], and
l2 =64.7[mm].
CHAPTER 4. ARTISTIC ROBOT 78

Figure 4.26: The experimental results of the Cam Charger jumping. The Cam Charger can
make the 5 [kg] weight (including own weight) jump up to approximately 200 [mm] height.

own weight) jump up to approximately 200 [mm] height. The required jump capability
that we assumed is 200 [mm] where the weight of robot is 10 [kg], therefore, the
developed Cam Charger suffices the requirement if the two Cam Charger can work at
the same moment.

4.4 Electric Circuit Design

4.4.1 Total System Architecture

The system architecture of the full function Jumping Joe is illustrated in Fig.4.27.
The commands from the host PC are transmitted by RS232 serial communication, and
converted to the Bluetooth wireless communication with “Bluestick"s. We adopted
the Bluetooth device because that is small and easy to use within the range of 10
[m]. The host command is sent to the FPGA board through a PIC, and then, the
FPGA sends the behavior of the robot to the two Master PICs (one for right side
and one for left side) and the control board of the Inertia Actuator (Fig.4.27-B). The
Master PICs communicate with all PICs of actuators control circuits using the I 2 C
protocol (Fig.4.27-A). After a behavior is transmitted to all actuator circuit boards,
the Master PICs send the trigger command and the robot starts the performance. As
a vision system, the silicon retina developed by Yagi et. al. [27] is introduced. The
silicon retina recognizes the motion in the front of robot such as up, down, left or right
motion, and sends a command to the FPGA board.
CHAPTER 4. ARTISTIC ROBOT 79

Figure 4.27: Electric system architecture of Jumping Joe. A: Communication between Master
PICs and the actuator modules and the power distribution, B: FPGA circuit board and C:
Control board of the hip joint.

4.4.2 Control Board of Each Actuator

Each actuator module includes its own microcomputer(s) PIC(s) and communicates
with the master PIC via I 2 C protocol. As an example, the design of the hip joint
control board is shown in Fig.4.27-C. The power source of micro controllers and servo
motors are isolated using photo couplers. The servo motors are controlled by PWM
signals which correspond to the joint angles. The joint angles are used as the feedback
signals. The sampling and control rates are 100 [Hz]. The developed control boards of
all actuators are illustrated in Fig.4.28.

4.5 Modular Architecture of “Jumping Joe”s

To achieve the reliability and easy to assemble and maintenance features, “Jumping
Joe"s are designed to have a modular architecture in their mechanical parts and eclectic
circuits (Applying principle 7: The mechatronic modularity principle). “Jumping Joe"s
are made of combination of four main modules: Upper body, Hip joint, Knee joint and
Jumping foot. All modules can be plugged to each other via a standard male/female
CHAPTER 4. ARTISTIC ROBOT 80

Figure 4.28: The developed control boards of Inertia Actuator, Hip joint, Knee joint and Cam
Charger.

developed connector, as showed in Figs. 4.16, 4.19 and 4.24-B, which is designed to
work as a mechanical joint and also an electric joint to transmit the power and data.
By using the modular architecture for “Jumping Joe"s we can realize the different
kinds of artistic robots, only, by plugging the different modules to each others (see
Fig.4.29). Upper body module itself divided into 3 modules, which are (a) the battery
part located in the left side, (b) the “Inertia Actuator" part in the middle and (c) the
electric circuit part in the right side (see Fig.4.30). These three modules can be plugged
to each others through their own connectors. Hip joint, Knee joint and Cam Charger
modules, as mention in fourth section, include their own microcomputer(s) PIC(s) and

Figure 4.29: By using the modular architecture for “Jumping Joe"s we can realize the different
kinds of artistic robots, only, by plugging the different modules to each others. All Jumping
Joe robots, at least, consist of a upper body and two cam chargers.
CHAPTER 4. ARTISTIC ROBOT 81

Figure 4.30: Upper body module contains battery part (a), Inertia Actuator part (b), electric
circuit part (c) that they plug to each other through their own connector. A standard devel-
oped female connector is installed in bottom side of electric part and battery part (not show
in the figure).

communicate with the master PIC, located on electric circuit part in the upper body,
via I 2 C protocol.

4.6 Simulation and Experimental Results

In this section we will show some samples of simulation and experimental results of
Jumping Joe robots. Figures 4.31 and 4.32 show the simulation and experimental
results of Jumping Joe-Junior 1 during the rapid wake up and somersault action,
respectively. For rapid wake up motion, the rotor is rotated up to 6000 [rpm] and
stopped suddenly using the brake mechanism in approximately 0.15 [sec]. In case of
somersault, the rotor is rotated up to 6000 [rpm] at first, then jumping action is done
by Cam Chargers, and after a few milliseconds the brake mechanism stops the rotor
suddenly. In the simulation results of the Jumping Joe-Junior 1 during the somer-

Figure 4.31: Simulation (A: side view) and experimental (B: front view) results of the rapid
wake up of Jumping Joe-Junior 1.
CHAPTER 4. ARTISTIC ROBOT 82

Figure 4.32: Simulation (A: side view) and experimental (B: front view) results of the somer-
sault of Jumping Joe-Junior 1. In the simulation results (A), the robot twisted in the frames
at 4.1 sec and 4.2 sec so that it is shown the simulation will correspond with the actual
experimental response (B).

Figure 4.33: Simulation (A: side view) and experimental (B: side view) results of the somer-
sault of Jumping Joe-Junior 2.

sault action (Fig. 4.32_A), the robot twisted in the frames at 4.1 sec and 4.2 sec so
that it is shown the simulation will correspond with the actual experimental response
(Fig. 4.32_B). The result of somersault simulation and experiment of the Jumping
Joe-Junior 2 is illustrated in Fig. 4.33. The hip joint action to make the robot as much
as possible close to the circle shape and jumping action is almost done in the same
time and after a few millisecond the rotor, which is rotated up to 6000 [rpm] in the
beginning, stopped suddenly by the brake mechanism.

4.7 Conclusion
In this chapter, we describe the design, modeling, simulation and implementation of
entertainment robots named “Jumping Joe". “Jumping Joe"s, artistic and agile robots,
can perform several rapid movements such as fast wake up, jumping and somersault.
In order to realize acrobatic movements, four different actuators which can create the
high-speed movements are developed. Inertia actuator, installed in the upper body,
is a torque generator that consists of Gyro and brake mechanisms. The each of hip
CHAPTER 4. ARTISTIC ROBOT 83

joints consists of a 2 DOFF actuator developed on a basis of parallel link mechanisms

for achieving light weight and high torque. The knee joints have 1 DOFF, high torque
mechanisms employing a lever-crank mechanism. Each jumping foot has 1 DOFF
and is designed based on a special mechanism named “Cam Charger " consists of a
special designed cam and torsion springs. we still have to perform further experiments
and improvements to achieve the proper control method for robot balancing during
somersault action. Calculation of the power consumption, required to execute wake-
up, jumping and somersault actions is on the future work.
 Chapter 5
Sewer Pipe Inspection Robot
Chapter 5

Sewer Pipe Inspection Robot

5.1 Introduction
Major cities invest considerable amounts of budget in underground infrastructure in-
cluding water and sewage pipes. Many of these pipeline system are prone to damage
due to aging, excessive traffic, earthquakes and chemical reaction. Sewage may leak
out, possibly polluting soil and ground water, and it may wash away soil, possibly
eroding the foundations of buildings or the underground of streets and pavements.
The rehabilitation of underground pipes is costly and slow in inspection process. The
size of most pipes eliminates the possibility of direct human inspection1 and requires
employment of sophisticated equipment in the process of pipe inspection. With regard
to the current sewer pipe inspection technology, all commercial robots are completely
tel-operated, usually via a tether cable, by a human operator. In addition, current
sewer inspection robots have a poor mobility function to pass any kind of pipe-bends
such as curves and junctions so that those robots are only capable to move into the
straight pipes. Inspecting the sewage pipes using the state of the arts inspection meth-
ods by the current robots is costly, mostly human cost, and not fast enough to check
and inspect the amount of sewage pipes will grow stronger than it has actually hap-
pened, specially in Japan. In this part of thesis we show our method for realization of
an innovative, fast and robust sewer inspection method by using a passive-active in-
telligent, fully autonomous, UN-tethered robot called "KAN TARO" which fits to the
pipes within a diameter range of 200-300 millimeters. KAN TARO prototype robot,
including a novel passive-active intelligent moving mechanism (nadir Mechanism), can
move into the straight pipe and pass different kinds of pipe bends without need to any
intelligence of the controller or sensor reading. In order to realize a fully autonomous
1
In Japan mostly 85% of the sewer pipes are under 600mm diameter and mainly more than 65% of this
amount have a diameter within range of 200-300mm [28]. This is the target diameter of the presented fully
autonomous mobile robot “KAN TARO” in this chapter.

85
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 86

inspection robot, we also developed a small and intelligent 2D laser scanner for detect-
ing of the navigational landmarks such as manholes and pipe joints independently with
main computer system, and fusion with a fish eye camera to assist the pipe state and
fault detection.

5.1.1 State-of-The-Art

In current conventional methods, the inspection of sewer pipes is carried out using a
cable-tethered robot with an on-board video camera system. An operator remotely
controls the movement of the robot including a video system (Fig. 5.1) and looks for
probable damages in the images of the pipe interior. All equipment necessary to supply
and control the robot are arranged inside of a car, out of the manhole (definition of
manhole has been specified on page 106), which causes a heavy and stiff cable for the
robot. In addition, all commercial sewer inspection robots have a poor mobility function
to pass any kind of pipe-bends such as curves and junctions so that those robots are
only capable to move into the straight pipes. These two main reasons, “heavy and stiff
cable" and “poor mobility function", make a limitation for robot movement that, in
most case, the robot is driven only from one manhole to the next, and the robot is
driven backward through the pipe to its entry point. Because the robot is tethered
to the control unit in the car and disconnecting the power cord in the next manhole
has a risk to expose the wires to water. Then the robot is retrieved, and the process
is repeated again at the next manhole. This method for inspection of the sewer pipes
makes the inspection process very slow (300 meters a day) and costly (2000 yen/20
dollars per meter) in Japan.[28] In order to realize inexpensive and effective inspection
system, an autonomous pipe inspection method should be introduce to improve the

Figure 5.1: In current conventional method, the inspection of sewer pipes is undertaken using
a cable-tethered robot with an on-board video camera system. An operator remotely controls
the movement of the robot including a video system.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 87

inspection efficiency by reducing the time and manpower of the inspection process.

5.1.2 Qualitative Degrees of Autonomy in Sewer Robots

To focus the study, also, for the sequel, we want to make more precise the different
qualitative degree of autonomy in developed inspection sewer pipe robots that we will
mention:
No autonomy: The robot is completely tel-operated, usually via a tether cable, by
a human operator. The pipe condition is assessed by the human operator who watches
the sensor data (usually video) as the robot drives through the pipe. Mostly all the
commercial sewer inspection robots are not autonomous system.
Semi-autonomy: The tethered robot (in a few case UN-tethered) is partially
controlled by automatic control programs and modules, or the assessment of the pipe
condition is partially performed by sensor data interpretation programs. There is
a number of researches and developments of the robots with the semi-autonomous
function capability.
Full autonomy: The UN-tethered robot carries all required resources on-board.
Navigation is performed completely by control programs running on on-board com-
puting equipment. Status messages may be communicated to a human inspector over
a radio link. Assessment of the pipe condition may be performed partially on-board,
or offline after retrieval of the recorded sensory data. A few research have been done
in development of a fully autonomous mobile robot for pipe inspection.Full auton-
omy: The UN-tethered robot carries all required resources on-board. Navigation is
performed completely by control programs running on on-board computing equipment.
Status messages may be communicated to a human inspector over a radio link. As-
sessment of the pipe condition may be performed partially on-board, or offline after
retrieval of the recorded sensory data. A few research have been done in development
of a fully autonomous mobile robot for pipe inspection.
In general, we observe that fully autonomous robots are not yet marketable system
today. There are two main reasons for this:

• The degree of development of complete autonomous sewer robots presently does

not warrant to use them safely and robustly in sewers. As mentioned in above,
most of these robots have a complex moving mechanism and Multi-sensor equip-
ment for navigation and motion control. These complexities in mechanism and
data processing make not easy to realize reliable commercial products specially
for small range of the pipes up to 300 millimeter in diameter.

• There seems to be some general skepticism about using fully autonomous systems
without a possibility to interfere the robot control at any time, as would be the
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 88

case for a tether-less sewer robot out of the range of some radio link.

However, semi-autonomy does make sense, implementing in a tele-operated sewer robot

separate modules (like special sensors with data interpretation software, or special
motion control module) that operate autonomously within the overall control of a
human operator. The purpose of such embedded autonomous modules would be to
leverage the work load of the human operator in controlling the robot and interpreting
sensory data, or to make certain measurements like in sewer inspection automatic, and
therefore operational and reproducible. This is a serious issue in ensuring the quality
of sewer maintenance, given that the pressure to provide service as cheep as possible
imperils the quality of the inspection results.[28, 29]

5.1.3 Sewer Inspection Robots

Autonomous Sewer Robots Platform2

As we mentioned before, toady’s state of the art in sewer maintenance involves tele-
operated, tethered vehicles both for inspection and manipulation. To enhance the ease
of use and the reflectivity efficiency of these platforms, there are in general two ways
of helping to handle and control them: first, support the human operator in his or her
control of the vehicle by making automatic part of the control; and second, make the
platform or parts of it completely autonomous, thereby taking the operator completely
out of the control loop or parts of it.3

KURT

The development of the experimental sewer robot test platform KURT (Kanal- Untersuchungs-
Roboter-Testplattform) started at the former GMD now Fraunhofer institute AIS in
1995 [30]. KURT is a monolithic, non-holonomic six-wheeled autonomous un-tethered
robot of approximate dimensions 30x45x30 cm. KURT version 1 has been successfully
employed for navigating autonomously in a dry sewer test net at the premises of the
Fraunhofer campus in Sankt Augustin. To achieve this, the robot is provided with a
map of the test net, representing the topology of the 80 meters of sewer pipes and the
nine manholes in between, with a start position (number of a manhole) and a goal
manhole. The robot can determine the sequence of manholes - or pipe junction types,
respectively - that it should pass on its path from start to goal. Since all pipe junctions
inside the sewer test net are ground level connections, the robot is mechanically able to
2
Using the term "platform" means to emphasize the fact that the respective systems do not include sensors
for finding damages or assessing the pipe state, but just sensors to warrant their safe navigation and control.
Such pipe-related sensors maybe be implemented on autonomous platforms.
3
Most of the materials presented in this section have been selected from material provided by “Steinbeis
Japan Inc., Kitakyushu foundation for the Advancement on Industry, Science and Technology”, March 2002.[28]
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 89

Figure 5.2: The commercial research robot platform KURT2. Figure shows a KURT2 system
with a custom mounted 2D laser distance scanner.

perform turns at such junctions. With its pivoted ultrasound sensor, KURT1 is able to
classify the type of a pipe junction, i.e. whether it is L-shaped, X-shaped, or T-shaped.
A special patented method for navigation under uncertainty enables KURT1 to detect
and correct errors due to false classification or due to errors while performing turns at
pipe junctions [31]. This work has been complemented by a method for probabilistic
mapping of objects - like landmarks - in the sewer [32]. Ever since its very first version,
the KURT type robot platform has been further developed for indoor applications.
The system is now being marketed by GMDs spin-off KTO as a customizable general
purpose robot research platform. The current version KURT2 can be equipped with
a variety of sensors. The standard configuration includes inclinometers and sensors
for odometry, either infrared or ultrasound distance transducers for obstacle detection,
and optional bumpers. Figure 5.2 shows a KURT2 system with a custom mounted
2D laser distance scanner. Alternatively, an industry standard PC/104 CPU can be
provided.

MAKRO

MAKRO (Mehrsegmentiger Autonomer KanalROboter / multi-segmented autonomous

sewer robot) is the prototype of a fully autonomous, un-tethered, self-steering artic-
ulated robot platform (Fig.5.3). It is designed for autonomous navigation in roughly
cleaned sewer pipes within a diameter range of 300 to 600 millimeters at dry weather
conditions. MAKROs case design, consisting of six segments connected by five motor-
driven active joints, allows for simultaneously climbing a step and turning, e.g. at a
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 90

junction consisting of a 600 millimeter pipe and a branching 300 millimeter pipe with
equal top levels. MAKROs autonomy and its kinematic abilities extend its potential
mission range enormously, compared to conventional inspection equipment that is lim-
ited by the cable and poor kinematics [33]. MAKRO carries all the needed resources
on-board. Standard NiCd batteries provide the power for its 21 motors, the sensors, and
the electronics, including an industry standard PC/104+ computer system and seven
micro controllers [34], allowing for an autonomous uptime of about two hours. The
goal of the MAKRO project was to prove that a robot is able to navigate completely
autonomously inside sewers under the above mentioned conditions. In a way similar
to that of KURT (see previous subsection), MAKRO takes a start and a goal manhole
position and a topological sewer map, computes a path from start to goal, and drives
autonomously along this path from the start to the goal position. Proprioceptive and
external sensors provide all the data that MAKRO needs to navigate and to control its
pose and locomotion. The robot is steered automatically by control programs running
on its CPU and micro controllers. MAKRO incorporates a number of internal, propri-
oceptive sensors. A thermometer for measuring the CPU temperature, pulse counters
for odometry, optical sensors for reading joint angle encoding, and more. Sensors for
data acquisition from the environment are mainly located in both identically equipped
head segments. Each of them contains a stereo camera pair, lighting equipment, an
ultrasound transducer for obstacle detection, four fixed infrared distance transducers,
and a custom IR scanner for household inlet detection. Low data rate sensor signals

Figure 5.3: MAKRO is the prototype of a fully autonomous, un-tethered, self-steering articu-
lated robot plaform. It is designed for autonomous navigation in roughly cleaned sewer pipes
within a diameter range of 300 to 600 millimeters at dry weather conditions.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 91

are gathered by the micro controllers and transferred via CAN bus to the CPU for fur-
ther interpretation. Only the camera signals are directly transmitted to the PC/104+s
frame grabber. The multi-sensor equipment is being used for navigation and motion
control. Self localization, that is the automatic determination of the robots position
inside the sewer, is based upon recognition of salient landmarks. Inside a sewer, only
few structures are present that may serve as recognizable landmarks. These are house
inlets, pipe junctions at manholes, and, less salient, pipe joints. The first two types
of landmarks can be detected by interpreting the data of the custom IR scanner [35].
Ground-level pipe junctions may also be detected by the robots fixed infrared transduc-
ers. The camera equipment has been employed for various image data interpretation
experiments, including 3D measurement of sewer pipe junctions, and visual recognition
of household inlets [36]. Each MAKRO segment is propelled by a motor that drives
a two-wheeled axis, and each joint connecting two segments contains three motors
that allow for rotations around the three axes in space. The joint motors are strong
enough to lift a part of MAKROs 50 kilograms heavy, 2 meters long body, which is
necessary for climbing steps at some pipe junctions. In order to perform turning and
climbing maneuvers, MAKRO has to simultaneously control most of its 21 degrees of
freedom. This is a highly complex control task, and it is performed automatically by
special control routines developed at Fraunhofer AIS. The routines rely on data from
proprioceptive sensors like joint angle readers, and on external sensors like the infrared
transducers, which can be employed to detect branching pipes at junctions and to es-
timate the distances of the head segment to the pipe walls during turning maneuvers.
The alignment of the head segment with respect to the sewer pipe axis is automatically
determined by a method that interprets images of a structured light pattern projected
onto the pipe walls [37]. The robot MAKRO has been developed by a German group of
two research institutes and two industrial partners. The MAKRO project was funded
partly by the German Ministry for Research and Education (BMBF) between 1997
and 2000. Since 2001, the MAKRO project is being continued as internal projects at
Fraunhofer AIS and FZI [38].

Semi-Autonomous sewer robots

KARO

KARO (KAnalROboter/ sewer robot) is an experimental semi-autonomous carrier for

sewer inspection sensory equipment [39, 40, 41]. It was developed by a group of re-
search institutes and some industrial partners in Germany. The project was partly
funded by the German Ministry for Research and Education (BMBF). The monolithic
KARO robot prototype (see Fig. 5.4) resembles much a standard non-autonomous
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 92

sewer inspection robot, and it is tethered via a cable to a surveillance unit. Using in-
clinometers and an on-board control program, KARO is able to automatically correct
for tilt in its pose and wheel slippage when driving inside a sewer pipe, thus freeing the
human operator from this crucial control task [42]. The main innovation of KARO is
its wide range of novel inspection and navigation sensors, namely, a microwave sensor
and a 3D optical sensor, complementing the standard video camera and some ultra-
sound transducers [43]. The method for 3D measurement of hollow spaces using axial
symmetric structured infrared light has been patented. It is applied for measuring
pipe deformations, larger pipe cracks and obstacles inside the pipe, the latter being
detected early by the US transducers. The microwave sensors are aimed at detecting
leakages. KARO has been continued as an internal research project within Fraunhofer
IITB at least until 2000. Most recent research deals with fuzzy methods for data fusion
of inspection sensors. A commercial version of KARO is not yet available.

Pipe Rover / Pearl Rover

The development of the semi-autonomous Pipe Rover [44, 45] inspection robot for water
filled pipes or ducts started as a research project in 1996. The specific application
was driven by the needs of the City of Hong Kong, with its many sewers that are
well accessible from Hong Kongs sea side. The project produced a spin-off company
called Pearl Technologies Ltd. that markets the Pipe Rover. This inspection robot
can operate in two locomotion modes. For flat-bottomed ducts, with little marine
growth or other obstructions, tracks propel the robot to its destination. For round
pipes, or where there significant marine growth or the possibility of obstructions, legs
are used. The robot is electrically powered, although the legs are powered by an in-

Figure 5.4: KARO is an experimental semi-autonomous carrier for sewer inspection sensory
equipment. It was developed by a group of research institutes and some industrial partners
in Germany.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 93

Figure 5.5: PIRAT is a semi-autonomous tethered robot for the quantitative and automatic
assessment of sewer condition.

house designed hydraulic system that make use of the surrounding water. The robots
sensor system includes an ultrasonic peripheral pipe contour scanner, one color zoom
camera with pan and tilt, and one black and white rear facing camera for piloting in
reverse. Communication can be achieved either via the fiber-optical tether, or, for the
untethered version of the robot, via an ultrasonic link that will transmit low-speed,
sampled color or black and white video to the control console with a sampling speed
of less than one frame per second [46, 47]. With its on-board batteries, the untethered
version of the Pipe Rover has a maximum mission duration of about four hours.

PIRAT

The PIRAT sewer inspection system has been developed between 1993 and 1996 within
a joint project of the Manufacturing Systems and Automation Division of Australian
CSIRO Manufacturing Science and Technology and Melbourne Water. PIRAT (Pipe
Inspection Real- Time Assessment Technique) is a semi-autonomous tethered robot
for the quantitative and automatic assessment of sewer condition [48, 49]. Just like a
conventional sewer inspection device, PIRAT (Fig. 5.5) is deployed to a sewer, and
tele-operated from a surveillance unit via a cable by a human operator. The maximum
cable length of 250 meters gives PIRAT a fair operating range. The added value of
the PIRAT system is its ability to perform automatic evaluation of its sensory data.
PIRATs innovative instrument system contains a video camera and a laser scanner. For
flooded sewers, the latter can be substituted by a sonar scanner, but at the price of less
resolution and inspection speed. In 600 millimeter sewer pipes and at PIRATs usual
inspection speed, the laser scanner produced a resolution of about 1.5 mm radially
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 94

Figure 5.6: PIPER is a semi-autonomous tethered sewer inspection robot. Two user-
programmable micro controllers provide for controlling locomotion and for automatically ad-
justing pan and tilt of the video camera head to the robots locomotion speed.

and 4mm axially and circumferentially. The sensory data are evaluated by means of
PIRATs interpretation system, that is an expert system that runs on a Sun workstation
in the mobile surveillance unit. From both types of scanner data, the interpretation
system generated in a first step a three-dimensional model of the scanned sewer pipe
section. In a second step, the interpretation system uses techniques from Artificial
Intelligence to detect and classify damages on the basis of the three-dimensional model
data. The result is a sewer condition assessment report that is readable for the human
operator.

PIPER

Piper is a semi-autonomous tethered sewer inspection robot (Fig. 5.6), manufactured

by Angelus Research Corporation. Two user-programmable micro controllers provide
for controlling locomotion and for automatically adjusting pan and tilt of the video
camera head to the robots locomotion speed. The video images can be either taped
or captured on a computer. Its dimensions (9 inches width and 13.5 inches height)
prohibit its application in smaller pipes. Pipers operating range is limited by the
mandatory cable of about 45 meters.

PipeEye

The company Automatika, based in Pittsburgh, PA, is developing the prototype of a

neutrally buoyant untethered pipe inspection robot called PipeEye (Fig. 5.7). The
system shall be capable of providing video and acoustic imaging of underground sewer
or water pipelines. This 12 inch robotic sphere is designed to operate floating in water
filled pipes with diameters of at least 24 inches. A wide-angle color CCD video camera
and spotlights are mounted in the part that stays above the water surface. Mounted
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 95

Figure 5.7: PipeEye is the prototype of a neutrally buoyant untethered pipe inspection robot,
capable of providing video and acoustic imaging of underground sewer or water pipelines.

in the lower part, ultrasonic transducers shall yield imaging data from the underwater
part of the conduit. The robot shall be steered by control programs running on an
on-board PC/104 industry standard computer powered by on-board Li-ion batteries.
Imaging data shall be either recorded to a flash disk or transmitted in compressed
form via radio link through a manhole. The PipeEye system is derived from an oil/gas
pipeline inspection module co-developed by Automatika and Shell Oil. Currently, the
system does not yet have a product status.

5.2 Summary of design based on “Mecha-telligence methodol-

ogy”
In this section the mecha-telligence methodology will be described for design and im-
plementation of a autonomous mobile sewer robot. The first step in “Mecha-telligence
methodology” as shown in Fig. 2.12 is to describe the tasks, desired behaviors, envi-
ronmental niche, and environmental niche constraints.

Robot task

In general the sewer maintenance is classified to two categories:

• Inspection: It means assessment of the sewer pipe interior network and detecting
the existing faults such as crack, water infiltration, root invasion, pipe break, and
etc.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 96

Figure 5.8: The definitions of robot task.

• Manipulation: It means repairing of the sewer pipe. It processes after the inspec-
tion the pipes.

In this industrial research study we concentrate to the inspection of the pipes with in
diameter of 200, 250, and 300 [mm]. Then the robot task can be defined as “inspection”
not “manipulation” (see Fig. 5.8).

Desired behavior

Based on state-of-the-art and autonomous sewer inspection scenario the desired behav-
iors can be described in three following items: (Fig. 5.9)

• Insert the robot from a manhole

• Robot moves in sewer network autonomously (without operator intermediary) as

the same water-flow direction.

Figure 5.9: The definitions of robot task.

CHAPTER 5. SEWER PIPE INSPECTION ROBOT 97

• After retrieval the robot, the number and position of the faults can be extracted
from the robot sensory data.

Environmental niche and relative constraints

Alternatively, the environmental niche and the relative constraints can be defined and
itemized as follows: (Fig. 5.10)
Environmental niche:

• Outdoor environment (Direct-physical meaning)

• Circle sewer pipe: main pipe (Direct-physical meaning)

• Manhole (Direct-physical meaning)

• Inlet (Direct-physical meaning)

• Pipe bend: Curve, Y- and T-Junction (Direct-physical meaning)

Environmental niche constraint:

• Diameter of main pipes are 20 or 25 or 30 [cm] (Direct-physical meaning)

• Diameter of manholes are 90 or 120 [cm] (Direct-physical meaning)

• There is a manhole over all pipe-bends (Direct-physical meaning)

• Maximum distance between to manhole is 50 [m] (Direct-physical meaning)

Figure 5.10: The definitions of environmental niche and relative constraint (Sewer robot).
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 98

Figure 5.11: The selected type of sewer robot.

• All pipe has at least 2 or 3 degrees slop: Water can flow into the pipe in one
direction (Direct-physical meaning)

Degree of autonomy and mobility

Degree of autonomy of the presented system can be summarized as follows:

• 1- The un-tethered robot carries all required resources on-board.

• 2- Navigation is performed completely by control programs running on on-board

computing equipment.

• 3- Status messages may be communicated to a human inspector over a radio link.

• 4- Assessment of the pipe condition may be performed partially on-board, or

offline after retrieval of the recorded sensory data.

Also, the robot should be designed as a wheel type-untethered robot (Fig. 5.11).

5.2.1 Simplification process of design functions

High-level specification layers (Main layer and Sub layer)

Regard to the description of four above terms, the high-level specification layers in-
cluding main-layer and sub-layer can be illustrated as Fig. 5.12.

Low-level specification layers (Layer 1, layer 2, and mono-spec layer)

Figure 5.13 shows the definition of layer 1 extracted from sub-layer. The significant
point in this generation is that because sewer inspection is done the same direction of
water flow, robot does not need to perform going step-up. Consecutively, Figs. 5.14
and 5.15 show the generation of layer 1 and mono-spec layer, respectively. In these
three figures the red color shows the terms which are on the process of simplification,
white color illustrates the terms that are not important for our design purpose, and
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 99

Figure 5.12: The high-level specification layers including main-layer and sub-layer (Sewer
robot).

Figure 5.13: Layer 1 generated from the sub-layer (Sewer project).

yellow color indicates the terms which recognized as a simple basic function or selecting
a sensor for a single task or behavior.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 100

Figure 5.14: Layer 2 generated from the layer 1 (Sewer project).

Figure 5.15: Mono-spec layer generated from the layer 2 (Sewer project).
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 101

Figure 5.16: The function priority of design (Sewer project).

Functions priority of design

Figure5.16 shows the result of our simplification process from “main layer” to “mono-
spec layer”. Also, this figure illustrates the function design priority in point of view of
mechanical parts design.

5.2.2 Design process: Applying the “Mecha-telligence principle”

This sub-section provides, only, an overview to our robot design and its architecture.
Also it shows how the “mecha-telligence principles” have been applied in the process of
the design. The whole detail designs and all definition and description of the related
concepts are explained in the next sections. To deep understanding about our approach
and detail of the robot design, reading the these sections is recommended.

Designing function 1 to 6

Figure 5.17 shows two possible model of water flowing inside of the pipe (Applying
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 102

principle 4: system-environment interaction). As a common sense, the right side model

Figure 5.17: Two possible model of the water flowing inside of the pipe and comparing with
a commercial vertical robot and naSIR mechanism presented in this thesis.

can be recognized as a intuitive model. By surveying on the existing commercial

mechanisms and considering these two models we can realize that the robot wheels
should be rotated 90 degree to get more national movement same as water flowing
inside of the pipe. In this approach we can find out that the robot, not only, can move
in straight pipe, but also can pass various kinds of pipe bends without any control
or sensor reading. Consecutively, figure 5.18 shows the extension of the design for
performing more smooth movement of “naSIR mechanism” during passing a curve, Y-
junction and T-junction (Applying principles 4 and 6). This mechanism can pass a
small step, depend on the robot size between 5 to 10 [cm], because of its suspension
system. Passing a big step is realized with installing a passive elevator inside of the
manhole. The robot can get on to the elevator and the elevator will be activated by
the robot weight (Applying principles 3, 4, and 6 ).
 CHAPTER 5. SEWER PIPE INSPECTION ROBOT

Figure 5.18: Design extension for performing more smooth movement of “naSIR mechanism” during passing a curve, Y-junction and
T-junction.
103
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 104

Designing the other functions

Figures 5.19 and 5.20 show our approach for passive adjustable the robot tilt angle angle
and the robot navigation. We applied the principle 4 in form of material morphology

Figure 5.19: The robot can adjust its tilt angle passively by using the different material for
the front and back wheels.

Figure 5.20: Shows the realization of a precise motion navigation for the robot by installed a
passive encoder on the top of the robot.

to remove the tilt sensor from the robot hardware. we used the different material for
the front and back wheels. In this approach the robot can adjust its tilt angle passively.
Also we installed a passive encoder on the top of the robot by a simple arm jointed to
the top-back of the robot. In this approach we could realize a precise motion navigation
for the robot includes counting the numbers of manholes that the robot are passing.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 105

Figure 5.21: The results of design (Sewer robot). The black color shows our solutions for the
listed problems in the mono-spec layer.

When the passive encoder does not send signal to the on-board main computer, it
means robot is moving in the manhole.
For easy understanding of what we did and what we designed, the results are visu-
alized in Fig. 5.21. The black color shows our solutions for the listed problems in the
mono-spec layer.

5.3 A New Approach to the Autonomous Sewer Pipe Inspec-

tion
5.3.1 An Overview to the Sewer Pipe System in Japan

In Japan mostly 85% of the sewer pipes are under 600mm diameter and mainly more
than 65% of this amount have a diameter within range of 200-300mm [28]. In this
range of pipes, a sewer pipe net is constructed, mostly, by combination of pipes, pipe
bends, manholes, pipe joints and inlets (Fig. 5.22). In general, any changing in the
direction of a pipe or any intersection between two or more pipes called pipe-bend.
Pipe-bends in a sewer, as a whole, can be classified in two main types depending on
their constructions and shapes:
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 106

Figure 5.22: An example of a network of sewage pipes, in Japan, within a diameter range
of 200-300 millimeters. In this range of pipes, a sewer pipe net is constructed, mostly, by
combination of pipes, pipe bends (Junction sand curves), manholes, pipe joints and inlets.
Red arrows show the direction of water flow into the pipe net.

• Curve: It defines as a pipe curvature and it has the maximum amount of pipe-
bends in sewer pipe system in Japan.

• Junction: It is formed where three or more pipes intersect each other and makes
different curves in intersection surface, such as Y-Junction, T-Junction, and X-
Junction.

All pipe-bends are hand-made and constructed by concrete in spite that straight pipes
are made by PVC or ceramic and factory-made. Therefore, in general, there is no rule
or standard for curves and junctions in sewer pipe system in Japan, especially, in old
sewer pipe construction. According to the sewer pipe construction lows in Japan, there
is a manhole at the beginning of the pipe lines, between each 50 meters pipe and over
all pipe bends. A pipe joint is formed in the connection surface of two straight pipes
and an inlet is used for connecting a house pipe to a main pipe.

5.3.2 Basic Concepts of Design

In the current sewer pipe inspection technology, all commercial sewer inspection robots
have a poor mobility function to pass any kind of pipe-bends such as curves and
junctions so that those robots are only capable to move into the straight pipes. Same
as MAKRO case, there is a number of efforts and researches to develop multi-joint
(snake-like) robots with capability of passing curves and junctions [50, 51, 52]. Most of
these robots have a complex mechanism and several sensors to detect the pipe bends
and pass them. These complexities in mechanism and data processing makes not easy
to realize reliable commercial products.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 107

In this industrial research project we design the robot platform based on two fol-
lowing concepts:

• Passing the different kinds of pipe bends such as curves and junctions.

• Moving into the pipe network as same as water-flow direction.

we proposed the idea of “water-flow motion” to point out the fact that if the robot can
move into the sewer network as same as the direction of water-flow, the robot does not
required sensor to select its movement direction in case of “junctions”.

5.4 Novel Sewer Pipe Platform: naSIR Platform

In this section, we will describe, in detail, our approach to design and realize a sewer
pipe platform (naSIR platform) including a novel passive-active intelligent moving
mechanism (naSIR Mechanism). This novel platform can move into the straight pipe
and pass4 the different kinds of pipe bends without need to any intelligence of the
controller or sensor reading. With regard to developing a platform with capability of
passing/navigating the pipe-bends, same as MAKRO case, there is a number of efforts
and researches to develop multi-joint (snake-like) robots [50, 51, 52]. Most of these
robots have a complex mechanism and several sensors to detect the pipe bends and
pass them. These complexities in mechanism and data processing makes not easy to
realize reliable commercial products. In this direction, there is no concrete evidence
with regard to logical reasons or theorems that why the current commercial robots can
not pass any kinds of pipe-bend, whereas the robot mobility function has become a
critical subject for the speed up the sewer inspection process in point of perspective of
related sewer companies.

5.4.1 Drawbacks of Current Commercial Sewer Inspection Robots

Most commercial inspection robots are built similar a car with four, six or more wheels
(See Fig. 5.23). All wheels are fixed related to the robot and the robot has no steering
function or special mechanism to turn in the pipe-bends. In such kind of robot, the
robot can only perform a turning movement by changing speed in its right and left
wheels. In this section, we will describe in detail why even a small car-like robot is
not successfully able to pass a pipe-bend. The following results are the outcomes of
considerable amount of work in simulation (simulation model shown in Fig. 5.24_A)
and also driving a prototype robot in the real pipe network (see Fig. 5.24_B). In the
4
Here, again, we point out that using the phrase of “passing” for a platform, while it is moving through a
pipe-bend, instead of “Navigating”, is that, the naSIR platform moves into sewer networks same as water-flow
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 108

Figure 5.23: Three samples of current commercial sewer inspection robots (Q.I company -
Japan).

Figure 5.24: Figure (A) shows the simulation model and figure (B) illustrates a prototype of
small car-like robot with four wheels and no steering function (L 230 W 165 H 160 [mm]).
This robot has six infrared sensors for detecting the pipe-bends and four motor-encoder to
control the wheels speed.

first experiment, there are no sensors used on the robot for detecting the pipe-bends and
controlling the wheels speed. In the next experiment, we evaluated the performance
of the car-like prototype robot using six infrared sensors, mounted around the robot,
to detect the pipe-bends in order to control the wheels speed and move through the
different curve angles.

No sensor used for the robot motion

Figure 5.25 shows three different starting point for the robot during passing a curve
(case A), Y-Junction (case B) and T-Junction (case C). Also this figure illustrates the
desired directions of the robot movement (black thick arrows) while passing the three
different kind of pipe bends. The black thick line, shown in Fig. 5.25, can be presented
as an optimal path-line for the robot motion during driving through the pipe network.
The robot will be, successfully, pass the pipe-bends, if it can roughly keep moving on
this line.
direction (the sewer net structure are constructed, so that, water has only one direction to flow in any kinds
of pipe-bends of sewage networks).
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 109

Figure 5.25: It shows three different starting point for the small car-like robot during passing
a curve (case A), Y-Junction (case B) and T-Junction (case C). The desired directions of the
robot movement marked by the black thick arrows. The black thick line can be presented as
an optimal path-line for the robot motion.

While the robot is driving through a curve, it will tilt over into the pipe because
the contact position between the robot wheels and the pipe-wall changes relative to
the robot (Fig. 5.26). For more detailed understanding, let’s consider the front wheels
contact position. In the straight part of the pipe, as shown in robot position #1 in Fig.
5.27, all four robot wheels have a contact area on the optimal path-line (black thick
line). In the beginning of the curve (see robot position #2 in Fig. 5.27) the contact
area of the left and right front wheels change from the optimal path-line, respectively,
to the side and bottom of the pipe as marked by white circles in the Fig. 5.27. On
the other hand, because of the pipe shape and the rigidity of the robot, a high level
of friction exists between the robot wheels and the pipe walls. In this condition, if the

Figure 5.26: Robot motion, shown at four positions, while driving through a curve (A: 3D
view, B: Top view). The robot will tilt over into the pipe, because the contact position
between the robot wheels and the pipe-wall changes relative to the robot.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 110

Figure 5.27: In the straight part of the pipe, the robot position #1, all four robot wheels
have a contact area on the optimal path-line (black thick line). In the beginning of the curve
(robot position #2) the contact area of the left and right front wheels change from the optimal
path-line, respectively, to the side and bottom of the pipe as marked by white circles.

robot continues to move, the left wheels start to tilt upward and the right wheels tilt
downward (see Fig. 5.26) till the robot tilts over into the pipe. For the same reasons
as noted above:

• The constant change of the contact position between the robot wheels and the
pipe-wall relative to the robot

• The high level of friction that exists between the pipe-wall and the robot wheels

the small car-like robot has not enough mobility function to pass the junctions. The
results of simulation and real experiments of the robot motion in a Y-Junction and

Figure 5.28: Robot motion, shown at three positions, while it is driving through a Y-Junction
(A: 3D view, B: Top view). Because of specific surfaces formed in the intersection junction,
the robot starts to turn toward the direction (n), (robot #2), and in the robot position #3 it
is unable to find the proper direction to move forward beyond the junction.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 111

Figure 5.29: Robot motion, shown at three positions, while it is driving through a T-Junction
(A: 3D view, B: Top view). Because of specific surfaces formed in the intersection junction,
the robot starts to turn toward the direction (n), (robot #2), and in the robot position #3 it
is unable to find the proper direction to move forward beyond the junction.

T-junction are shown in Fig. 5.28 and 5.29, respectively. In both cases, because of
specific surfaces formed in the intersection junction, the robot starts to turn toward the
direction (n) (shown in robot position #2 in Figs. 5.28 and 5.29). At the intersection
junction, each front wheel has the tendency to move toward a different direction (see
the robot position #3 in Figs. 5.28 and 5.29). Therefore the robot is unable to find the
proper direction to move forward beyond the junction or in another expression, pass
the junction.

Six infrared sensors-four encoders used for the robot motion

In the second experiment, there are six infrared transducer sensors, which are mounted
at the robot chassis to cover most of its front and back and almost side area of the robot
(Fig. 5.24_B). They measure distances to objects up to 15 cm at a resolution of 2
mm. Then, we used four encoders, which are installed in the motor-gearhead-encoders
combination, measuring the robot wheel speed, independently. After a number of ex-
periments we realized that the installed sensors could use for detecting and controlling
the robot during driving it through the curve with specific angle but are unable to dis-
tinguish the different degrees of the pipe curvature. Therefore distinguishing different
degrees of curvature in a pipe network would require specific sensors with separate pro-
graming and algorithm which is not practically applicable to realize reliable commercial
product.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 112

5.4.2 Novel Moving Mechanism (naSIR Mechanism)

In this section we describe a special patented moving mechanism called “naSIR Mecha-
nism" [53]. naSIR is a passive-active intelligent, compact and novel moving mechanism
that can move into the straight pipes and passes different kinds of pipe bends without
need to any intelligence of the controller or sensor reading. This passive-active mech-
anism as a robot platform itself can move into the pipe and passes wide variety of the
pipe bends even without controller for the wheels actuator. In addition, this moving
mechanism has capability to pass the different size of pipes in diameter even from a
bigger diameter pipe to smaller diameter and also can pass a small obstacle and go
down a small step.
The design is based on the concept “passive adaptation of robot wheels to the
bends in the pipe". As mentioned in previous section, because of the constant change
of the contact position between the robot wheels and the pipe-wall relative to the robot
and the high level of friction that exists between the pipe-wall and the robot wheels,
current robots can not pass the pipe-curve without an exact speed control on their
wheels. This problem has been solved in new design by a proper wheels orientation
and passive damping of springs. The wheels orientation in new design was changed
90 degree compared with a car-like robot and each wheel has enough flexibility in
horizontal level. Therefore naSIR mechanism has four wheels parallel with horizontal
level which are connected to the four arms and each arm is jointed to the base plate,
independently, and they connected to each other by using four springs (see Fig. 5.30).
In naSIR mechanism, the contact positions between the wheels and pipe-wall relative
to the robot are steadily the same, therefore it can move through the different degrees
of the pipe curvature smoothly without any control and sensor reading (eg. a 90 degree
curve, shown in Fig. 5.31). Moreover, while the robot moves through the curve, there

Figure 5.30: naSIR mechanism has four wheels parallel with horizontal level which are con-
nected to the four arms and each arm is jointed to the base plate, independently, and they
connected to each other by using four springs.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 113

Figure 5.31: naSIR mechanism motion during passing a 90 degree curve (A: 3D view, B: Top
view). Robot can adapt itself into the pipe interior to have a continuous movement on the
optimum path-line.

Figure 5.32: naSIR mechanism motion during passing a Y-Junction (A: 3D view, B: Top
view). The white arrows indicate the direction of wheels rotation.

is a contact point on each of the four wheel circumference which at each moment in
time has exact similar condition relative to the robot so that it can be encircled by
the pair of optimal path-line, as shown in Fig. 5.31. Figures 5.32 and 5.33 illustrate
the robot in Y-junction and T-junction, respectively, and the white arrows indicate
the direction of wheels rotation. In both cases, because of specific surfaces formed in
the junction intersection, first the robot starts to turn toward the n direction (robot
position #2), immediately upon the contact of the right front-wheel with the optimal
path-line (robot position #3) the robot corrects its path and moves forward in the
m direction. Fig. 5.34 indicates the naSIR mechanism movement when it is moving
from a bigger diameter to smaller diameter pipe. The design of naSIR mechanism was
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 114

Figure 5.33: naSIR mechanism motion during passing a T-Junction (A: 3D view, B: Top
view). The white arrows indicate the direction of wheels rotation.

Figure 5.34: Motion of naSIR mechanism when it is moving from a big diameter to small
diameter pipe.

developed by changing the wheels orientation from 0 degree to 5 degree with respect to
the horizontal level (see Fig. 5.35). In this approach, the robot has a smoother motion
in T-junction and also it is capable to move on flat surfaces.

5.4.3 naSIR Platform

Calculation of proper dimensions of the naSIR mechanism for making a prototype

needs a precise study of dynamic and behavior of robot into the pipe. Because of the
complexity of friction condition and the contact area between the platform wheels and
pipe-wall, considering and driving of the platform dynamic equations is difficult and
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 115

Figure 5.35: Second version of naSIR mechanism. The robot wheels have a 5 degree
angle with respect to the horizontal level.

time-consuming. Using a powerful simulation such as Visual Nastaran 4, included of

a precise modeling of friction, can make an easier study of the robot motion and very
close to the actual condition which happen for mechanism in the real pipe. Based on
simulation results, for the pipe in range of 250-300 mm in diameter, a prototype with
following specification was made for employing naSIR mechanism as a platform in real
sewer pipe network, Fig. 5.36:

• Length of the prototype is changeable from 20 till 50Cm.

Figure 5.36: Prototype of naSIR mechanism, includes four motor-gearheads, four wheels and
a battery pack, is able to pass wide variety of the pipe bends in a sewer pipe network.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 116

• The wheels orientation is changeable in five different angles, -5, 0, 5, 10, 15 degree.

• Width of the prototype is changeable from 10 till 20 Cm.

• The spring stiffness coefficient is adjustable.

• Prototype has enough free space, for the extra weight, to get the different weight.

After several experiments of naSIR mechanism with different size, weight and spring
stiffness coefficient into the real pipe net, naSIR mechanism with the dimension of L400
W180 H60 [mm] and a spring coefficient 0f 0.5-1 [N/mm] had a smooth and robust
movement while passing the curves and junctions of a sewer network arranged by pipes
which have with diameter within range of 250-300 [mm]. The practical results of naSIR
mechanism movement with the different wheels angle with respect to the horizontal
level and the different wheels in shape and material, are illustrated in Tables 5.1 and
5.2, respectively. In addition, naSIR mechanism, because of its special morphology,
has an intelligence in selecting the correct direction of its movement same as water-
flow direction with no control or sensor reading. After all we developed the naSIR
mechanism as a platform called “naSIR Platform” (Fig. 5.37). naSIR platform consists
of naSIR mechanism, a lithium polymer battery pack fitted under the platform, and a
power switch can demonstrate a robust movement into the straight pipe and smooth
and rapid motion while passes different kinds of pipe bends. In this approach naSIR
platform does not request any kinds of sensor for its motion inside of the pipe, even
encoders for motors of wheels, and can move same as water-flow direction into the sewer
networks including the wide variate of pipe-bends . As an example, Fig. 5.38 shows
a sample of sewer network including a T-junction, a Y-junction, and six 90 degrees
curves. Red arrows, in this figure, show the direction of water-flow. The prototype of
naSIR mechanism and naSIR platform, started from point (A) or (B), are employed to

Table 5.1: The practical result of naSIR mechanism movement with the different wheels angle
with respect to the horizontal level into the pipe network.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 117

Table 5.2: The practical result of naSIR mechanism motion with the different wheels in shape
and material through the pipe network.

Figure 5.37: naSIR platform consists of naSIR mechanism, a lithium polymer battery pack
fitted under the platform (is not show in the figure) and a power switch.

move same as the red arrows direction into this network without any control or sensor
reading up to 2 hours without any stocking in the pipe-bends.

5.5 Autonomous Sewer Inspection Robot: KANTARO

KANTARO is designed as a fully autonomous, un-tethered robot, which fits to the
pipes within a diameter range of 200-300 millimeters (see Fig. 5.39). KANTARO’s
autonomy and its kinematic abilities extend its potential mission range enormously,
comparing to conventional inspection equipment that is limited by the cable and poor
kinematics. KANTARO carries all the necessary resources on-board. Standard lithium
polymer batteries provide the power for its 4 motors, the sensors, the light and elec-
tronics, including a developed computer system and an optical underground wireless
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 118

Figure 5.38: A sample of sewer network including a T-junction, a Y-junction, and six 90
degrees curves. Red arrows shows the direction of water-flow.

Figure 5.39: Developed KANTAROs fit to the 250 [mm] (left robot) and 200 [mm] (right
robot) diameter of sewer pipe.

communication module, allowing for an autonomous up time of about one hour.

5.5.1 KANTARO Architecture

To realize a reliable, robust robot and an easy maintenance system, KANTARO is

designed to have a complete modular architecture in its mechanic, hardware and soft-
ware. KANTARO, as shown in (Fig. 5.40), consists of two main modules: Bottom and
Upper box modules.
naSIR platform is employed as a bottom box of KANTARO. Electronic boards,
the sensors and light are installed in upper box which be connected via the main
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 119

Figure 5.40: KANTARO has a modular architecture consist of two main modules: Bottom
box and Upper box. Modularity of bottom and upper boxes showed in right and left of the
figure, respectively.

connector to the bottom box . In addition, KANTARO has IP67 waterproof standard
that it achieved by waterproof design of KANTARO’s modules including the upper
and bottom box, motor boxes and battery pack.

5.5.2 KANTARO Hardware

We designed all necessary electronic boards used for control, saving sensory data and
wireless communication as a module,called E.B. module, fit to the upper box of KAN-
TARO (Fig. 5.41). E.B. module contains a mother board, power and motor controller
and CPU boards, hard disk and an optical underground wireless communication board.
All boards are designed as separated modules to communicate to each other via their
own connectors through the mother board. In the newest version of E.B. module,
power board and hard disk are combined with mother board to be a more compact
and wiring less module. Wiring less was one of our target in this project to realize
a compact, reliable and robust hardware for KANATRO. To achieve the wiring less
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 120

Figure 5.41: The newest version of KANTARO electronic boards module (E.B. module).

Table 5.3: KANTARO electronics boards specification

electronics, KANTARO hardware is designed base on modular architecture. Table 5.3

illustrates the specification of KANTARO electronics boards.

5.5.3 KANTARO Sensors

Autonomous sewer robots must include sensors for their own control, navigation and
localization, not only those for sewer state assessment and damage detection. In ad-
dition, localization is an issue not only for the proper robot control, so that the robot
knows where it is, but also for inspection, as detected damages have to be reported
with their location. Therefore navigation and localization for the sewage inspection
robots, as a whole, can be classified in two main group: "motion-navigation" used for
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 121

control and robot motion and "damage-navigation" applied for the location of damages
in the pipe. The sensors used for motion navigation may overlap with the inspection
sensors (e.g. a camera may be used for both motion navigation and damage detec-
tion). Motion-navigation sensors, mostly, mounted in front of the robot that it makes

Figure 5.42: First (left side) and second (right side) version of KANTARO. Kantaro’s sensor
system consists a camera, a laser scanner, a tilt sensor mounted on the mother board and two
IR sensors.

complex design to avoid the overlap of their work spaces. KANTARO’s sensor system
includes an intelligent laser scanner, one fish eye camera, two IR sensors and a tilt
sensor. The first two sensors are only used for damage detection and navigation, be-
cause KANTARO, including naSIR mechanism, dose not request any kind of sensor for
motion-navigation. In this approach we had no constraint for mounting the sensors,
mentioned above, into the upper box of KANTARO. Using a tilt sensor mounted on
the mother board, and on-board control program, KANTARO is able to automatically
correct for tilt in its pose when driving inside a sewer pipe. To avoid driving the
robot in a step down more than 10 [cm] and protection from obstacle, two IR sensors
are installed in front of robot in two side of the camera. Figure 5.42 illustrates the
arrangement of KANTARO sensors. Rotating laser scanner has been developed as a
small and intelligent sensor installed in the robot rear to scan radially the inner side
of a pipe wall. If the KANTARO moves along the pipe while measuring, the laser
beam measures along a spiral on the pipe inside, where the resolution of the spiral
depends on the turn rate of the reflection and the speed of robot (Fig. 5.43). To re-
duce the data processing on the E.B. module, the developed laser scanner is improved
to detect the navigational landmark such as manholes, inlets and pipe joints used for
damage-navigation, independently, by using a powerful Microprocessor (SH-2) as well
as measuring distance and scanning angle, linearized and filtering and modeling data.
For realization of a 360 degrees infinitive rotation in laser scanner, we developed a
small and special magnet coupling to supply power to rotary part and transmit the
laser signal to the Microprocessor box.
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 122

Scanning directions 360 degrees

Scanning speed 0-1800 (rpm)
distance range 70-190 (mm)
Accuracy ±1(mm)
Beam radius 0.5 (mm)
Transmission method Full duplex serial transmission (2Mbps)
Sampling speed More than 10 KHZ
Measuring signal output Distance, Scanning angle
Wight 200 g
Size 37×48×166 (mm)
Power ±12V(0.5A) +5V(1A)

Figure 5.43: The newest laser scanner is capable to detect the navigational landmarks, inde-
pendently with the main computer, as a sensor module.

5.5.4 KANTARO as A Fully Autonomous Sewer Inspection System

In this section, first, we describe how KANTARO can be employed as a fully au-
tonomous robot for finding the pipe damages and fault detection. Second, we will
show our approach to make a safe and robust system in point of view of sewer inspec-
tion companies as users. KANTARO is capable to detect the candidate faulty images
(called feature-images) on-board, by using a fast and simple edge detection program
running on its E.B. module. Feature-images may include three different types of image:

• Landmarks (manholes, pipe joints and inlets)

• Faulty images (Fig. 5.45-left)

• Non-faulty images (eg. images including trashes)

When robot detects a feature-image, an autonomous control program will start to

decrease the robot speed for saving enough images, up to 5 frames, from feature-image
area. Saving the images, directly, is done by ADV202 image compression chip, installed
in CPU board, to the hard disk without involving CPU in saving process. At this time,
the laser scanner, that has been kept switch off during the normal pipe condition (no
feature-image detected by on-board program), will start to extract the possibility of
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 123

Figure 5.44: The sequence of on-board inspection process.

happened feature fault as a land mark. In case of detection of fault as a land mark, the
type of landmark, only as a code, will be saved via CAN interface into the hard disk.
These start and stop process of saving data from camera and switch on/off the laser
scanner perform a longer inspection process where we consider the capacity of hard disk
and energy consumption. In addition, the time, robot speed and encoder data from
four DC motors are saved to the hard disk during whole inspection process. Figure
5.44 shows the sequence of on-board inspection process. After the robot is retrieved, all
data saved in the robot hard disk will transmit via the LAN interface to the host PC.
In this PC, three different programs are applied to extract the real faulty images from
the retrieval feature-images and calculate the position of happened fault. First, the
feature-images are divided to the landmark images and the faulty-non faulty images
by applying the laser scanner data and the time to the images. In the next step, a
special fault detection patented algorithm extracts up to 60-70% of the faulty images
from the faulty-non faulty images (Fig. 5.45) [55]. Last, the rough position of the faults
will be calculated by using the encoder data, robot speed and time. This program has
possibility to accept the map data of the sewer network (position of manhole, inlets
and pipe joints) and fusion with laser scanner data to decrease the error of wheels slip
and more accurate calculation of the happened fault position. After of all, the remain
faulty images including the necessary specification will save as a result of inspection
process. Figure 5.46 describes the sequence of off-line inspection process.
To make a practical, safe and robust system, we introduced a life-optic-untethered
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 124

Figure 5.45: Nine different types of fault, possibly happening in the sewer pipe (left side).
An example of result of special fault detection patented algorithm applied on a faulty image
including a crack (right side).

Figure 5.46: The sequence of off-line inspection process.

cable for KANTARO. With this cable that act as a tail for robot, operator can connect
the robot to the control units located in out of the manhole via the fiber optic while
the robot is driving into the pipe. In this case, operator can observe the condition of
the pipe interior and also has possibility to control the robot and use the KANTARO
as a semi-autonomous robot for pipe state assessment. Also this cable can be used as
a life cable to pull out the robot in the case of stocking it inside of the pipe or power
failure. The length of the cable is designed 50 meters as the same length as maximum
distance of two manholes (see Fig. 5.47). In this approach, we could solve general
skepticism about using fully autonomous systems without a possibility to interfere the
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 125

Figure 5.47: KANTARO with the life-optic-untethered cable.

robot control at any time. In addition, the operator can send the basic commands such
as a move/stop the robot, turn on/off the light and save/not save data to the hard disk
and so on via the optical underground wireless communication module.

5.6 Experimental Results

We evaluated the performance of naSIR mechanism and our sewer inspection robot by
driving KANTARO in a sewer test field, in our RRI laboratory, which is made by PVC
material with a diameter ranging from 250 to 300 mm. Our test field, including all kind
of pipe bends, small steps and also navigational landmarks such as manhole, inlets and
pipe joints, can present as a sample of real world sewer network in Japan (Fig. 5.48).
naSIR platform (see Fig. 5.37) is able to move in this sewer network up to 4 hours
with one battery pack, charged in advance, without any control or stocking in the pipe
bends and steps. Also KANTARO, as a whole system, has been successfully employed
for autonomous damage navigation and fault detection up time about one hour (400
meters in advance). In addition, we had several chance to drive and test KANTARO
in real world sewage pipe network. Figure [cap:Real world test field] shows KANTARO
while passing a straight pipe with 250 millimeter in diameter. In this real test field
experiment, KANTARO could successfully pass 80 meters pipe network, including 3
manholes and one junction, and perform all task mention in previous section.

5.7 Conclusion
In this chapter we proposed an innovative, fast and robust sewer inspection method by
using a passive-active intelligent, fully autonomous, un-tethered robot,called "KAN-
TARO", which fits to the pipes within a diameter range of 200-300 millimeters. KAN-
TARO prototype robot, including a novel passive-active intelligent platform (naSIR
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 126

Figure 5.48: Our sewer test field including all kind of pipe bends at RRI laboratory. Two
down pictures showed KANTARO when passing a junctions (Left side) and a 90 degree curve
(Right side).

Figure 5.49: KANTARO in real world sewage pipe network while passing a manhole.

platform), has a robust movement into the straight pipe and smooth and rapid mo-
tion while passes different kinds of pipe bends without need of any intelligence of the
controller or sensor reading. In this approach KANTARO does not request any kind
of sensor for its motion inside of the pipe. In addition, we developed a small and
intelligent 2D laser scanner for detecting of the navigational landmarks such as man-
holes and pipe joints independently with main computer system and fusion with a fish
eye camera, mounted on the KANTARO, used for assessing the pipe state and fault
CHAPTER 5. SEWER PIPE INSPECTION ROBOT 127

detection. Realization of KANTARO as a fully autonomous, reliable and robust pipe

inspection robot has been achieved by designing an architecture based on intelligence
at its modules and definition and implementation of a life-optic-untethered cable to
make the inspection process easily and safely. To prove that KANTARO is able to use
as a commercial product, We still have to perform further experiments in real world
sewer pipe network with different in size and state pipe condition. Improving the offline
fault detection software to get high accuracy is on the future work.
Chapter 6
Conclusion
Chapter 6

Conclusion

In this thesis the concept of "intelligent mechanical design" is presented to show how
a mechanical structure can be designed to have an effect to the robot controllability,
simplification and its tasks performance. The description of this concept will lead
us to establish the landmarks of the territory of mechanical designing in the form of
seven design principles. The design principles, named “Mecha-telligence principles”,
provide guidance on how to design autonomous mobile robots mechanics. They incor-
porate the insights gained in this large design field in a compact and coherent form.
Macha-telligence principles guide us in asking the right questions when investigating
issues concerning a self-controllable, reliable, realizable, and compatible mechanics for
autonomous mobile robots (principle 2). To show how the Mcaha-telligence princi-
ples can be applied on the processes of the design of robots mechanics, we proposed
a novel methodology, "Mecha-telligence" methodology. Mechanical design in the pro-
posed methodology is done based on preference of classification of the robot specifica-
tion described by interaction of the robot with its environment and, also, the physical
parameters of the robot mechatronics itself (principle 1). In this approach the robot
specification is classified to a set of high-level- and low-level-specifications which can be
expressed as tasks (desired behavior) and physical parameters of the robot, respectively.
In fact the high-level-specification, including the two layers: main and sub layers, is
described by considering the robot tasks and its interaction with the environment in
which it is used. A low-level-specification may, consequently, consist of a set of layers
explaining the physical setup of the robot, its body, sensory, and motor systems.
A main goal in these analyzing process of the robot specification, is that, figure out
to a Mono-spec layer (the last layer of low-level-specification) including a simple basic
function or selecting a sensor for a single task or behavior. The mechanical design will
be started based on an important defined robot mono-spec, and it will be extended to
realize the other mono-specs by adding the minimum actuators and sensors to the de-

129
CHAPTER 6. CONCLUSION 130

sign (principle 6). In each step of the design process, also, we are trying to come up to
the proper solution by considering the robot morphology, design the suitable actuators,
applying the description of environmental and physical-morphological constraints, and
employing passive mechanisms which have inherent intelligence characteristics (princi-
ple 3 and 4) . In this methodology the mechanical design will be finalized by considering
the proper and sufficient positioning of sensory-motor system (principle 5) and concept
of modularity in robot mechatronics (principle 7).
To prove the validity of proposed method we concentrate on developing of the
mechanical structure and devices of three types of autonomous mobile robots used in
the three different educational, entertainment and industrial applications. In these
three projects we describe, not only, the mechanical design process based on "mecha-
telligence " principles and methodology, also, the detail design will consider to lead
readers to get familiar as much as possible to our vision related to the presented
design methodology. Here we, briefly, summarize the results of these three projects
which are realized by considering the Mecha-telligence theory in all design, simulation,
manufacturing, and assembly processes.
The first project was on developing an autonomous mobile soccer robot which has
fully mechatronics modular architecture including a strong ball-kicking device with
capability of lifting the ball and a ball-holding mechanism, aiming at getting champion
in RoboCup Midsize League. We developed a strong kicking device with capability
of shooting (up to 5 [m/s]), lifting (up to 120 [cm]), and holding the ball. One of
the specific features of the developed kicking device is that, the ball-kicking and ball-
holding are accomplished by only employing a simple DC motor-gearhead and a limit
switch. This kicking device, also, has an inherent design characteristic to lift the ball
by using the different shape of plates installed on the kicker.
The second challenge was on the design, modeling, simulation and implementation
of a series of entertainment robots named "Jumping Joe", exhibited in Aichi Expo-
sition 2005. Jumping Joe robots, artistic and agile robots, can perform several rapid
movements such as fast wake up, jumping and somersault. The key point for realiza-
tion such kinds of rapid movements, is developing two new types of actuators: Inertia
Actuators and Cam Charger. Inertia Actuator absorbs mechanical energy by increasing
its rotor angular velocity and delivers energy by decreasing its rotor velocity. Inertia
Actuator can generate a small internal torque by changing the speed of the rotor and a
big internal torque in short time by using a brake to stop the rotor at high speed. Cam
Charger including the same mechanism used as a kicking device for RoboCup robot, is
designed to fit to the foot shape with capability of. In this case we designed a unique
and compact cam including a releasing part to avoid the distribution of the force at
the moment of springs energy releasing to the foot structure.
CHAPTER 6. CONCLUSION 131

In third project we propose an innovative, fast and robust sewer inspection method
by using a passive-active intelligent, fully autonomous, un-tethered robot called KAN-
TARO which fits to the pipes within a diameter range of 200-300 millimeters. KAN-
TARO prototype robot, including a novel passive-active intelligent moving mechanism,
can move into the straight pipe and pass different kinds of pipe bends without need
to any intelligence of the controller or sensor reading. The proposed method, design
the morphological mechanism and sensor less robot will achieved completely based on
"mecha-telligence principles and methodology.
Bibliography

[1] K. Goris, Autonomous Mobile Robot: Mechanical Design, Vrije University Brussel, Brus-
sels, Belgium, 2004-2005

[2] Article of “AI - Classwork Assignment 1”, c Philip Iezzi / s997143, 2001.

[3] R. Pfeifer and C. Scheier, Understanding Intelligence, Massachusetts Institute of Tech-

nology, 1999.

[4] Y. Wu and C. Van, Proc. Of the Southern Association of Information Systems Conference,
pp: 289-294, 2005.

[5] RoboCup homepage, http://www.robocup.org/

[6] H.Matsubara, M.Asada, et al, “The RoboCup: The Robot World Cup Initiative”, In
Proceedings of IJCAI-95.

[7] T.Ichinose, Y.Takemura, K.Azeura, I.Godler, “HIBIKINO- MUSA SHI”, RoboCup 2006
Bremen, CD-ROM Proc. of RoboCup 2006.

[8] A.T.A Peijnenburg, et. al, “Philips CFT RoboCup Team Description” in preliminary
proceedings 2002 RoboCup conference, July 2002.

[9] F. Riberio, et. al, “New improvements of MINHO Team for RoboCup Middle Size League
in 2003”, IN Proceedings of CD-ROM, RoboCup 2003.

[10] T.Buchhei, et. al., “Team Description Paper 2005 CoPS Stuttgart”, RoboCup 2005 Osaka,
CD-ROM Proc. of RoboCup 2005.

[11] K. Matsuoka, “A mechanical model of repetitive hopping movements,” Biomechanisms,

Vol.5, pp. 251-258, 1980.

[12] M. H. Raibert, H. B. Brown, Jr., and M. Chepponis, “Experiments in balance with a 3D

one-legged hopping machine,” Int. J. Robot. Res., vol. 3, no. 2, pp. 75-92,1984.M.

[13] M. H. Raibert, “Legged Robots that balance," MIT Press, Cambridge, 1986.

132
BIBLIOGRAPHY 133

[14] P. Fiorini and J. Burdisk, “The development of hopping capabilities for smell robots,"
Auton. Robots, vol.14, no.2-3, pp.239-254, Mar./May 2003.

[15] Sandia National Lab., 2000, Hop to it: Sandia Hoppers Leapfrog Conventional Wisdom
about Robot Mobility [Online]. Available: http://www.sandia.gov/Labnews/LN10-20-
00/hop_story.html

[16] P. E. Rybski, N. P. Papanikolopoulos, S. A. Stoeter, D. G. Krants, K. B. Yesin, M. Gini,

R. Voyles, D. F. Hougen, B. Nelson, and M. D. Erikson, “Enlisting Rangers and Scouts
for recconnaissance and surveillance,” IEEE Robot. Autom. Mag., vol. 7, no. 4, pp. 14-24,
Dec. 2000.

[17] S. A. Stoeter and N. P. Papanikolopoulos, “Kinematic Motion Model for Jumping Scout
Robots," IEEE Transaction on robotics, vol.22, no.2, pp. 398-403, 2006.

[18] N.M. Mayer, RS. Guerra, M. Ogino, M. Asada, "Stabilizing a Biped Robot by Using a
Symmetric Rotor, CD-ROM Proc. of RoboMec2005, 1P2-S-045, 2005.

[19] N. Michael. Mayer, K. Masui, M. Browne, M. Asada, “Gyro Stabilized Biped Walking,"
Proc. of IROS, 2006.

[20] Joseph E. Shigley, Charles R. Mischke, Mechanical Engineering Design, McGraw-Hill

Higher Education, 2001.

[21] M. Uchiyama, “Mechanisms and Characteristics of Parallel Manipulators," Journal of the

robotics society of Japan, vol. 10, no. 6, pp. 715-720, 1992.

[22] Y. Ota, Y. Inagaki, K. Yoneda and S. Hirose, “Research on a Six-Legged Walking Robot
with Parallel Mechanism," Proc. of IROS 98, pp. 241-248, 1998.

[23] Y. Sugahara, T. Endo, H. Lim and A. Takanishi, “Design of a Battery-powered Multi-

purpose Bipedal Locomotor with Parallel Mechanism," Proceedings of IROS’02, pp. 2658-
2663, 2002.

[24] S. Tadokoro, “Control of Parallel Mechanism," Journal of the robotics society of Japan,
vol. 10, no. 6, pp. 721-726, 1992.

[25] K. Monden, “Machine Engineering," Morikita, pp. 10 - 13.

[26] “POWER SPRING CHARGER MECHANISM," Inventor: A. A. F. Nassiraei, Japan

Patent No. 2005-315994, Oct. 2005.

[27] K.Shimonomura, S.Kameda, K.Ishii and T.Yagi, “A Novel Robot Vision Employing a
Silicon Retina," Journal of Robotics and Mechatoronics, Vol.13 No.6, pp. 614 - 620,
2001.
BIBLIOGRAPHY 134

[28] Material provided by Steinbeis Japan Inc., Kitakyushu foundation for the Advancement
on Industry, Science and Technology, March 2002.

[29] K.P. Bolke:Kanalinspekiton. Schaden erkemen und dokumentieren. Berlin (Spring), 1996.

[30] F. Kirchner and J. Hertzberg, “A prototype study of an autonomous robot platform

forsewerage system maintenance,” Autonomous Robots 4(4), pp. 319-331, 1997.

[31] J. Hertzberg and F. Kirchner, “Landmark-based autonomous navigation in sewer-

agepipes,” in Proc. First Euromicro Workshop on Advanced Mobile Robots (EU-
ROBOT96). 1996, pp. 6873. Kaiserslautern, IEEE Press.

[32] J. Hertzberg F. Schherr and W. Burgard, “Probabilistic mapping of unexpected objects

bya mobile robot,” in Proc. of the 1999 IEEE/RSJ International Conference on Intelli-
gentRobots and Systems (IROS 99), vol. 1. 1999, pp. 474-481, IEEE Press, Piscataway,
NJ, -ISBN 0-7803-5184-3.

[33] E. Rome, J. Hertzberg, F. Kirchner, U. Licht, H. Streich, and Th. Christaller, “Towards
autonomous sewer robots,” The MAKRO Project. J. Urban Water, 1, pp. 57-70, 1999.

[34] K.-U. Scholl V. Kepplin and K. Berns, “A mechatronic concept for a sewer inspection
robot,” in Proc. IEEE/ASME International Conference on Advanced Intelligent Mecha-
tronics (AIM 99). 1999, pp. 724-729, IEEE Press, Piscataway, NJ.

[35] E. Rome, H. Surmann, H. Streich, U. Licht, and K.-L. Paap, “A custom IR scanner for
landmark detection with the autonomous sewer robot MAKRO,” in Proc. 9th Interna-
tional Symposium on Intelligent Robotic Systems (SIRS 2001). Devy, M. and Lerasle,
F. (eds.), LAAS-CNRS, Toulouse, July 18-20, 2001, pp. 457-466, ISBN 2-907801-01-5,
Toulouse, France.

[36] E. Rome L. Paletta and A. Pinz, “Visual object detection for autonomous sewer robots,”
in Proc. 1999 IEEE/RSJ International Conference on Intelligent Robots and Systems
(IROS 99), vol. 2. 1999, pp. 10871093, IEEE Press, Piscataway, NJ, ISBN 0-7803-5184-3.

[37] M. Kolesnik and H. Streich, “Visual orientation and motion control of makroadaptation
to the sewer environment,” in To be published in Proceedings of SAB2002 - Simulation
of Adaptive Behavior. Edinburgh, UK, August 4-9, 2002.

[38] K. Berns K.-U. Scholl, V. Kepplin and R. Dillmann, “Controlling a multijoint robot
for autonomous sewer inspection,” in Proc. 2000 IEEE/RAS International Conference
on Robotics and Automation (ICRA 2000). IEEE/RAS, Piscataway (NJ), April 24-28,
2000, ISBN 0-7803-5886-4. San Francisco.
BIBLIOGRAPHY 135

[39] H.-B. Kuntze and H. Haffner, “Experiences with the development of a robot for smart
multisensoric pipe inspection,” in Proc. 1998 IEEE International Conference on Robotic-
sand Automation (ICRA 98). May 1998, pp. 1773-1778, EEE Press, ISBN O-7803-4300-
X-5198, Leuven, Belgium.

[40] H. Haffner H.-B. Kuntze, D. Schmidt and M. Loh, “Karo - a flexible robot for smart
sensor-based sewer inspection,” in Proc. 12th International No-Dig conference, 1995, pp.
367374, Hamburg, Messe und Congress GmbH.

[41] H.B. Kuntze, H. Haffner, M.Selig, D.Schmidt, K. Janotta, and M. Loh, “Entwick-
lung einesflexibel einsetzbaren roboters zur intelligenten sensorbasierten kanalinspektion
(karo),” Dokumentation 4. Internationaler Kongre- Leitungsbau, pp. 513-528, 1994.

[42] R. Kerker H.B. Kuntze and U. Hirsch, “Zur lageregelung eines mobilen kanalroboters,”
Autonome Mobile Systeme, vol. 9, pp. 369-379, 1993.

[43] M. Hartrumpf R. Munser, H.-B. Kuntze and C.W. Frey, “Ein modulares
multisensorsystemf- rohrinspektions- und rohrsanierungsroboter,” proc. 16, in Fachgespr-
h Autonome MobileSysteme (AMS 2000), 2000, Karlsruhe, Germany.

[44] R. Bradbeer, S. Harrold, B.L. Luk, B. Li, L.F. Yeung, and H.W. Ho, “A mobile robot for
inspection of liquid filled pipes,” Workshop on Service Automation and Robotics, June
2000, City University of Hong Kong.

[45] R. Bradbeer, “The Pearl Rover underwater inspection robot,” Mechatronics and Machine
Vision, pp. 255-262, 2000.

[46] R. Bradbeer, F. Nickols, and S.O. Harrold, “An ultrasonically controlled autonomous
model submarine operating in a pipe environment,” in Proc. Mechatronics and Machine
Vision in Practic e (M2VIP 97). 1997, IEEE Press, ISBN 0-8186-8025-3, Toowoomba,
Australia, September 23-25, 1997.

[47] S.O. Harrold Y. Li and L.F. Yeung, “Experimental study on ultrasonic signal transmission
within the water-filled pipes,” in Proc. 4th Annual Conference on Mechatronics and
Machine Vision in Practice (M2VIP 97). September 23-25, 1997, IEEE Press, ISBN 0-
8186-8025-3, Toowoomba, Australia.

[48] K. Rogers G. Campbell and J.Gilbert, “PIRAT - A system for quantitative sewer assess-
ment,” in Proceedings of the 12th International No-Dig Conference (No-Dig 95), 1995,
pp. 455-462, Hamburg Messe und Congress GmbH.

[49] [29] K.J. Rogers, R. Kirkham, P.D. Kearney, “PIRAT A System for Quantitative Sewer
Assessment,” in Proc. Int. Conf. Field and Service Robotics (FSR 99), 1999, pp. 7-12,
CMU,vPittsburgh (PA).
BIBLIOGRAPHY 136

[50] S. Fujiwara, R. Kanehara, T. Okada, and T. Sanemori, “An articulated multi-vehicle

robot for inspection and testing of pipeline interiors," In Proceedings of the IEEE/RSJ
International Conference of Intelligent Robots and Systems, pp. 509-516, 1993.

[51] E.F. Fukushima, and S Hirose, “How to steer the long articulated body mobile robot
kr-ii," In International Conference on advanced robotics, 1995.

[52] St. Cordes, K. Berns, M. Eberl, W. Ilg, and R. Suna. “Autonomous sewer inspection with
a wheeled multi-articulated robot," Robotics and Autonomous Systems, pp. 21:123-135,
1997.

[53] “Mechanism of movement inside of the pipes and mechanism of pipe inspection," Inventor:
Amir A. F. Nassiraei, Japan Patent No. 2005-17384, January 2005.

[55] “A device and method for fault detection in sewer pipe system ," Inventor: Alireza Ahrary,
Japan Patent No. 2005-209677, July 2005.