Department of Electrical and Control Engineering

beaverswimmingΤεχνίτη Νοημοσύνη και Ρομποτική

14 Νοε 2013 (πριν από 3 χρόνια και 11 μήνες)

187 εμφανίσεις

Department of Electrical and Control Engineering


T
itle
:
The Research on the Intelligent Control System of Remote Video Monitoring

Principal Investigator
:
Lee Tsu
-
Tian

Sponsor
:
National Science Council

Keywords
:
Neural Networks
,
Monitoring Control
,
Wavelet

Transform
,
JPEG2000
,
Embedded System
,
Encryption Technology
,
Chaotic System
,
Face
Recognition
,
Vehicle
License
Plate
Identification


The Intelligent Control System of Remote Video Monitoring (ICSRVM) will
combine the research results of a high quality ima
ge compression technology and an
intelligent
information

encryption system (IIES) to promote the e
-
home security and
transportation security. The ICSRVM will control a rotary platform with CCD
cameras and apply the image
reorganization

techniques to search
, track and identify
the targets automatically. The research result of this project can be utilized in the
Intelligent Transportation Systems (ITS). Moreover, considering the research trend in
the post
-
PC age and the demand of data security, we will implem
ent our research
result in an embedded system with IIES

This project includes 9 topics below:

1.

Image Compression Technology

2.

Image Processing Technology

3.

Object Segmentation Technology

4.

Remote Image Transmitting Technology

5.

Rotating Camera Platform Control Tec
hnology

6.

Face Recognition Technology

7.

Vehicle
License
Plate
Identification

8.

Design and Implementation of Embedded System

9.

Intelligent Information Encryption Technology

NSC91
-
2213
-
E009
-
088

91R319


-------------------------------------------------------------------------------------------------------


T
itle
:
Research on
Control, Sensing and Information Technology of Intelligent
Vehicles

Principal Investigator
:
Lee Tsu
-
Tian

Sponsor
:
National
Science Council

Keywords
:
Intelligent Vehicles, Intelligent Control


This is he group project of “Research on control, Sensing and Information
technology of an intelligent vehicle,” In this research, we shall focus on the study of
some key technologies of
passive collision warning and avoidance systems of a
intelligent vehicle. These passive collision avoidance systems will be designed,
developed and integrated together to form a humanistic vehicle safety system.
Specifically, we shall integrate the linguis
tic man/machine interface, the longitudinal
collision avoidance radar, the real
-
time monitor system, the image recognition and
tracking system, the intelligent guidance systems, and the battery energy management
system to form the passive collision avoidan
ce system of the intelligent vehicle.
Various different sensors and intelligent peripheral devices are linked via the
CAN
-
bus to achieve the integrated environment detection and warning capability of
the system.

NSC91
-
2213
-
E009
-
028

91R116


-------------------------------------------------------------------------------------------------------


T
itle
:
Behavior Transformation and Real
-
Time Simulation for Multifunctional
Dynamic VR
-
Based Exercise and Rehabilitation Training Systems

Prin
cipal Investigator
:
Lee Tsu
-
Tian

Sponsor
:
National Science Council

Keywords
:
Motion Simulator, Virtual Reality, Washout Algorithm, Motion Cue,
Reinforcement Learning, Real
-
Time Simulation.


In the research of motion simulator, the reaction of the considere
d system must
be truly emulated in the notion platform. In this project, we attempt to develop
algorithm for transferring the behaviors of the considered system in the simulated
space into that of the emulated space. Traditionally, those tasks are fulfille
d by the
so
-
called washout filters . However, most existing research is for flight simulators, In
this project, we will try various approaches in designing washout filters and motion
cue transformation systematically. Due to the major concern is the feelin
g of the
operator, we also attempt to incorporate reinforcement learning concept into the
construction of washout filters and/or motion cue transformation rules. In order to
have a more accurate simulation, ways of building models are also studied. The ide
a
is not to propose a accurate models by means of reinforcement learning.

In the first year of this three
-
year project, we shall study the ideas and
variations of washout filters and implement neural fuzzy systems in modeling various
targets. In the second
, we shall focus on the reinforcement learning and its
applications to the design of motion cue and the construction of system models. In the
final year the research focus will be on the integration of subprojects and the design
of the human
-
machine interf
ace.

NSC91
-
2213
-
E009
-
036

91R312


-------------------------------------------------------------------------------------------------------


T
itle
:
I
ntelligent Multi
-
Objective Optimal Decision and Control of Unknown
Nonlinear Dynamical Systems

Principal Inv
estigator
:
Lee Tsu
-
Tian

Sponsor
:
National Science Council

Keywords
:
Uncertain Nonlinear Systems
,
Multi
-
Objective Decision &Control


This project is the subproject no.1 of the group project “ Research on
Control

S
e
nsing and Information technolog
y for a inte
lligent vehicle.” This project is
focused on the development of an algorithm for intelligent multi
-
objective optimal
decision and control of unknown nonlinear dynamical systems. The developed results
will serve as a major tool of the decision and control o
f a intelligent vehicle.

Major research works of this three
-
year project include the development of a
hierarchical structure for solving the proposed multi
-
objective decision and control
problem. Firstly, the observer
-
based output feedback control law and
update law to
rune on
-
line the adjustable vectors of the adaptive fuzzy
-
neural controller for output
tracking for a class of unknown dynamical system is developed. Then a multi
-
layer
fuzzy
-
neural system, serving as the upper level coordinator and decision
making
purpose, is properly design to fuse the n
-
actions determined by the n low
-
level
adaptive fuzzy
-
neural controller to determine the optimal action acting on the plant at
each time step. The Simulated Annealing will be employed to solve the proposed
mu
lti
-
objective optimal decision and control problem. Results will be compared with
those obtained by other different approaches.

NSC91
-
2213
-
E009
-
029

91R117


----------------------------------------------------------------------------------------------------
---


T
itle
:
A Study of Intelligent Networking Power Electronics for Aerospace Systems

----

S
ub
-
Project IV: System Programming, Analysis, Monitoring Design
and The Build Up of A Test
-
Bed Electronics (1/3)

P
rincipal Investigator
:
Der
-
Cherng Liaw


Sponsor
:
N
ational Science Council

Keywords
:
Repetitive Control, Learning Control, Iterative Inverse Operator


Due to the constraint of the finite energy and the growing demands of power
load, high efficiency, reliability, maintainability, fault tolerance and prevent
ion, and
system monitoring become most important issues in aerospace power electronics
system. In this project, an intelligent networking scheme is proposed to tackle these
issues. The proposed works will cover the following subjects: building up of a powe
r
electronics test
-
bed, topological analysis and design for networking, control scheme
for system monitoring, system analysis and design for resonant power converter, the
development of the database and related AP for system auto
-
testing, and the
intellige
nt scheme for system integration. In the first year, we propose to work on
system modeling, analysis, and control programming of the intelligent power
modules. In the second year, we will focus on the improvement of the performance
for power conversion dev
ices and supervising network programming and the design
of the intelligent network for aerospace power electronics system. In the third year,
we propose to complete the functional verification and performance enhancement of
the designed networking system.

NSC91
-
2213
-
E009
-
034

91R121


-------------------------------------------------------------------------------------------------------


Title
:
Development of the Techniques for Individual Multi
-
loop Control and
Electronic System Stability

Principle Investiga
tor
:
Der
-
Cherng Liaw

Sponsor
:
Precision Machinery Research & Development Center

Keywords
:
reliability

control

monitoring


R
ecently, it is known that system reliability become one of important issues in
engineering applications.
I
n this project, we propose
to study such an issue from
theoretical as well practical implementation points of view, specifically, for
microcomputer or computer based control system.
T
he effects of the number of
redundant controllers and corresponding topological hierarchs on system
performance

and system reliability will be first discussed.
A

crucial guideline of
system architecture for guaranteeing system reliability will be constructed, which will
then be applied to a small test
-
bed for experimental verification.
I
n addition, a stu
dy
of the technology of MP
-
based control system connecting to internet will also be
covered in the proposed works.

C91031

(
9
1
.
3
.1
-
9
1
.12.31)

---------------------------------------------------------------
------


Title
:
neural fuzzy network, genetic algorith
m, speech segmentation and
enhancement, adaptive time
-
frequency parameter, two
-
dimensional
cepstrum.

Principal Investigator
:
Chin
-
Teng Lin

Sponsor
:
Chunghwa Telecom Laboratories

Keywords
:
N
eural fuzzy network
,
genetic algorithm
,
speech segmentation and
enh
ancement
,
adaptive time
-
frequency parameter
,
two
-
dimensional
cepstrum.


To solve
the problem
s

that

background noise
caused by the channel
communications
can decrease

the performance of
the
speech segmentation and
enhancement
,
new methods
have

been develop
e
d
.

First,

we proposed a new
(ATF
-
based SONFIN algorithm)

in fixed noise
-
level environment.

This method
contains an adaptive time
-
frequency (ATF)

parameter for extracting both the

time and
frequency features of noisy speech signals.

The ATF parameter can ex
tract useful
frequency

information by adaptively choosing proper bands

of the mel
-
scale
frequency bank.

W
e further proposed a new word boundary detection algorithm

by
using a self
-
constructing neural fuzzy inference network

(called SONFIN)

for
identifying
word signals in noisy environment.

Due to the self
-
learning ability of
SONFIN,

this ATF
-
based SONFIN algorithm avoids the need of empirically

determining thresholds and ambiguous rules
.
In fact,

the background noise level may
vary

during the procedure of r
ecording

due to movements, engine running, speed
change, braking, slams, etc.

To solve this problem,

we proposed a minimum
mel
-
scale frequency band

(MiMSB) parameter
,

which can estimate the varying
background noise

level by adaptively choosing

one band wit
h minimum energy from
the

mel
-
scale frequency bank.

With the MiMSB parameter,

some preset thresholds
used to find the boundary of word signal

are no longer fixed in
entire

recording
interval.

These thresholds will be

tuned according to the MiMSB parameter.

We also
add

an enhanced time
-
frequency (ETF) parameter
, which can
extract
more
useful
frequency information
. Finally, we integrated the above parameters with GA
-
based
MTDC methods. T
he two
-
dimensional

(2
-
D) cepstrum (TDC) has the advantages

of
simple comp
utation and less storage space.
W
e

propose the GA
-
based M_TDC
(modified TDC) method to

improve the representativeness and robustness of the
selected TDC

coefficients in noisy environments. Furthermore, in the GA
-
based
M_TDC method, we apply the genetic

alg
orithms (GAs) to find the robust
coefficients in the M_TDC

matrix

in order to improve the recognition rates.

C91004

91.1.16
-
92.1.15


-------------------------------------------------------------------------------------------------------


Title
:
Multifuncti
onal Dynamic VR
-
Based Exercise and Rehabilitation Training
System

Principal Investigator
:
Chin
-
Teng Lin

Sponsor
:
National Science Council

Keywords
:
Virtual Reality, Dynamic Simulation,

Rehabilitation, Intelligent Control,

Motion Cue,

Force Feedback,

Intell
igent Tutor, Expression Recognition


This project is to develop a multifunctional exercise

and rehabilitation training
system using

virtual
-
reality (VR) based dynamic real
-
time

simulation techniques.
This system consists of

a multifunctional exercise
-
train
ing mechanism

mounted on a
stewar
t

platform.

By using the training mechanism and

immersing in the 3D VR
scene,

the user can feel that he/she is doing

exercise at a real outdoors site,

e.g.,
rowing a boat in a beautiful lake.

The proposed system is a highly

integrated system

of electrical/mechanical engineering,

and information science.

It includes the 3D VR
display, real
-
time simulation,

stewar
t

platform, exercise
-
training mechanism,

and
force feedback subsystem.

This system can provide the user with realis
tic feeling

(including vision and motion) of the real cases.


The major feature of the proposed system

is that it can sense the physical and
mental status

of the user from time to time such that

the scene and content of exercise
can be changed

properly and

automatically to meet the user's need.

For example, if
the user has a fast heartbeat

and tired expression when doing the jogging exercise,

our
system will change the scene from a uphill road

to a downhill road, and tune the
attitude of the motion

platform

properly. This will encourage the user

to do the
exercise continuously,

and achieve the effect of proper training.

This system can be
used both for regular exercise

and for rehabilitation purpose.

In this system, there is
also an intelligent machine
-
tutor
.

With the knowledge gained from exercise experts

through machine learning techniques,

the tutor can give comments and advise to the
user

to help him/her to achieve more successful exercise.


This project consists of six subprojects working

on eight major
subjects to set
up a multifunctional

dynamic VR
-
based exercise and rehabilitation training system.

These eight subjects are as follows:

1.

Physical modeling of the simulated scene and mechanism

for a specific exercise
item

2.

Behavior transfer between real world

and virtual world

3.

3D imaging of real scene and its integration

with interactive 3D VR scene

4.

Mechanism, motor driving system, and control

of an electrical stewar
t

platform

5.

Multifunctional mechanism for exercise and

rehabilitation, and its force feedback
co
ntrol

6.

Real
-
time detection and analysis of the physical and

mental status of the user

7.

Human
-
machine communication scheme

and feedback learning system

8.

Intelligent exercise tutoring system

In this three
-
year project, we shall set up the

hardware and software
of the
proposed system.

Based on it, we shall focus on three exercise items, i.e.,

run
n
ing
exercise, boat
-
rowing exercise,

and horse
-
riding exercise.

NSC91
-
221
3
-
E009
-
035

91R311


-------------------------------------------------------------------------------------------------------


Title
:
Intelligent Sensory and Control of The Dynamic VR
-
Based Exercise and
Rehabilitation Training System

Principal Investigator
:
Ch
in
-
Teng Lin

Sponsor
:
National Science Council

Keywords
:
Virtual Reality, Stewart Platform, DSP,

Real
-
Time Embedded OS,
Optimal Fuzzy Control,

Kinematics Analysis, Exercise Patterns Analysis,

Human Physical Signal Analysis.


This project is the Subproject 2

of the group project

“Multifunctional Dynamic
VR
-
based Exercise and

Rehabilitation Training System.”

It aims at the design,
analysis, and control of

the electric motor
-
driven
S
tewart platform,

the sensoring and
control of the exercise
-
training

mechanism m
ounted on the Stewart platform,

and the
detection and monitoring of

the user's physical status.

Based on our past experience
in developing

a hydraulic Stewart platform,

we shall design an electrical
S
tewart
platform

driven by electric motors in this subpro
ject.

This motion platform will
become the base platform

of the whole exercise and rehabilitation training system.

We shall design the mechanism, and study the problems

of kinematics, singularity,
and working space

of the target motion platform.

We shall a
lso develop the position,
velocity,

and acceleration control schemes for this platform.

The second major target
of this subproject

is to design and implement a multifunctional

mechanism for
exercise training purpose,

which can mimic several exercise items
such as

run
n
ing,
boat
-
rowing, and horse
-
riding.

This mechanism will be mounted on the Stewart
platform

to form a complete mechanism for exercise and

rehabilitation training
simulation.

We shall also study the sensory and control problems

of this mechanism.

The third major target of his subproject

is to monitor the physical status of the user

doing exercise by analyzing the physical signals

of the user such as heartbeat rate,
pace speed.

According to this information,

the control unit will change the exercis
e
content

properly to fit the user's need.

For example, if the user has a fast heartbeat

when doing the jogging exercise,

the system will change the scene from a uphill road

to a downhill road, and tune the attitude of the motion

platform properly. This wi
ll
encourage the user

to do the exercise continuously,

and achieve more successful

exercise and rehabilitation training.

We shall also create a real
-
time computation
environment

for this training system in this subproject.

We shall develop a
stand
-
alone co
ntrol card

with DSP chip or Arm chip,

and implement a real
-
time
embedded OS for this card

to achieve the goal of real
-
time simulation.

NSC91
-
221
3
-
E009
-
037

91R313


-------------------------------------------------------------------------------------------------------


Title
:
Intelligent Image Processing Technologies on DSP Platforms

Principal Investigator
:
Chin
-
Teng Lin

Sponsor
:
NSC, Advance and Wis
e Technology Corporation.

Keywords
:
AI technology,

Human Visual System, Noise Removal,
Image
Interpolation
, Digital Signal Processor.



With the rapid advance of digital technology, the
consumer

digital image
capture, display
, and printing product
s such as

digital still cameras (DSC),
color
printer, and
multiple
-
function peripherals (MFC), etc.

become more and more popular.
In these systems, noise removal and image interpolation are two common and
necessary technologies.
Images are often corrupted by noise
in capturing procedure
due to noisy sensors

such as dead pixels or hot pixels
.

The challenge of noise removal
is
how
to remove the noise
without blurring and destroying the image. Image
interpolation is resizing an image from low resolution to high resolut
ion. It is
indispensable for format/resolution conversion. The challenge of image interpolation
is how to enhance the resolution of an image without
blurring and jaggedness
.

In this project, we shall develop the intelligent noise removal and image
interpo
lation technologies to solve the problems of conventional approaches and to
meet the
requirement
s of the related
Taiwan industries
. Human visual system (HVS)
combined with AI technology shall be the base of developing intelligent image
processing technolog
ies to
perform higher visual quality.

In addition, powerful digital
signal processors (DSP) will become the computing and control kernels of digital
imaging systems. The developed
technologies

shall be
optimized

and simplified to
operate on the DSP platfor
ms in real time.
It is expected that this research can elevate
the add
-
on values of the imaging products in the related industries, and push Taiwan
industries to move forward to the goal of “knowledge
-
based” economics.

91
-
2622
-
E
-
009
-
011
-
CC3 (91R771)

C9112
6

(
91.06.01
-

92.05.31
)

-------------------------------------------------------------------------------------------------------


Title
:
Realization of an Advanced Flight Dynamical Simulator

Principal Investigator
:
Chin
-
Teng Lin, Kuu
-
Young Young

Sponsor
:
NSC
,
Best System Inc. (BSI)

Keywords
:
FAA, 6
-
DOF motion platform, force
-
reflection joystick, HLA.


This project is to follow up the previous three
-
year project “Development of
real
-
time dynamic simulator,” and will promote our technique level developed in the
previous project to meet the FAA (Federal Aviation Administration) standard for
flight simulators. In the previous project, we have developed a real
-
time dynamic
simulation system by integrating the technologies of virtual reality, information
science, mec
hatronics, and control to provide the user with realistic feeling (including
vision and motion) of the real cases. However, the developed technologies and
system can be used for hi
-
tech entertainment and simple training purposes only. To
meet the strict re
quirement of professional personal operation training, we shall
follow the FAA standard to improve the mechanism, control module, dynamics
simulation, force
-
reflection joystick, etc. of the current system to meet the
international standard of a flight simu
lator. This new integrated system consists of
five subsystems:

1.

6
-
DOF motion platform

2.

Force
-
refection joystick

3.

Interactive virtual reality

4.

Real
-
time dynamics modeling and simulation

5.

HLA
-
based system integration.

To meet the FAA standard, the new system will

be constructed by performing
the following five major tasks:

Task 1: Mechanism improvement, analysis, and control of the 6
-
DOF motion
platform

Task 2: Real
-
time Dynamics Simulations and Virtual
-
Reality (VR) Scene
Generation

Task 3: Force Information Proce
ssing Techniques and Force
-
Reflection
Manipulators

Task 4: Real
-
time Control Cards and Embedded OS for 6
-
DOF Motion
Platform

Task 5: High
-
Level Distributed Network Architecture

In this three
-
year project, the industrial people will join us to develop the w
hole
flight simulation system. We shall follow the FAA standard and the industrial need to
finish each key component in the system step by step.

91
-
2622
-
E
-
009
-
002 (
9
1R707)

-----------------------------------------------------------------------------------
--------------------


Title
:
Control Technology Development

Principal Investigator
:
Chin
-
Teng Lin

Sponsor
:
National Science Council

Keywords
:
V
irtual
R
eality (VR)
,
VR Video Processing and Control Technologies
,
VR Audio Processing and Control Technologies
,
VR Multimedia
Communication Technology
,
VR
-
Based Dynamical Motion Simulators
,
VR
-
Based

Exercise and Rehabilitation Training System,
ITS Smart Car


Following the rapid growing of computer industries and multimedia
technologies, the researches and applicatio
ns of virtual reality (VR) is booming. The
applications of VR cover a wide range from driving/flying motion simulators to the
home entertainment and security of everyday life. By combining 3D video, 3D audio,
and sensing/computation devices, VR creates a v
irtual environment, in which humans
can interact with the virtual objects as if they were touching the real world. On the
other hand, through the inspiring from the signals such as audio, video, and force
information produced by the VR system, humans can i
mmerse in the virtual world. In
this world, people can communicate with the machine freely in any 3D form, not
limited to keyboard, mouse, and screen. This constitutes a new
-
generation technology
of human
-
machine interface.

The advanced developments of inf
ormation technology (IT) and artificial
intelligence (AI) have pushed the research on VR toward a new stage. IT can bring
higher reality and enhanced real
-
time processing to VR, and AI can make VR more
humanistic and intelligent with self
-
constructing and
learning abilities. The basic
measure of VR is to treat the perceptions of humans in a virtual world in such a way
to make them feel to be in some real environment. Since vision and hearing are two
most important perceptions of human being, a complete VR s
ystem should include
the processing and controlling of video and audio signals. To meet the needs of
next
-
generation VR systems, this project aims at developing the advanced intelligent
VR multimedia technologies, including video and audio signal processin
g and control.
In addition, since the data communication network is the backbone of a complete
virtual world, the multimedia communication technology is also a major research
topic in this project. Hence, three major research topics will be performed in th
is
project:

1.

Topic 1: VR Video Processing and Control Technologies.

2.

Topic 2: VR Audio Processing and Control Technologies.

3.

Topic 3: VR Multimedia Communication Technology.

To embed a VR system with intelligent reaction power, the ability of human
location i
s a must for producing real
-
time human
-
environment interaction. To identify
the person(s), and
his/her

location as well as moving tendency in an environment,
Topic 1 of this project will focus on three major problems: face detection/recognition,
human iden
tification, and moving object detection. For VR multimedia systems, two
major kinds of audio data are speech and non
-
speech audio signals. Topic 2 of this
project will process and control these two signals using different techniques. The
three major resear
ch problems in this topic are speaker identification, speech
recognition, and 3D sound (virtual concert). Topic 3 of this project is to develop a
wireless multimedia communication network for the transmission and exchange of
the data produced by the system
s created by the others two research topics of this
project. The short
-
range wireless communication networks we consider include IEEE
802.11 WLN, HomeRF WHN, and Bluetooth WPN.

The VR multimedia processing and control technologies developed in this
project

can be applied to many practical systems such as:

1.

VR
-
based Dynamical Motion Simulators: By combining the units of VR
scene, VR sound, motion platform, force
-
feedback joystick, and real
-
time
simulation, we can build a VR
-
based motion simulator for the purp
oses of
high
-
tech entertainment and manipulation training.

2.

VR
-
based

Exercise and Rehabilitation Training System: We
can develop a
multifunctional exercise and rehabilitation training system using
VR
-
based dynamic real
-
time si
mulation techniques. By using t
he training
mechanism and immersing in the 3D VR scene, the user
can

feel that
he/she is doing exercise at a real outdoors site, e.g., rowing a boat in a
beautiful lake.

ITS Smart Car: The smart car will play a key role in the next
-
generation
intelligent t
ransportation system (ITS). The technologies such as speech recognition,
speaker identification, face recognition, 3D sound, etc., developed in this project can
be used in the smart car to provide the riders with friendly car
-
human communication,
security,

and in
-
car entertainment etc.

NSC91
-
2213
-
E009
-
145

91R
543


---------------------------------------------------------------
------


Title
:
Circuit Interconnect Model Reduction


A Balanced Realization Based
Appoach (II)

Principal Investigator
:
Ching
-
An Lin

S
ponsor
:
National Science Council

Keywords
:
VLSI, Circuit Interconnect, Model Order Reduction, Balanced
Realization


Drastic increase of signal frequency and decrease of design feature size in
today’s high
-
speed electronic circuits has made circuit intercon
nects one of the
dominating factor in determining the circuit performance and reliability of deep
submicrometer designs. Even though in the GHz range interconnects are
transmission lines, computational reality requires that they be modeled as lump
circuit
s. In general, circuit interconnect is modeled as a lumped

two
-
port or
multi
-
port of very high order (through discretization, e.g. RLCG model). To
effectively simulate and analyze circuit performance, interconnect model order
reduction becomes very importa
nt. The goal of this research is to develop model
reduction method for effective interconnect modeling. The theoretical basis of the
proposed approach is balanced realization. Topics of this investigation include
balancing the modified node analysis equat
ions (MNA), passivity preservation,
efficient partial solutions of the associated Lyapunov equations, exploiting sparsity of
the associated circuit matrices and frequency weighting.

MATLAB will be used for
algorithm development and SPICE simulations will b
e used for model verification
purpose. It is expected that this research will contribute to the understanding and
techniques of interconnect model order reduction.

NSC91
-
2215
-
E009
-
069 (91R496)

---------------------------------------------------------------
--------------------------------
--------


Title
:
Blind Identification And Equalization of Wireless Communiation Channel

Principal Investigator
:
Ching
-
An Lin

Sponsor
:
National Science Council

Keywords
:
Blind Equalization, Wireless Communication, OFDM


We pr
opose to investigate blind identification and equalization of wireless
communication channel using cyclostationarity of second
-
order statistics. Topics
include recursive algorithms for combined blind identification and equalization, blind
adaptive equaliza
tion

algorithms for slowly time
-
varying channel and analysis of
identification and equalization problem of OFDM system

using
multivariable

system
theory.

NSC91
-
2219
-
E009
-
0
4
9

91R
373


--------------------------------------------------------------------------
-----------
------------------


Title
:
A Design of Supporting Dynamic System with Interactive Real
-
Image
-
Based
Scenes (

)

Principal Investigator
:
S. F. Lin

Sponsor
:
National Science Council, R.O. C.

Keywords
:
V
irtual
R
eality,
I
mage
-
B
ased
R
endering, Stewart
P
latform
.


The main goal of the subproject is to provide a vivid immersion virtual scene,
which is synthesized by 3D real
-
image
-
based rendering technologies and virtual
reality technologies.


This project will take three years, so we divide the project int
o three parts:
Last

year

(the

first year)
, we support
ed

the system of a dynamic jogging machine with
interactive real
-
image
-

based scenes.
This year
, we will design the virtual scenes of
the system of a

dynamic ship machine with interactive real
-

image
-
bas
ed scenes.
Finally, next

year, we will be

in charge of designing the interactive real
-
image
-
based
scenes of a dynamic rodeo machine.

To sum up, according to the different specific requirements of each system, we
do the following research: First, a suitable

scene environment is chosen, and the
image data have to

be settled down. Second, we use the obtained image sequence to
construct an environment

model by applying image
-
based methods, and study the
techniques of image morphing and

some other technologies r
egarding view syntheses.
Meanwhile, we will do researches to find

the relationship among user’s jogging speed,
the movie speed, and the gesture of the Stewart

platform. Finally, we will study an
efficient way for storing and retrieving massive image data.

NSC91
-
2213
-
E009
-
107

91R
323


-----------------------------------------------------------------------------------------------
--------


Title
:
System
-
Level Verification Interacting with Architecture Exploration

Principal Investigator
:
Lan
-
Rong Dung

Sponsor
:
National Science
Council

Keywords
:
SOC
,

System
-
Level
Verification
,

HW/SW Co
-
Simulation
,
Cost
Estimation
,

Performance Modeling
,

Architecture Exploration


The System
-
On
-
Chip(SOC) design encompasses a large design space.
Typically, the designer explores the possible architec
tures, selecting algorithms,
choosing architectural elements, and constructing candidate architectures.
Designing such a complex system is hard; designing such a system which will work
correctly is even harder. Design errors should be removed as early as

possible;
otherwise, errors detected at the later stages will result a costly, time
-
consuming
redesign cycles. Thus, the designer should face two distinct tasks in SOC design;
carrying out design process itself and establishing the correctness of a desig
n.
Design correctness is the main theme of this project. Interacting with architecture
exploration the proposed verification methodology partitions the design process into
three phases: cost estimation, performance modeling and h
ar
d
ware
-
software(HW/SW)

c
o
-
simulation. The first phase estimates the architecture costs in terms of power,
area, and delay. By interacting with architectural allocation, the cost estimation
verifies that the architecture adhere to system requirements. Following the first
phase,

performance modeling then simulates candidate architecture at the
performance level of abstraction. Models at this level of abstraction do not concern
actual data in the system, but rather the flow of data through the system. Hence, the
second phase tak
es the advantage of simplifying the complexity of the simulation and
modeling. Using the simulation results in terms of time
-
critical constraints, the
designer is able to verify the candidate architectures and, if necessary, refine the
design. Once finis
hing the first two phases, HW/SW co
-
simulation starts to perform
the simulation of heterogeneous architectures which contain hardware and software
instances. This phase, therefore, verifies the behavior of the SOC design. Up to
this point, the system
-
lev
el verification has been done and establishes the correctness
of a design at the system
-
level.

NSC91
-
2215
-
E009
-
079 (91R506)

----------------------------------------------------------------------------------------
--------------


Title
:
(Subproject IV) A Stu
dy of Real
-
Time Human Emotional Understanding and its
Mechanism for Intelligent Human
-
Computer Interface (II, III)

Principal Investigator
:
Jyh
-
Yeong Chang

Sponsor
:
National Science Council

Keywords
:
Virtual Reality
,
Facial Emotion Recognition
,
Facial Imag
e Sequence
Processing
,
Abstract Concept Representation


Various technologies produce numerous equipment and systems to mimic the
real world, which have affected our daily life, especially in entertainment, safety, and
training. Virtual Reality (VR) technol
ogy plays the vital role to the success of these
systems. An ideal VR system would allow full immersion of all of the user’s feeling,
thereby presenting a simulated system that the user can intuitively interact with. True
immersion of feeling and interacti
on of man and machine are the main issues to be
addressed when investigating any VR technology.

The purpose of this project is to construct a human
-
machine communication
interface between the users and the simulators, which can transfer user’s feeling sens
e
of to project 1, to transform the control variables for the project 2 control loop
design of the simulator of Steward platform. In this way, the performance of the
whole control loop and the method of adjusting parameters can be interacted with the
user

and the reality of the dynamics can be enhanced. Therefore, the project can be
viewed as a higher
-
level feedback unit of the whole simulator system.

VR system can be more genuine if the truly sense of users can be
communicated to the dynamic systems. This

can be achieved by on
-
line monitoring
the user’s facial emotion and transferring the user‘s feeling to the machine and
updating the appropriate control parameters. The feeling of human being is an
abstract concept and there is no corresponding physical q
uantity, i.e., universe of
discourse, for the membership function to define with. Research on how to extract the
user facial emotion, represent the abstract concept and devise a method that can
communicate with machine simulator constitutes the major conce
rns to be
investigated in this project.

In this three
-
year project, emotion extraction and feeling communication
techniques between users and dynamic simulators tailored for crane, boat, and
single
-
engine plane will be respectively developed. The emotion c
ommunication
mechanism is intended to be hooked up with the MPEG
-
4 standard so that it can also
be useful for the multimedia applications.

NSC91
-
2213
-
E009
-
108

(91R324)

----------------------------------------------------------------------------------------
---------------


Title
:

滿足
H
-
infinity
性能之低階穩健控制器設計

Principal Investigator
:

Sponsor
:
National Science Council

Keywords:

H

Controller, Robust Control, Controller Reduction


In this project, we will develop a low
-
order robust controller design procedure
using nu
merical methods for single
-
input, single
-
output linear time
-
invariable (LTI)
systems with disturbances. The object is to decrease the gap between the robust
controllers theory and applications. In general, the disadvantage of controllers design
method base
d on

H
-
theory is high order of the optimal controllers. The high
-
order
property makes controller realization hard. Therefore, we hope to get a low
-
order
controller with similar performances comparison with

H

opti
mal controllers’. We
will use The Control System Toolbox,

-
Analysis and Synthesis Toolbox, and
SIMULINK in MATLAB to analyze many kinds of controllers’ performances in
frequency domain and time domain. We will try to show that we hav
e the chance to
replace the

H

optimal controllers by low
-
order robust controllers. For some case
studies, we really got information showing that the performances of low
-
order
controllers’ are similar to the

H

opt
imal controllers’. Based on this approach, we
will apply it on the controller reduction of

H

high
-
order controller, to obtain low
order controllers for implementation of industrial application.

NSC91
-
2213
-
E009
-
059

(91R315)

-----------
-----------------------------------------------------------------------------
---------------


Title
:
On Fast Parallel
-
Processing CMAC Control Chips Which Contain Large
Amount of On
-
Chip Memory

Principal Investigator
:
Fu
-
Chuang Chen

Sponsor
:
National Scienc
e Council

Keywords
:
CMAC Neural Networks, IC Chip, Control Systems


We have been doing research on the CMAC neural network control system for
some years. Two major topics are:

1. Design the CMAC neural network controller.

2. Study the CMAC control law and

learning law.

Recently, we have a major progress in the CMAC neural network control
scheme [24]. The CMAC neural net controller we are designing is something
between PID controllers and Lyapunov
-
based controllers. Our CMAC neural net
controller is much mo
re powerful, yet more complex, than the PID controllers. The
important point is that our CMAC neural net controller is conceptually acceptable for
industrial engineers.

The major purpose of this project is to design the CMAC IC chip. This is the
only way t
o release the parallel
-
processing power of the CMAC neural network, and
to achieve the demand in real
-
time control. The major challenge is in the design of
the on
-
chip memory, which is not a simple task technically.

In order to increase parallelism, and to

increase the amount of on
-
chip memory,
we have to modify the CMAC algorithm, such that we can simplify the circuit. We
propose to use the systolic array and the LFSR to replace or reduce the number of
multipliers and dividers used.

After the CMAC chip is
completed, it will be used to build a six
-
degree control
board, which also includes a DSP chip. The purpose of this DSP chip is to compute
the PD control, handle feedback signals, coordinate the 6 CMAC IC chips, and
schedule all the tasks of the control bo
ard.

The main goal of this research is to replace the industrial PID controllers with
our CMAC neural network controller, so that the industry can really handle
nonlinearty and uncertainty problems.

NSC91
-
2213
-
E009
-
058

(
91R314)

----------------------------
------------------------------------------------------------
---------------


Title
:
A Study on Compensation and Analysis of Lithography Overlay Error

Principal Investigator
:
Mu
-
Huo Cheng

Sponsor
:
Industrial Technology Research Institute T

Keywords
:
Lithog
raph Overlay

Correlation Analysis

Response Surface

Modeling


Micro
-
lithography is a critical process in wafer fabrication of

semiconductor
manufacturing. Most steppers employ, recently, the

step
-
and
-
repeat technique for
increasing the resolution, robustnes
s, and

reliability in wafer mask exposure.The most
critical factor of the

step
-
and
-
repeat process is the lithograph overlay error that
should be

within an allowable tolerance. Accurate analysis and modeling of the

lithograph overlay error, therefore, is im
portant for improving the yield

and
manufacturing reliability by adjusting adequately the stepper

parameters.The
lithograph overlay error is commonly modeled as the sum of

interfield and intrafield
errors; the interfield error characterizes the

global effe
ct while the intrafield error
represents the local effect.

These

errors are all modeled as the linear combination of effects including the

translation, expansion, rotation, magnitude, and the third
-
order and the

fifth
-
order
trapezoids. This project aims t
o create a model including the

correlation between
fields such that more detailed analysis of the

lithograph overlay error can be obtained.

We plan to apply the technique of

correlation analyses and response surface
modeling to characterize the

effect betw
een fields on the overlay error. This project is
expected to

enable us to obtain a more accurate model of the overlay error, and thus

reduce the error by choosing appropriate stepper control parameters.

C91018

(9
1.03.01
-
91.
12
.
15)

-------------------------
---------------------------------------------------------------
---------------


T
itle
:
Research on Realization of Precise Motion Control Design

P
rincipal Investigator
:
Pau
-
Lo Hsu


Sponsor
:
National Science Council

Keywords
:
CNC Control, AC Servo Motor Cont
rol, Intelligent Supervisory System


In past 10 years, our group has been sponsored by the National Science
Council to conduct research on motion control design. In CNC motion systems, we
have developed a series of control design to improve motion accuracy

and related
reports are published on international transactions [Hsu and Houng, 1996; Yeh and
Hsu, 1999a, 1999b, 1999c]. Recently, we also involved in AC permanent magnetic
servo motor control design. With the DSP
-
based implementation, we have
successfull
y improved velocity response to 85 Hz and position loop to 42 Hz,
respectively.

Although the past achievements in motion control are solid, the problems in
realization of the developed results still exist. Basically, the more elaborate the
control design
is, the more sensitive of the control system subject to model
uncertainties and external disturbances may occur. Therefore, how to realize the
developed control techniques in practice will be our future
research

topics. There are
three
research

directions
in this project as in the following:

1.

To improve tracking accuracy for each servo axis, it is desired to design a digital
disturbance observer (DDOB) to overcome the model uncertainties and external
disturbances.

2.

To improve motion accuracy for CNC multiple

axes, it is desired to construct the
network through controller area network CAN bus and the proper compensation
value among multiple axes for contouring error can be thus transferred.

3.

A supervisory system through the network will be constructed in this
study. Thus,
current of the servo motor and the system identification results for the CNC
system can be sent to the supervisory system. The condition of the servo motor
and CNC machine can be thus monitored to ensure the quality of manufacturing.

This proj
ect is scheduled to conduct a practical research for two years. Based
on the past theoretical results, we will have a deeper study on the above three topics.
It is expected that the research results of this project will be beneficial to both
academies and
industries.

NSC91
-
2212
-
E009
-
035 (91R109)

------------------------------------------------------------------------------------------------------


T
itle
:
Research on Realization of Precise Motion Control Design

P
rincipal Investigator
:
Pau
-
Lo Hsu


Sponsor
:
Na
tional Science Council

Keywords
:
Intelligent Vehicle, Real
-
Time Supervisory System, CAN
-
Bus, Linux
Kernel


Recently, intelligent transportation system is concerned because of the
increasing car accidents and heavy traffic on the turnpike. To decrease the t
raffic
accident and increase driving safety, a real
-
time supervisory system which integrates
the CAN
-
Bus and Linux kernel technology is proposed in this project.

The sensor signals in the vehicle, such as the GPS data of navigation system,
road images of l
ane departure warning system, and sonic signals of collision
avoidance warning system, are integrated and transmitted by the CAN
-
bus linking to
the Linux kernel
-
based supervisory system with real
-
time and multi
-
tasking operating
enviro
n
ments. Then, the sup
ervisory system analyzes the collected data to issue the
right decision, and controls the vehicle taking corresponding actions to prevent traffic
damages.

Through this project, the proposed real
-
time supervisory system not only
effectively reduces the traf
fic damages in experiments but also increases the
capability of research and development for automobile industries in Taiwan.

NSC91
-
2213
-
E009
-
086 (91
R135
)

--------------------------------------------------------------------------------------------------
---
--


T
itle
:
Telepresence and Teleoperation of An Advanced Home Robot System

P
rincipal Investigator
:
Kuu
-
Young Young

Sponsor
:
National Science Council

Keywords
:
Home Robot, Telepresence, Manipulation System, Two
-
Axis Robot
Manipul
ator


Along with the develop
ment of network technologies and information
appliances, intelligent home management system becomes possible. Via computer
networks, family members can then operate the computers and electric devices in
home, take care of daily affairs, proceed with inform
ation transfer, and execute
security monitor, etc. When an intelligent home robot is introduced to this system, its
mobility and flexibility can highly enhance the interaction between family members
and home devices. Motivated by this idea, we propose an a
dvanced home robot
system in the integration project. In this project, we concentrate on the development
of the telepresence generation and manipulation system in the advanced home robot
system.

For the teleoperation involved in the project, it is crucial
to make the operator
feel actually present at the remote site. Due to the varying and uncertain home
environment, the robot inevitably would contact with the environment when
executing the task. Under this circumstance, telepresence generation becomes much

more challenging. On the other hand, variations in position and control during the
interaction between the robot and the environment will also influence the design of
the manipulation devices and the development of the manipulation strategies.

Based on t
he discussions above, this project includes the development of a
telepresence emulation system and a force
-
reflection manipulation device. We will
also construct a two
-
axis robot manipulator to equip on the home robot, acting as its
arm. The virtual realit
y techniques will be used to implement the telepresence
emulation. Images of computer vision are utilized to increase the resemblance
between the virtual and real environment. The demonstration of force feeling induced
from the interaction between the robo
t and the environment is a key subject for the
development of the virtual environment. And, the force
-
reflection manipulative
device is developed to let the operator feel the contact force from the remote
environment, so that effective manipulation can be
achieved. With these software and
hardware systems aforementioned, a three
-
dimensional real
-
time interactive
manipulation system is then accomplished.

NSC91
-
2213
-
E009
-
025

91
R113


-------------------------------------------------------------------------------------------------------


Title
:
The MultimediaCard Control System Design and The Application in Computer
Audio and Video Systems (3/3
)

Principal Investigator
:
Bing
-
Fei

Wu

Sponsor
:
National Science Council

Keywords
:
Chaotic Systems, MultiMediaCard
,

Information Encryption


The research combines chaotic system theory and Data Encryption System,
DES to be so called Chaotic DES, CDES. A CDES can enhance the capability of
Dif
ferential attack and Linear attack of DES by the uncertainty principle of chaotic
systems. And the results can be applied to a portable storage flash memory card,
MultimediaCard. So far, we integrate the core of the CDES, information encryption
engine, Mul
timediaCard controller, FAT manager, and EPP controller to be an
FPGA
-
based
CryptoMedia
.
Part results of the hardware design are

better than the
related
results

of a paper in IEEE conference.

NSC91
-
2213
-
E009
-
010 (91
R110
)

-----------------------------------
-------------------------------------------------------------------


Title
:
The FPGA
-
Based Design of an IrDA Protocol Chip with Encryption

Principal Investigator
:
Bing
-
Fei Wu

Sponsor
:
Chung
-
Shan Institute of Science and Technology

Keywords
:
IrDA

IrCOMM


The IrDA has wide wireless communication applications on Notebooks, PDAs,
mobile phones and some peripherals like printers, modems, LANs, or cameras. The
embedded famous OS, Windows CE, Linux, and Palm, also support standard IrDA
protocol. Because

of the diverse application areas, high speed and low cost, it reveals
that IrDA has very good growth opportunity in the future.

The system will implement the standard protocols of physical layer, the IrLAP,
and IrLMP. The physical layer solves different c
ommunication speeds and frame
formats of IrDA protocol. The IrLAP constructs the communication link between
primary and secondary IrDA devices. The IrLMP includes IAS and link management.

The objective of the project is to develop and implement an FPGA
-
bas
ed IrDA
chip with an embedded microcontroller that can support standard IrDA protocol with
IrComm. Since the data security is very important in IrFM (Infrared Financial
Messaging) system, we propose an encryption architecture so that the data flow in the
I
rDA chip will be secured.

I 91002
(91.0.01
-
91.12.31)

------------------------------------------------------------------------------------------------------


T
itle
:
Information And Networking Platform Research for Home Robot System

P
rincipal Investigator
:
J
wu
-
Sheng Hu

Sponsor
:
National Science Council

Keywords
:
Robotics, Home Robot, Embedded System, Real
-
Time Multi
-
Tasking
Control


The goal of this project is to develop an information platform for home robot
systems with special emphasis on power saving and
modern information and
connection capability. Using embedded system deign technology, the project is going
to construct a development environment based on real
-
time multi
-
tasking control
architecture. A three
-
layer control structure is proposed to define t
he hardware and
software technology involved in the system. From hardware viewpoint, we have: 1.
Real
-
time Control Layer; 2. Gateway and Server Layer; 3. Internet Layer. On the
other hand, the software layers are defined as: 1. Reflective Action Layer; 2.
Unitary
Intelligence Layer; 3. Cooperative L
earning

Layer. These will enable an infinite
internet resource sharing and a highly flexible growing space for the robot.

The project will use RISC CPU to design the central control unit for the robot.
Embedded L
inux will be used as the operating system. Meanwhile, based on the idea
of thin client and server, this research will establish a robot house to provide more
computing power. To make the overall software platform more transparent, we will
use JAVA and XML
techniques to design the message exchange and development
interface. Therefore, any computer connected to internet can easily become part of
the brain of the robot. Via internet connection, the robot can own an infinite resource
and computing power to esta
blish a huge and complex learning and cooperative
mechanism.

NSC91
-
2213
-
E009
-
024

91R112


-------------------------------------------------------------------------------------------------------


T
itle
:
Development of A Real
-
Time Jmage System and Image Track
ing Scheme of
Intelligent Vehicles

P
rincipal Investigator
:
Kai
-
Tai Song

Sponsor
:
National Science Council

Keywords
:
Image Processing ,
I
mage Tracking,
I
ntegrated
Image
Sensor,
I
ntelligent

Vehicles,
Cruise

Assistance


The real
-
time image system
can
provide
the driver
the
video information of the
vehicle’
s

surroundings
.

The
processed
image
data
can

indicate dangerous situations
.
This can be useful

for
assist
ing

the driver with more prompt
and
timely response.

This project
involves

the
design, implement and te
st
of
a real
-
time image system for
intelligent vehicles. The developed image processing platform will integrate
advanced
hardware and software to demonstrate image tracking applications.

In the first phase of this project,

CCD cameras integrate
d

with dedi
cate
processing unit are u
tilized
to

develop the stand
-
alone image system. A
car
-
computer
-
based real
-
time image system will be realized. In the second phase of
the project CMOS image sensors will be developed to replace conventional CCD
cameras. This self
-
developed sensor features a combination of image processing

c
ircuit to enhance system performance. The focus is on the design and fabrication of
integrated sensors with
small
er

size,
l
ower power consumption and higher
performance
than the conventional syst
em. This will be more practical and
beneficial

for
applications in
advanced intelligent vehicle
s
.

In this project we will develop image tracking scheme based on active contour
model. Moving object with arbitrary shape can be tracked and followed in order t
o
detect the
location
and
motion
tend
ency
of the surrounding vehicles. The obtained
information can be further used to generating warning signal to enhance the cruises

safet
y. We plan to complete the project in three years
.

For

the first year,
main
objecti
ve

is to develop image tracking method for moving vehicles.
The focus of the
second year is
the

design and implementation of
a real time image system for
intelligent vehicles.

In the third year,
integrated
CMOS image sensor
s

with on
-
chip
image processing
c
apability will be developed. In summary, this project will develop
real
-
time image processing module for
intelligent

vehicles. The results of this project
can be used by other projects of the intelligent vehicle team. The
application of image
tracking
to

c
ruise assistance

will be demonstrated
.

NSC91
-
2213
-
E009
-
033

91R120


-------------------------------------------------------------------------------------------------------


T
itle
:
Development of An Advanced Home Robot System

P
rincipal Investigator
:
Kai
-
Tai Song

Sponsor
:
National Science Council

Keywords
:
Robot Control, Image Processing, Embedded System, Teleoperation,
Tracking Control, Robot Learning


In 1970’s, industrial robots appeared on the market. These programmable
machines have been one of the most important equipment in automatic manufac
turing
factories. However, industrial robots are limited to factory environment. Nowadays,
there remains very few research challenges. Recently, the rapid progress in
microelectronics and sensing technology has promoted robot applications to home
appliance
. Intelligent robots are working together with human beings in office and at
home. This project involves the development of an advanced home robot system. The
developed intelligent robots will demonstrate human
-
like behaviors and have the
capacity of machi
ne intelligence. They will serve people and in the meantime
constantly interact with people in completing the specified task. The advanced home
robot research team consists of five projects:


1). The development of information and internet access platf
orm of a home
robot system.


2). The research on the telepresence and teleoperation of a home robot system.


3). The research on the computer vision of a home robot system.


4). The development of intelligent interaction system of a home robot.


5). The research on the image
-
based tracking control system of a home robot.

We plan to complete this research work in three years. The research contents
involve many challenge issues in control and information technologies, i.e. embedded
computing p
latform, embedded real
-
time Linux and its application to home robots,
image recognition, voice recognition, Intelligent man
-
robot interface, image tracking,
machine learning, multiple robot systems. The results of this study are related the
information app
liance (IA) products and can be transferred to local companies in IA
and 3C products development. It is greatly expected hat a closer interaction between
the university laboratories and company sectors can be achieved through the progress
of this home robo
t project.

NSC91
-
2213
-
E009
-
023

91
R111


-------------------------------------------------------------------------------------------------------


T
itle
:
Development of The Intelligent Interaction System of An Advanced Home
Robot




P
rincipal Investigator
:
Ka
i
-
Tai Song

Sponsor
:
National Science Council

Keywords
:
Robotics, Man
-
Machine Interface, Intelligent Systems, Multi
-
Robot
System,Machine Learning


About one year ago, Sony
announced

her first commercialized intelligent
robot
-
the AIBO. It has been a very suc
cessful product. Since then, Sony, Honda and
other companies, mostly in Japan, introduced
several

new types of humanoid robots.
Although these are not yet successful models on the market, one can forecast that in
the near future there will be robots workin
g together with human being in the office
and at home. These intelligent robots will demonstrate human
-
like behaviors and
have the capacity of machine intelligence. They will serve
people

and in the
meantime constantly interact with
people

in completing th
e specified task.

This project involves in the development and construction of an intelligent
interaction system for an advanced home robot system. In this study, the interactions
between the home robot and respectively, human being, the environment and o
ther
robots are considered. Therefore, in this project, the focus will be on the following
issues of a home robot: configuration, interface circuits of various sensors, signal
processing, architectures of robot behavior, machine learning and multi
-
robot
co
operation. In particular, as this robot is developed to work at home as an appliance,
the interaction between the robot and
people

determines whether it is accepted or not
at home. Consequently, the first item of this project will be the design and
constru
ction of a suitable mobile platform. Secondary, the interaction with the
environment will be studied. This includes the communication with intelligent home
appliance, such as TV set, telephone and the safety devices. Specifically, a fire
detection system u
sing special designed CMOS image sensor will be developed and
integrated with the robot. As for the communication and interaction with the members
of the family, the voice recognition and image recognition techniques will be
developed. It is sometimes more

efficient to allow several robots to work together for
certain tasks. A multi
-
robot cooperation architecture will be required for such
applications. By means of a reliable communication protocol between robots via
wireless LAN and an algorithm for robot t
o self
-
dispatch subtasks, robot cooperation
can be realized. We plan to complete this project in three years. The first year will
focus on interaction with the environment. A mobile platform and a self
-
localization
method will also be developed. The main g
oal of the second year will be the study of
the interaction with the human being. An intelligent machine
-
man interface will be
designed and realized. The purpose will be easy operative and interactive features
between machine and operators. The third year
involves in the interaction with other
robots. A robot cooperation architecture will be developed and tested.

NSC91
-
2213
-
E009
-
027

91R115


-------------------------------------------------------------------------------------------------------


T
itle
:
An Experimental DWDM System by Using Novel Devices: Subproject (VIII)

The Study of MEMS Based Optical Cross
-
Connect
Device for DWDM


P
rincipal Investigator
:
Jin
-
Chern Chiou

Sponsor
:
National Science Council

Keywords
:
Optical Cross
-
Connect, Optical Switch
,
MEMS
,
DWDM


The main objective of the present proposal is to design, manufacture, test, and
control design of MEMS b
ased optical cross
-
connect / optical switch for DWDM in
three years. The main purpose of these devices is to cross
-
connect and/or switch
optical light path in optical fibers for optical communication and data transmission.
The research will focus on the me
chanical design and analysis, the driving control
circuit design, the MEMS based process design and manufacture, and the
performance of the devices once they are manufactured. Furthermore, the developed
devices will integrate with other sub
-
projects to for
m optical module such as WADM
(wavelength adder/dropper module). We are expected to obtain an excellent optical
switch once the present sub
-
project is completed.

Abstract for the 1st year:

The research topic of the 1st year is divided into two parts: (1) s
urvey and
collect existing journey papers, thesis, and technical reports for DWDM using
different types of optical switches and/or cross
-
connect. With the information, we are
able to realize the principles and characteristics of the role of optical switche
s and/or
cross
-
connect in DWDM. Furthermore, we wish to study the design of the switching
mirrors and corresponding control driving circuits for our future design references. (2)
discuss our designed optical switches with other sub
-
projects with system int
egration
in mind. Upon completing the design, a 1 dimensional optical switch will be
manufactured using MEMS technologies. A self
-
developed optical measurement
system is used to measure the static and dynamic performance of the optical switch.
The research

will discuss the possibility of control the optical switch in digital and in
analog with high performance in mind.

Abstract for the 2nd year:

In order to fully comply the needs in optical switch for DWDM, we will
provide manufactured 1
-
D optical switch to

other sub
-
projects for application
evaluations. Upon all the tests are completed, we are able to modify our designs
according to the feedback from other sub
-
projects. Furthermore, with the
experimental experience gained in the 1st year, a 2 dimensional op
tical switch will be
under investigation, design, and manufactured. A new optical measurement system
for the 2
-
D optical switch will be developed for the purpose of measuring its
performance. Again, the research will look for the possibility of control the

optical
switch in digital and in analog with enhancing the performance of DWDM in mind.

Abstract for the 3rd year:

Under the supervision of the main project, we will provide manufactured 2
-
D

optical switch to other sub
-
projects for application evaluations
. Upon all the
tests are completed, we are able to modify our mechanical and control circuit designs
according to the feedback from other sub
-
projects. In order to make the system more
compact, we will utilize CIC’s chip design capability to implement our
electronic
circuits using Mix
-
mode VLSI design. Furthermore, for the sake of realizing the final
DWDM, we will study an appropriate packaging method for integrating the optical
switch module with other sub
-
projects. A series of tests will be performed to e
xamine
the reliability of the developed module.

NSC91
-
2215
-
E009
-
028

91R151


-------------------------------------------------------------------------------------------------------


T
itle
:
Intelligent Autonomous Mobile Robot Via Network
-
Based Control

P
rincipal Investigator
:
Yon
-
Ping Chen

Sponsor
:
Nationa
l Science Council

Keywords
:
Vehicle Navigation System, Global Positioning System,Inertial
Navigation System, Geographic Information System


This
proposal mainly focuses on the development of intelligent vehicle
navigation system, using global positioning s
ystem (GPS), inertial navigation system
(INS), geographic information system (GIS), and communication interface.

Recently, due to the development of GPS, to achieve the geographic
information is getting easier and more precise; However, although GPS is usa
ble all
the day, its signal is not receivable at locations covered by building, wood, etc. To
compensate such weakness, INS has been adopted simultaneously to record locations
when GPS signal is lost. Once the vehicle location is determined, combining with

GIS, it could automatically provide the vehicle user the information about
surroundings, such as street map, under path, railway, highway, hospital, .station,
park, etc.

Except the development of intelligent vehicle navigation system, this proposal
also
focuses on the system function design to make it more useful and realizable.

NSC91
-
2213
-
E009
-
032

91R119


-------------------------------------------------------------------------------------------------------


T
itle
:
DMT Transceivers for Frequency Selectiv
e

Channels

P
rincipal Investigator
:
Yuan
-
Pei Lin

Sponsor
:
National Science Council

Keywords
:
Discrete

Multitone Modulation (DMT), Optimal DMT
,
Multicarrier
M
odulation (MCM), Orthogonal Frequency Division

Multiplexing

(OFDM)
, Multiplexing


The DMT transceive
r is now one of the most commonly used

modulation
techniques for channels like digital subscriber lines,

e.g., ADSL and HDSL. It has
been reported that DMT systems can

approach within 8
-
9 dB of channel capacity of
digital subscriber

loops. Recently the DMT

transceiver systems have been adopted in

several standards for transmission over twisted pairs.

In our earlier results we established a polyphase framework for

M
-
band DMT
transceivers over arbitrary distorted channel with

general colored noise. It has be
en
shown that, for the important

subclass of block based orthogonal DMT system, the
optimal

solution can be obtained in closed form. Performance evaluation

has
demonstrated the optimal solution outperforms the widely used

DFT based
significantly. In this
research plan, we will consider

many important applications
using the multirate polyphase

approach. In the first year we will consider
extensions to the

general biorthogonal block based system. Non block based
systems

will be explored in the second years
. In particular, we will

investigate
DMT systems with maximum bandwidth efficiency, systems

robust to narrowband
interference. In the third year, we will

consider the complete DMT systems,
including the time domain

equalizer at the receiving end. The ti
me domain
equalizer at the

receiving end will be optimized for DMT applications. Performance

analysis will also be conducted for the proposed systems.

NSC91
-
2219
-
E009
-
031

91R169


-----------------------------------------------------------------------------
--------------------------


Title
:
Fault Tolerance, Signal Analysis and Reliable Control

Principal Investigator
:
Yew
-
Wen Liang


Sponsor
:
National Science Council

Keywords
:
Power Electronic Systems, Load Estimation, Reliable Control, Fault
Detection and Dia
gnosis


As the space vehicle technology
develop
ing,
the vehicle is designed more and
more complicated. Although the space technology provides the human being a rapid
and comfortable

transportation
, the space missions are still potentially highly
dangerous.

Due
to the potentially
highly
dangerous feature, the space vehicles are
inevitably
required to be more reliable, safe and efficient than other missions in land
and sea. Among the many

reliable
requirements of the space missions, the power
supply quality,
which is one of
fundamental and

important demands, must be stable
and efficient. However, due to the space vehicle has a limitation on its
total
weight,
it

s power supply system can

provide
only a limited

power.
Under this finite power
restriction, the spa
ce system might exhibit
undesirable
phenomena
such as bifurcation,
chaos and sub
-
harmonic oscillation
if
the following situations occur:
some of the
power generators outage,
system
under
went

a large disturbing signal or experienced
a
large load variation.
Therefore, it is an important issue to study
a

power distribution
strategy so that the power electric system
can

provide a stable and efficient power
quality in spite of the outage of
some of the
power

generators
,

large disturbing signals
and large load va
riations.
In addition to studying the reliable control issue and

building a model to be analyzed,
in this project,
we are also
about

to estimat
ing
load
variation
and investigating the fault detection and diagnosis issues. The objective of
the fault detecti
on and diagnosis issue is to monitor the power electric system and to
detect the occurrence of undesirable phenomena. Once an undesirable phenomenon is
detected, a warning signal and the recognition of where and what faults have
occurred will be provided s
o that the engineer can take an appropriate control strategy
to recover or prevent the system from occurring these undesirable phenomena.

NSC91
-
2212
-
E009
-
037

91R306


------------------------------------------------------------------------------------------
-------------


Title
:
Adaptive Backstepping Motion Control of

Induction

Motors

Principal Investigator
:
Shir
-
Kuan Lin

Sponsor
:
National Science Council

Keywords
:
Induction Motor,
DTC, Backstepping, Sliding Mode
.


In this proposal, an adaptive backstepping

controller
will be

proposed for the
position tracking of a mechanical system driven by an induction motor. The
mechanical system is a single link fixed on the shaft of the induction motor.
In this
proposal
, we directly adopt the sliding
-
mode
torque contro
l

as the inner loop torque
controller,

and
directly deal with

the motion control.
The induction motor drive is
influenced by the uncertainties, which usually composed of unpredictable plant
parameter variations, external load disturbances, unmodeled and no
nlinear dynamics
of plant
. T
o

overcome these problems, an adaptive
sliding
-
mode
backstepping
controller is
developed
. The
proposed

motion
control
ler

is shown to be stable and
robust to the parameter variations and external disturbances by simulation.
In th
is
project, a
n experimental system will be established
to implement the proposed
controller,
which
is
compose
d

of motion controller and driver controller.
To verify the
robustness,
a
DC motor driver will be established to generate

the

external
disturbance
.

NSC91
-
221
3
-
E009
-
071

91R317


-------------------------------------------------------------------------------------------------------


Title
:
Improvements on Ordinal Optimization

Principal Investigator
:
Shin
-
Yeu Lin

Sponsor
:
National Science Council in Taiwan

Keyword
:
Ordinal
Optimization, Computational Intelligence, Power Systems


In this project, we intend to propose methods for the following two significant
subjects in Ordinal Optimization Theory: (i) Finding the better N samples and (ii)
Constrained Ordinal Optimization Pro
blems.

In the first subject, we consider two types of hard optimization problems. The
first type is for systems with structural information and the second one is for the lack
of structural information system. Thus, we intend to use two years to deal with t
hese
two types of problems. In the first year, we will consider an example system that we
are familiar with and propose a heuristic method to find the better N samples and then
to find a good enough solution, which will be better than that can be found in
the
current Ordinal Optimization method. In the second year, we will combine the
Ordinal Optimization method with an intelligent computing method to find the better
N samples for a lack of structural information system's hard optimization problems
and then

to find a better good enough solution. In the third year, we will propose a
new method to solve constrained ordinal optimization problem and provide rigorous
theoretical support for the proposed method.

NSC91
-
2213
-
E009
-
070 (91R316)

-----------------------
--------------------------------------------------------------------------------


Title
:
High Speed Link Driver Circuit Design

Principal Investigator
:
Chauchin Su


Sponsor
:
Taiwan Semiconductor Manufacturing Company Ltd

Keywords
:
High Speed Data Communicat
ion, Integrated Circuit Design, Signal
Integrity, Serial Link.


In this project, we proposed a research proposal on the design of high speed
data link. When the data rate exceeds giga bits per
second
, not only the design of the
driver and receiver circuits

but also the topological design of the package and
interconnects on the printed circuit board are challenging tasks. The main theme of
this project is to study the impact of the package and interconnects on the high speed
data links to conclude design gui
de lines for the driver and receiver circuit designs.
We will study the circuit modeling techniques, derive the design guide lines, and
implement the driver circuits to verify the feasibility of the techniques. The major test
vehicle is on the bus driver d
esign for the data rate of 2.5Gbps/3.2Gbps.

C91214

(9
2.01.01
-
92.
12
.31
)