Citation preview
TS5-4 Mobile Haptic Interface for Large Immersive Virtual
Environments: PoMHI v0.5 The 4th International Conference on
Ubiquitous Robots and Ambient Intelligence (URAI 2007)
Mobile Haptic Interface for Large Immersive Virtual Environments:
PoMHI v0.5
1Chaehyun Lee, 2Min Sik Hong, 1In Lee, 2Oh Kyu Choi, 2Kyung-Lyong
Han, 2Yoo Yeon Kim,
1Seungmoon Choi, and 2Jin S. Lee 1: Haptics and Virtual Reality
Laboratory
POSTECH, Republic of Korea
2: Robotics and Automation Laboratory POSTECH, Republic of
Korea
{xlos, mshong, inism, hyh1004, sidabari, yooyeoni, choism,
jsoo}@postech.ac.kr
Abstract - We present initial results of research toward a novel
Mobile Haptic Interface (MHI) that provides an unlimited haptic
workspace in large immersive virtual environments. When a user
explores a large virtual envi- ronment, the MHI can sense the
configuration of the user, move itself to an appropriate
configuration, and provide force feedback for haptic display,
thereby enabling a vir- tually limitless workspace. Our MHI (PoMHI
v0.5) is featured with omni-directional mobility, a collision-free
motion planning algorithm, and force feedback for general
environment models. We also provide experimental re- sults that
show the fidelity of our mobile haptic interface. Keywords - mobile
haptic interface, mobile robot, omni-directional wheel, PID control
with fuzzy logic control, haptic rendering
1. Introduction
A force-feedback haptic interface has an inherent limit on its
workspaces and cannot render virtual objects larger than its
workspace. This is tolerable in desktop applica- tions where the
size of a virtual environment is comparable to the workspace of a
desktop haptic interface, but can be a severe limitation in large
virtual environments such as the CAVETM. The traditional solution
to this problem has been using a large haptic interface of a
manipulator type [1] or of a string type (SPIDAR) [2][3]. However,
their work- spaces are still limited and cannot be easily scaled to
vir- tual environments of different sizes.
A promising alternative is a Mobile Haptic Interface (MHI) that
refers to a force-feedback haptic interface with a mobile base. The
mobile base can move the haptic in- terface to an adequate place to
render large virtual objects and thus can provide an unlimited
workspace. Compared to other large haptic interfaces, the MHI is
easier to be installed or removed due to its mobility, considerably
smaller, and safer. Furthermore, the MHI can provide high-fidelity
force feedback since a desktop haptic inter- face with greater
precision is used for it.
The concept of the MHI was firstly proposed by Nitzsche et al
[4][5]. Although their MHI, Walkii, had omni-directional mobility,
a simple motion planning al- gorithm was used and only primitive
objects could be rendered. Barbagli et al. presented two new MHIs
in 2004 [6]. Both of the two MHIs used a commercial robot as its
mobile base and adopted more complex motion planning
algorithm than Walkii. More recently, a MHI with two desktop haptic
interfaces was proposed for two-point ma- nipulation [7].
In this paper, we present an initial version of POSTECH Mobile
Haptic Interface (PoMHI v0.5) especially de- signed to be used with
large visual environments such as the CAVETM. The PoMHI can move in
any direction using three omni-directional wheels while avoiding
collisions with a user, load general virtual environment models,
and provide acceptable force feedback. The mobile base with the
three omni-directional wheels was controlled via a PID-control with
an additional fuzzy logic control. A mo- tion planning algorithm
guiding the movement of the mobile base was also developed. We also
confirmed through experiments that the dynamics of the mobile base
brings little interference on the forces delivered to the
user.
The reminder of this paper is organized as follows. Section 2
briefly reviews the overall architecture of the PoMHI. Section 3
describes the hardware components and the control methods used in
the system. In Section 4, we describe the software structure and
the motion planning strategy. The effect of the mobile base
dynamics on the final rendering force is also examined.
Applications of the PoMHI are discussed in Section 5. Finally, we
conclude this paper in Section 6.
2. System Architecture
The overall system architecture is shown in Figure 1. The IS-900
Tracking System (a processor and two track- ers; InterSense Inc.,
USA) is responsible for tracking the current configurations
(position and orientation) of both the PoMHI and the user. Tracked
information is sent to the
Fig. 1. Architecture of the PoMHI.
106
The 4th International Conference on Ubiquitous Robots and Ambient
Intelligence (URAI 2007)
tracker server, and the server forwards it to the laptop inside the
PoMHI via UDP protocol. Based on the user’s configuration, the
laptop determines an appropriate con- figuration of the mobile base
and controls the base to place it to the configuration. Graphic and
haptic rendering modules are also managed in the laptop. The user
wears a head mounted display (HMD) and sees stereoscopic scenes of
a virtual environment. Relevant communication and rendering rates
are specified in the figure.
3. MHI Hardware
3.1 MHI Hardware Design A. Overall Structure of the Hardware
The hardware structure of the PoMHI is shown in Figure 2. It
consists of four main parts: omni-directional wheels and geared DC
motors, a DSP control board and power amplifiers, a laptop, and a
desktop 3 DoF haptic interface (PHANToM Premium 1.5A; SensAble
inc., USA). The mobile base is responsible for moving the whole
PoMHI to face the user, and the PHANToM is responsible for
providing appropriate force feedback to the user. B.
Omni-directional Structure
Figure 3 shows the mobile-base actuation design using three
omni-directional wheels. Our design follows the Y-shaped structure
for holonomic motion of the mobile base. This is more adequate than
a bidirectional mobile robot to follow possibly abrupt motions of a
user. Each wheel is connected to a motor by a timing belt and a
pulley.
The kinematics of the mobile base is derived from the relation of
parameters represented in Figure 4. Here, the
origin of the local coordinate is set to the center of the base and
the parameters are defined as:
iv : Linear velocity of each wheel (i = 1, 2, 3)
[ , ]Tx y=v : Linear velocity of the mobile base
φ : Angular velocity of the mobile base L : Distance between a
wheel and the orgin.
Then, the mobile base kinematics can be derived as
follows [8][9]:
, (1)
where T TTS x y φ φ = = v .
3.2 MHI Control
A. Mobile Base Control Overview
The motion of the mobile base is based on PID control for the
desired linear and angular velocities of the base. As shown in
Figure 3, each omni-directional wheel has six free-rolling
sub-wheels that may cause a slip. Moreover, it was observed that
the mobile base contains inherent nonlinearity induced from the
wheel and belt-pulley structure. We thus also use a fuzzy logic
control (FLC) and other supplementary control techniques to
overcome the problem. The overall control equation for motor i
is:
( 1) ( ) ( ) ( )i i i iU k U k U k k+ = + +Γ , (2) where Ui(k) is a
command to be sent to the motor amplifier and
i(k) is an output of the supplementary control. The
sampling rate of the control loop is set to 50 Hz. B. PID-based
Velocity Control
Once a desired trajectory of the mobile base is determined from the
motion planning algorithm (will be explained later), the desired
velocity of each wheel is calculated by the kinematics in Eq. (1).
Using these
Fig. 2. Hardware structure of the PoMHI.
Fig. 3. An omni-directional wheel (left) and the placement of three
wheels in the mobile base (right).
Fig. 4. Omni-directional structure on local coordinates.
107
The 4th International Conference on Ubiquitous Robots and Ambient
Intelligence (URAI 2007)
velocities, each motor is controlled by the PID control method. The
velocity of each wheel is estimated using the exponential filter
as:
, ,( ) (1 ) ( 1) ( )i filtered i filtered iv k v k v kα α= − − + ,
(3) where 0 1α≤ ≤ is the parameter of the filter.
To obtain the desired motor velocity, we use the PID control
equation as follows:
1
i p i i i d i i
U k K e k K e i K e k =
+ = + + ∑ , (4)
where , ,( ) ( ) ( )i i desired i filterede k v k v k= − is the
velocity error
of the motor, and Kp, Ki, and Kd are the PID gains. Using only the
PID control did not bring enough per-
formance in our system. This is illustrated in Figure 5 where the
red lines represent references and blue lines measured values. The
left graph shows the position of the PoMHI, and the right three
graphs show the velocities of the wheels. We can observe the system
nonlinearity in the black circles where the measured velocity
rather decreases in spite of the increase of the reference
velocity. The nonlinearity seems to be caused by the increase of
friction when the contact point of a wheel between a sub-wheel and
the floor is changed. Because this nonlinearity is in- herent
characteristics of the wheel structure, eliminating this phenomenon
perfectly may not be possible. We thus minimize the effect of the
nonlinearity through additional control methods that will be
described in the next subsec- tions.
C. Fuzzy Logic Control
We use a singleton fuzzifier, a product inference engine, and a
center average defuzzifier in FLC of the mobile base. The inputs
are the velocity error ei(k) and its increment
ei(k). We use triangular membership functions (eµ and
eµ ) as shown in Figure 6. Membership functions consist of five
sections; negative big (NB), negative (N), zero (ZO), positive (P),
and positive big (PB). These sections are determined
experimentally.
Table 1 shows the rule set of the FLC derived from the input
membership functions. Each element of the table,
( , )i jy , represents the output of each rule. Using the rules and
input variables, the updating output of FLC is [3]:
5 3 ( , )
m n i
k k
µ µ
µ µ
We adopt additional supplementary control when the following
condition is satisfied:
, ,( ) ( 1) & ( ) ( 1)i i i filtered i filteredU k U k v k v
k> − < − . (6) This condition detects when the contact point
of a wheel between a sub-wheel and the floor is changed. While this
occurs, excessive errors are accumulated in the PID con- trol which
causes an overshoot in position control. An algorithm to minimize
this behavior is described below.
If Eq. (6) is satisfied at time index k, the controller stores the
sum of updating outputs
Ui(k) and the current velocity
( )e filteredv v k= . In each control time l after k, the
controller checks whether a condition, ( )e filteredv v l< , is
met. If it is, the controller compensates the excessive error
by:
, ( ) l
U U k =
= ∑ and (7)
, ( ) , ( )
k k l
Fig. 5. Experimental results using the PID-based velocity
control only (x = 0.0 m/s, y = 0.2 m/s, φ = 0.0 rad/s).
Fig. 6. Membership functions of velocity error, eµ , and
its variation, eµ .
e(k) e(k) NB N ZO P NP
N (1,1)y (2,1)y (3,1)y (4,1)y (5,1)y
ZO (1,2)y (2,2)y (3,2)y (4,2)y (5,2)y
P (1,3)y (2,3)y (3,3)y (4,3)y (5,3)y
108
The 4th International Conference on Ubiquitous Robots and Ambient
Intelligence (URAI 2007)
E. Experimental Results
Figure 7 presents the results with the adapted FLC and the
supplementary control. In the velocity graph of each wheel, fine
trembles due to the change of contact points still show up, but we
can see that the effect of the inherent nonlinearity has
significantly decreased. Moreover, in the position graph, measured
position values almost perfectly followed the reference values (a
straight line).
4. MHI Software
The software in the MHI system consists of three parts.
The tracker server program which is executed in the tracker server
receives the configuration of each IS900 tracker and sends this
information to the MHI via wireless LAN. Using this information,
the software running on the laptop in the MHI controls all devices
in the MHI (except for the motors controlled by the DSP board) and
renders visual and haptic information.
The tracker server program and the MHI communicate through wireless
LAN. Due to the update rate of the IS900 tracker, the network
update rate is set to 190 Hz. To maintain this speed, the UDP
protocol is used instead of the TCP protocol. TCP, which is more
reliable than UDP, is sometimes too slow and even exhibits severe
jitters. Whenever a packet is missed in TCP, the protocol re-
transmits all packets after the missed packet and causes a long
delay which is unacceptable in our application. The UDP is not as
complex as the TCP, and is therefore faster. Although some packets
can be missed and reordered dur- ing transmission in UDP, its
effect can be made trans- parent to a user.
The haptic server program is comprised of networking, motion
planning and haptic rendering parts. First, the networking part
receives the tracker information from the tracker server and
transforms the coordinate system from IS-900 to the world
coordinate frame. This updates the configuration of the user, the
mobile base and the haptic interface point (HIP; a point modeling
the haptic tool tip) and sends the new information to the visual
server through network. Note that since the haptic and visual
server pro- grams communicate with each other via network, the two
servers can operate on separate machines for purposes such as
higher performance or improved convenience. Second, the motion
planning part calculates the next proper configuration of the
mobile base and commands angular and linear velocities to the
mobile base. Finally, the haptic rendering part detects a collision
between the HIP and virtual objects and calculates haptic feedback
force. All of these parts run at different rates, so the haptic
server is designed as a multithreaded program.
The other server program which also runs on the laptop is the
visual server. The visual server receives the con- figuration
information from the haptic server and renders visual scenes. In
our current system, a HMD is used for visual rendering, so the
visual server also runs on the laptop with the haptic server. If
other visual systems are used (e.g. CAVETM), the visual sever can
be easily moved to another machine that controls the visual display
and communicate with the haptic server through network, as is done
in [11].
Fig. 7. Experimental results with the modified velocity
control (x = 0.0 m/s, y = 0.2 m/s, φ = 0.0 rad/s).
Fig. 8. Software structure of PoMHI.
Fig. 9. Configuration space of the MHI.
Fig. 10. Motion planning algorithm.
-1 -0.5 0 0.5 1
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
y
x
0 1 2 3 4 5 6 7 8 9 10 -0.02
0
0.02
0.04
V 1
0 1 2 3 4 5 6 7 8 9 10 0
0.2
0.4
V 2
0 1 2 3 4 5 6 7 8 9 10 -0.4
-0.2
0
The 4th International Conference on Ubiquitous Robots and Ambient
Intelligence (URAI 2007)
4.2 Motion Planning Algorithm
Basically, the mobile base has to place the haptic device in a
proper position where the device can give force feedback to a user
most effectively while avoiding colli- sions with a user.
Considering these constraints, we set the goal position of the MHI
on a line which passes the user and the user’s hand grasping the
haptic tool. The distance between the user and the MHI is
maintained within the typical arm length of an adult. Moreover, the
MHI always faces the user so that the haptic interface tool is
positioned in front of the user.
To represent the motion of the MHI, we introduce a 3D configuration
space where its x and z axes represent the 2D position of the
mobile base and its y axis the direction of the MHI in the
workspace. In the configuration space, a configuration of the MHI
is represented as a point, and a user is represented as a
cylindrical obstacle (see Figure 9).
To move the MHI to a goal configuration without col- liding with
the user, we developed the following motion planning strategy
consisting of three cases. First, when the MHI is too close to the
user, it moves away from the user as soon as possible to avoid
possible collisions with the user. Second, when the goal position
is close to the current position, the MHI moves toward the goal.
Finally, when the goal position cannot be reached directly (e.g.,
when the user is on the path), the MHI sets sub-goals between the
goal and current position of the MHI and moves towards the nearest
sub-goal. An implemented algorithm for this strategy is shown in
Figure 10.
To evaluate the performance of the motion planning algorithm, we
simulated the motion of the mobile base and the user and
represented the results with graphs in Figures 11 and 12. To
simplify the simulation process, we as- sumed the mobile base could
always move with maximum velocity. In each graph, the red point
represents the current position of the MHI and the blue region
represents the next position of the user where the mobile base can
catch up with the user in 1 second. The results indicate that the
change of maximum angular velocity does not signifi- cantly matter
to the MHI while the change of the maxi- mum linear velocity
does.
4.3 Force Rendering Algorithm
To calculate feedback force for a user given the HIP po-
sition and a virtual environment model, we use the virtual proxy
algorithm which is widely used for haptic rendering with static
objects. In a MHI, the movement of a mobile base can cause unwanted
forces that may be perceived by the user. However, if the mobile
base is tightly posi- tion-controlled and significantly heavier
than a haptic in- terface on top of it, we can ignore the effect of
mobile base dynamics. To verify this fact, we conducted an
experiment with a force sensor attached between the lank link of
the PHANToM and the puck held by the user.
The results of the experiment are shown in Figures 13 and 14 where
the red lines represent commanded force to the haptic device and
the blue lines recorded force by the force sensor. According to the
results, the differences between the commanded and recorded force
in the case of the moving mobile base are not greater than those in
the case of the stationary mobile base. This means that the mobile
base dynamics does not significantly affect the feedback force
delivered to the user. Therefore, we calculate torque commands sent
to the haptic interface only considering the static torque-force
relationship of the desktop haptic interface.
Fig. 11. Regions that the mobile base can move in 1 second with
different maximum angular velocities.
Fig. 12. Regions that the mobile base can move in 1 second with
different maximum angular velocities.
Fig. 13. Comparison between commanded and meas-
ured forces when the mobile base was stationary. Fig. 14.
Comparison between commanded and meas-
ured forces when the mobile base was moving.
110
The 4th International Conference on Ubiquitous Robots and Ambient
Intelligence (URAI 2007)
4.4 Virtual Environment Model
Not only primitive models such as a plane and a cyl-
inder but also complex triangular mesh models can be loaded and
rendered in the PoMHI. The visual server also shows the current
position of the HIP to let the user know where s/he interacts with
and the wall for a warning about the physical workspace limit in a
room. The model of the mobile base is also loaded and visually
rendered via a HMD in order to help the user avoid colliding with
the mobile base for the safety purpose.
5. Applications
The current MHI system can provide the user with
visual and haptic feedback about large virtual environ- ments. One
example is shown in Figure 15. A complex dolphin model consisting
of 4488 faces is loaded, and the user sees and touches the
real-sized dolphin model (1258 mm wide, 426 mm high, and 352 mm
deep). For the sta- bility of the mobile base, we set the maximum
linear and angular velocity to 0.2 m/s and 40 degree/s,
respectively. With this maximum velocity values, the MHI can follow
the user in most cases, unless the user moves abruptly.
6. Conclusions and Future Work
We have develop an initial version of new mobile haptic interface
named PoMHI and its applications. The PoMHI can place itself to an
appropriate configuration to provide boundaryless haptic feedback
while avoiding collisions with the user, and handle general virtual
environment models. We adopted several motion control methods to
stably and correctly control the motion of the mobile base. We also
examined the fidelity of force display in the PoMHI by comparing
actual force outputs with desired values.
We are currently working on a next version of the PoMHI. This
version has four omni-directional wheels with advanced design and a
lift for the desktop haptic interface for the extension of its
workspace in the height direction. We are also upgrading the
software in terms of more sophisticated motion planning algorithm,
more precise kinematics calibration, and force computation
algorithm considering the effect of the mobile base dynamics. Once
all of these are completed, we will integrate the PoMHI into the
CAVETM that is the most
immersive large virtual environment platform among the
present.
Acknowledgements
This work was supported by a research grant No.
R01-2006-000-10808-0 from the Korea Science and En- gineering
Foundation (KOSEF) funded by the Korea government (MOST).
References
[1] F. P. Brooks, M. Ouh-Young, J. J. Batter, and P. J.
Kilpatrick, “Project GROPE - Haptic Displays for Scientific
Visualization,” In Proc. SIGGRAPH 90, pp. 177-185, 1990.
[2] L. Buoguila, M. Ishii, and M. Sato, “Multi-Modal Haptic Device
for Large-Scale Virtual Environ- ment,” In Proc. 8th ACM
International Conference on Multimedia, pp. 277-283, 2000.
[3] N. Hashimoto, S. Jeong, Y. Takeyama, and M. Sato, “Immersive
Multi-Projector Display on Hybrid Screens with Human-Scale Haptic
and Locomotion Interfaces,” In Proc. International Conference on
Cyberworlds, pp. 361-368, 2004.
[4] N. Nitzsche, U. D. Hanebeck, and G. Schmidt, “Mobile haptic
interaction with extended real or virtual environments,” In Proc.
RO-MAN 2001, pp. 313-318, 2001.
[5] N. Nitzsche, U. D. Hanebeck, and G. Schmidt, “Design Issues of
Mobile Haptic Interfaces,” Jour- nal of Robotics Systems, Vol. 20,
No. 9, pp. 549-556, 2003.
[6] F. Barbagli, A. Formaglio, M. Franzini, A. Gianni- trapani, and
D. Prattichizzo, “An experimental study of the limitations of
mobile haptic interfaces,” In Proc. ISER 2004, 2004.
[7] Formaglio, M. D. Pascale, and D. Prattichizzo, “A mobile
platform for haptic grasping in large envi- ronments,” Virtual
Reality, Vol. 10, No. 1, pp. 11-23, 2006.
[8] Leow Y.P., Low K.H., Loh W.K., “Kinematic modeling and analysis
of mobile robots with omni-directional wheels,” In Proc. ICARCV
2002. Vol. 2, pp. 820-825, 2002.
[9] Yong Liu, Xiaofei Wu, J Jim Zhu, Jae Lew, “Omni-Directional
Mobile Robot Controller Design by Trajectory Linearization,” In
Proc. American Control Conference, 2003.
[10] L. X. Wang, A Course in Fuzzy Systems and Control,
Prentice-Hall, Inc., 1997.
[11] E. Dorjgotov, S. Choi, S. R. Dunlop, and G. R. Bertoline,
“Portable Haptic Display for Large Im- mersive Virtual
Environments,” In Proc. HAPTICS 2006, pp. 321-327, 2006.
Fig. 15. A user playing with the virtual dolphin using
the PoMHI (left) and the visual scenes displayed to the user via
the HMD (right). The red sphere represents the HIP.
111
Main
Information
Plenary Talk I : Symbolic AI + Embodied AI = Dependable Robot
Intelligence?
Plenary Talk II : Visual SLAM in large environments
Plenary Talk III : Maximal Information Systems
Plenary Talk IV : Telehaptics - Principle, Design and
Implementation
Plenary Talk V : Mobiligence: Emergence of Adaptive Motor Function
through Interaction among the Body, Brain and Environment
Plenary Talk VI : Force and Visual Control for Physical Human-Robot
Interaction
Oral Session
TS1: Dependable manipulation
TS1-1 Design of a Humanoids Specific Anthropomorphic Robot Hand by
Imitating Human Finger’s Motion
TS1-2 Robot Skill Learning Strategy for Contact Task
TS1-3 Design of the Safe and Speedy Robot Arm with Variable
Stiffness Joint
TS2: Mobile Robot & Application
TS2-1 A User Study of a Mobile Robot Teleoperation
TS2-2 An approach for Modularization Design of Mobile Robot
TS2-3 Controller design of the propulsion system used in the
autonomous vehicle
TS2-4 Inspection of Insulators on High Voltage Power Transmission
Line
TS3: Trends in the North American Personal, Service and Mobile
Robotics Market
TS3-1 Trends in the North American Personal, Service and Mobile
Robotics Market
TS4: Humanoid & Application
TS4-1 Adaptive Communication Selection for Multi-Robot
Interaction
TS4-2 Structural Optimization of the Pelvis in a Humanoid
Considering Dynamic Characteristics
TS4-3 Grasp Planning for Three-Fingered Robot Hands using
Taxonomy-Based Preformed Grasps and Object Primitives
TS4-4 Inertial Sensor Alignment with Human Arm Orientation for
Humanoid User Interface
TS5: Ambient Intelligence
TS5-3 Unified S/W Platform for Ubiquitous Robot, AnyRobot
Studio
TS5-4 Mobile Haptic Interface for Large Immersive Virtual
Environments: PoMHI v0.5
TS5-5 Modeling of Indoor Space and Environment Sensor
TS5-6 Context Aware Service Agents for Ubiquitous
RobotsApplications
TS6: Navigation & Localization
TS6-2 A Simultaneous Localization And Mapping Algorithm for
Topological Maps with Dynamics
TS6-3 Square Root Iterated Kalman Filter for Bearing-Only
SLAM
TS6-4 Artificial Landmark Design and Recognition for
Localization
TS6-5 Range Sensor Scan Matching for Localization and Map
Building
TS7: Swarm & Robotics
TS7-1 Neighbor-Referenced Formation Control for a Team of Mobile
Robots
TS7-2 Kalman Filter based ZMP Estimation Scheme for Balance Control
of a Biped Robot
TS7-3 Predicting the Size of Spring Network Swarm in Quadratic
Potential Fields
TS8: Autonomous Vehicle
TS8-1 Fault Diagnosis of Aircraft Engine Sensor Based on Sequential
Probability Ratio Test and Kalman filter
TS8-2 A Shared Fate Approach to Improve UAV Pilot Performance and
Minimize Accidents
TS8-3 Unmanned Aerial Vehicle Rotorcraft: Identifying Landing Zones
in the Presence of Obscurants
TS8-4 Backward Parking Control of a Car with Passive Trailers
TS9: Bio-mimetic Mechanism
TS9-1 A Development of Wall-Climbing Robot with Pneumatic
System
TS9-2 Determination of Joint Angles for Fitting a Serpentine Robot
to a Helical Backbone Curve
TS9-3 An Insect-Like Flapping-wing Device Actuated by a Compressed
Unimorph Piezoelectric Composite
TS9-4 An Undulatory Locomotion Controller for Snake-like Robots
Based on Adaptive Central Pattern Generators
TS10: Sensor & Actuator
TS10-2 A New Adjustable Actuator for Vehicle Engine
TS10-3 Fusion Method for Multi-sensor Dynamic Data
TS10-4 Personal Information Extraction Using A Microphone
Array
TS11: Vision & Application
TS11-1 Recognition of Moving Objects in Mobile Robot with an
Omnidirectional Camera
TS11-2 Recognition of Objects with Legs Using Vision-based
Candidate Generation and Evaluation
TS11-3 Recognition of Human action in Dynamic image using Entropy
measurement based on Frequency of Motion region
TS12: Bio-Signal Processing
TS12-1 Traffic Sign Recognition for Intelligent Vehicle/Driver
Assistance System Using Neural Network on OpenCV
TS12-2 Design a Reject Output for Pattern Recognition Neural
Network by Using a Dynamic Radial Basis Function Network
TS12-3 Improvement of Real-Time Object Tracking Performance of The
ROBOKER Head by RBF Network Compensation
TS12-4 Pattern Recognition of Typical Grasping Operations Using a
Wavelet Feature of EEG Signal
TS13: Design and Synthesis of Mechanical System
TS13-1 Visualization of Cornea Dynamics
TS13-2 Scenario Analysis of Sustainability in Global Resource
Circulation
TS13-3 Preliminary Study on Flexible Product Family Deployment by
Optimal Anticipatory Design of Common Components
TS13-4 Development of Quasi-3D-Rehabilitation-System, “Hybrid
PLEMO”
TS14: Industry Session I
TS14-1 Advanced CMOS Image Sensor for Robot Eye
TS14-2 The Development Concepts and Business Model for Intelligent
Service Robot
TS14-3 Development of the Humanoid Research Platform,
URIA(Ubiquitous Robotics Information Assistant)
TS15: Industry Session II
TS15-2 Improvement of iGS Positioning Sensor for Ubiquitous Robot
Companion
TS15-3 EDS-EDRS(Exciting & Dynamic Robot control Simulation
software)
TS16: Towards New Trends of Robotics
TS16-1 Collaborative Monitoring Using UFAM and Robot-2ndreport :
Development of integrated management system-
TS16-2 Architecture of Navigation System Using CBD Method and UPnP
Middleware
TS16-3 Basic Consideration on Robotic Food Handling by Using Burger
Model
TS16-4 Robot Town Project: Sensory Data Management and Interaction
with Robot of Intelligent Environment for Daily Life
TS16-5 Minimally Invasive Orthopedic Surgery System Based on
Biologically Compatible Processes
TS17: Interactive Human-Space Design and Intelligence
TS17-1 Intelligent Ambience-Robot Cooperation-Closing Door with
Humanoid Robot-
TS17-2 Multiple Objects Localization in Intelligent Space -
Utilizing User Hands Position Information from Position Server
-
TS17-3 Information Recommendation Module for Human Robot
Communication Support under TV Watching Situation
TS17-4 Evaluation of Mental Stress by Analyzing Accelerated
Plethysmogram Applied Plethysmogram Applied Welfare Space based
on
TS17-5 A common platform for Robot Services through Robot Services
Initiative(RSi) activities
TS17-6 Development of the Home Appliance Component using
RTC-Lite
Poster Session
PO-1 Development of Intelligent BioRobot Platform for Integrated
Clinical Test
PO-2 Development of Flexible Mobile Agents for BioRobot System in
the Clinical Laboratory
PO-3 Parameter Identification of Dynamic Systems using Hamiltonian
Regressor
PO-4 Conflict Evaluation Method for Building Grid Maps with Sonar
Sensors
PO-5 Active Resampling for Rao-Blackwellized Particle Filter
PO-6 Multivariate Fuzzy Decision Tree for Hand Motion
Recognition
PO-7 A Map Merging Technique using the Fingerprint Matching Method
in Multi-Robot FastSLAM
PO-8 A New Motion Planning Algorithm for a Biped Robot Using
Kinematic Redundancy and ZMP Constraint Equation
PO-9 Example-based Spoken Dialog Processing for Guidance
Robots
PO-10 Robust Localization for Mobile Robots in Dynamic
Environment
PO-11 Real-time Simultaneous Localization and Mapping using
Omnidirectional Vision
PO-12 Systematic Error Calibration of Mobile Robot using Home
Positioning Function
PO-13 Expression-Invariant Face Recognition Using Tensor Based
Image Transformation
PO-14 Development of Omnidirectional Human Detection System for
Mobile Robot
PO-15 Realization of a Service Robot HSR-I for Environment
Security
PO-16 Laser Scanner-based Navigation for Indoor Service Robot
PO-17 A Digitally Trimmable Continuous-Time Capacitive Readout ASIC
for MEMS vibratory gyroscope for Robot Applications
PO-18 State Estimation with Delayed Observations Considering
Uncertainty in Time
PO-19 Robust Data Association for the EKF-based SLAM
PO-20 Development of a scheduling algorithm for efficient tests on
Bio Robot
PO-21 Navigation Strategy for the Robot in the Elevator
Environment
PO-22 A High-Performance, Low-Cost INS for Autonomous Mobile
Robots
PO-23 Robust Multidimensional Scaling for Multi-Robot
Localization
PO-24 Detection of Moving Objects by Optical Flow Matching in
Mobile Robots using an Omnidirectional Camera
PO-25 Probabilistic Grid Matching Method for Grid-Based SLAM Using
Sonar Sensors
PO-26 Robust Control of Coordinated Motion for Underwater
Vehicle-Manipulator System with Minimizing Restoring Moments
PO-27 Efficient Feature Tracking for Visual SLAM by Pairwise
Constraints
PO-28 The Binary Recognition Algorithm Using Point Correlation
Template
PO-29 Flexible Docking Mechanism with Error-Compensation Capability
for Auto Recharging System
PO-30 Terrain Classification Using Texture Recognition for
autonomous robot
PO-31 Efficient Area Coverage Method for a Mobile Robot in Indoor
Environments
PO-32 The implementation of URC Robot Management Server
PO-33 Landing Platform for Experiment of Stable Humanoid
Walking
PO-34 Redundancy-Based Navigation Algorithm for Swarm Robot
Systems
PO-35 Gomy: The Baby Bear-like Robot for Emotional Human-Robot
Interaction
PO-36 KOBIE: A Pet-type Robot with Emotional Behaviors
PO-37 Multi-Arm Path Generation Method for Humanoid Robots
PO-38 Position Detection of KHST(Korean High Speed Train) using GPS
System
PO-39 Temperature Characteristics of Main Transformer for KHST
using Sensors
PO-40 A Study on Propulsion System Characteristics of KHST(Korean
High Speed Train)
PO-41 Traction System Characteristics of TTX(Tilting Train eXpress)
Using Measurement System
PO-42 The Study of Tilting System Combination Test of EMU Tilting
Train
PO-43 Robot Arm Tele-operation using Wearable Electronic
Device
PO-44 A Study on Motor Temperature Characteristic of TTX(Tilting
Train eXpress)
PO-45 Study on the URC robot simulation that used intelligent robot
simulator
PO-46 Contact Stability Analysis and Coordinative Manipulation of a
Dextrous Robotic Finger Mechanism
PO-47 Cosegmentation as a Generalized Correspondence Problem-A
Motion Model Driven Approach
Search
Help
Exit