BEE4333 Intelligent Control

siennaredwoodIA et Robotique

23 févr. 2014 (il y a 3 années et 5 mois)

61 vue(s)

BEE4333 Intelligent Control

Hamzah

Ahmad

Ext: 2055/2141

Course Outline

Reference

ARTIFICIAL INTELLIGENCE:

WHAT IS IT ALL ABOUT?

Chapter 1

Achievements

:PO1, CO1, CO5, LO1

Today contents and achievements


LO1

: Able to understand
intelligent system and its
application

1.1

Overview of artificial intelligence (AI)

According to the Oxford and Penguin English Dictionaries
the word “
intelligence
” can be defined as follows:


ability to understand


reason


perceive


quickness in learning


mental alertness


ability to grasp relationships


clever


information


news

One way to understand “intelligence” is by looking at our
own capabilities,
which means that
humans are able to:


think


understand


recognize


perceive


generalize


adapt


learn


make decisions


solve daily problems

Artificial Intelligence (AI) ?



AI

is

a

study

about

inventing

machines/computers

that

capable

of

mimicking

human/animal

intelligent

behavior
.



http://xmb.stuffucanuse.com/xmb/viewthread.php?tid=6825


The ultimate objective is
to develop a system that
can think and act
rationally like humans.

How do we design Intelligence?


Study from biological models (brain, genetic, DNA, life,
Molecular biology, ….)


neural nets, GA, Artificial Life, DNA
Computing, Quantum Computing, Robotics, etc.



Study from human phenomena (common sense, reasoning,
predicting, observing, inference, …)


fuzzy logic, expert
systems, search techniques, etc.



Need to develop mathematical/logical algorithms based on the
above biological models or phenomena


ARTIFICIAL INTELLIGENCE:
COMPARISON BETWEEN CLASSICAL
CONTROL AND INTELLIGENT CONTROL

Chapter 1

Achievements :PO1, CO1, CO5, LO2

Today contents and achievements


LO1

: Able to understand intelligent system
and its application


LO2

: Able to compare classical control
system and modern intelligent system

1.2

Artificial intelligence applications

1.3
Comparison with classical controller

Intelligent Control


Classical Control

Classical Control
Intelligent Control
Basic Concept
Mathematical Modeling - Designer
designed the system which includes
system dynamics
Abstract Modeling - Designer input the
behavior to the system and then
system attempt to abstractly define the
system
Need to know prior information about
the system dynamics
Does not need to know all about the
system dynamics and conditions
Suitable for system that can be
easily model
Appropriate for complex system
Open loop system
Fuzzy logic
Closed loop system
Artificial Neural Network
System Modeling
Genetic Algorithm
Support vector machine
Swarm Intelligence
Particle Intelligence
Characteristics
Examples of
Methods
Intelligent Control


Classical Control

Software

Designer

Designer

Software

Classical Control

Intelligent Control

INTELLIGENCE

Some Pre
-
requisites in understanding
AI


Some mathematical background


HLL Programming


Discrete
-
time systems


Some aspects of biological systems as well
as philosophy and psychology


………..


Where AI can/should be applied?


Data is overwhelming/abundance


Too many manual operations/procedures


Optimization is possible


Parallel/Distributed procedures/architectures
are needed


Decision making is required


When current techniques are too complicated
to be used/designed


Where AI can/should be applied?

.. Cont’d


Mathematical models are too
complex/impossible


To increase efficiency


To reduce cost


To improve performance and reliability




Some Important Facts, you need to
know….


AI is not the only solution


AI is only one part of technology


AI is just a tool for improvement


You must know your domain/target
application



Intelligent
Systems; Conceptual
Design

Intelligent
Man
-
Machine
Interface

Intelligent machine

Execution

TASKS

Algorithms,
computations

Cognition

(Algorithms
)

Perception

(sensors)

Expert
systems

Fuzzy logic

Neural
networks

GA

………….

Expert
systems

Fuzzy logic

Neural
networks

GA

………….

AI : Some of the approaches


Expert system


Fuzzy Logic


Genetic Algorithm


Swarm Intelligence


Ant Colony


etc

Expert System


Expert System (ES)

is a branch of Artificial
Intelligence that
attempt to mimic human
experts
specifically in
decision making
process based on
prior knowledge.



Expert systems can either
support

decision makers
or completely
replace

them.



Expert systems are the most widely applied &
commercially successful AI technology
.

Types of ES


Ruled Based Expert System


Represented as a series of rules


Frame
-
Based System


Representation of the object
-
oriented programming approach


Hybrid System


Include several knowledge representation approach


Model
-
Based System


Structured around the model that stimulates the structure and
function of the system under study


Ready
-
Made (Off
-
the
-
shelf) System


Custom
-
made, similar to application package such as an accounting
general ledger or project management in operation mgmt.

Rules as a knowledge representation technique


The term

rule
in AI, which is the most commonly used
type of knowledge representation, can be defined as
an IF
-
THEN structure that relates given information or
facts in the IF part to some action in the THEN part. A
rule provides some description of how to solve a
problem. Rules are relatively easy to create and
understand.


Any rule consists of two parts: the IF part, called
the
antecedent
(
premise
or
condition
) and the THEN
part called the
consequent
(
conclusion
or
action
).


Relation

IF


the ‘fuel tank’ is empty

THEN

the car is dead


Recommendation

IF


the
sea is very deep
AND

the sky is cloudy

AND

the forecast is
danger
THEN

the advice is
‘do not go to the sea’


Directive



IF


eat too much
raya

cakes,
rendang

AND

the
stomach is always aching
THEN

the action is
‘fasting in
S
yawal


Rules can represent relations, recommendations,
directives, strategies and heuristics:

Fuzzy logic
: human
reasoning
process
(approximation)


Differs from binary set
theory(true or false, or
1 or 0)


Similar to probability
but not same in
concept.


Fuzzy (degree of truth)


Probability (likelihood)


An elements might have
partial characteristics of
others or a subset of
something but non
-
fuzzy is more
deterministic.

Biological Network

An

artificial

neural

network


i
s

a

universal

function

approximator
.

Weights

W
ij

are

“learned”

to

fit

any

function
.


1 n









• • •

W

1

ji

W

n

kj

O
1

O
2

O
m

x
1

x
2

x
j

a
1

a
2

a
k

Artificial Neural Network

ANN

Computation

INPUT

Weights

Node(s)/Neuron

OUTPUT

AI IN INDUSTRIES

Briefing about…

Companies engaged in manufacturing


Automotive Parts


䥮摵獴s楡氠䕱E楰浥湴


High
-
tech Industry


䡥慶y 䕱E楰浥湴


Planes






and others

3 General Types of Industries

Advantages of Adding Intelligence in
Products/ Systems


Better performance


Longer Life


Reliability


Simpler operation


Cost effective


Higher efficiency


Self
-
organizing / self
-
optimization


Simpler design

Is there really a need for AI?


Manufacturers need to improve on their products


Need to satisfy customers


Need to improve products’ reliability


Need to improve products’ performance


Need to improve products’ features


Need to distinguish their products away from their
competitors

Made in Japan

(based on a video on innovations and the need for
change)

1980s


Excellent Quality


Expensive


Leadership


Balance of Payments


High Technology


Innovations

1960s


Junk


Cheap


Poor Quality


Copies


Low Technology


Imitation

World Industrial Leaders by Country

(1975
-
2000
)
(MIS for the Info Age)

In 1994 alone Japan sold US$34 billion worth of

consumer products using fuzzy logic technology

ASIMO

A
dvanced
S
tep in
I
nnovative
MO
bility

Camera Eyes
[AI]

Antenna

Battery (Fuel Cell)

Gyro Sensor Measuring
Body Angle

Actuators and Other Peripheral

Systems Controlling leg

movements [AI]

Load Sensors In Leg

Intelligent Real
-
time

Flexible Walking [AI]

The New ASIMO

The key features of the new ASIMO include:

Advanced communication ability thanks to pattern
recognition technology

1.

Recognition of moving objects



4.

Sound recognition

2.

Posture/gesture recognition



5.

Face recognition


3.

Environment recognition


Mitsubishi’s
-

Annanova

Sony’s AIBO

Fujitsu’s


HOAP

(Miniature Human Robot)

PINO ROBOTS

Sony’s SDR (Sony Dream Robots)

SDR
-
4X II

Introducing 4 new technologies


Small robot actuators (ISA
-
4)


Real
-
time Integrated Adaptive Motion
Control


Motion creation software


Real
-
time Real World Space perception


Multi
-
modal Human
-
Robot Perception

SDR
-

specifications

Intelligent Servo Actuators (ISA
-
4)

Other Features


Multi
-
face Detection


Emotional
Expression

Some Postures of SDR

Future Research in Humanoids



ROBOT


Speed (Fast)


Not tired
-
Can do repetitive job
(Fuel Cell)


Not imaginative/Not creative


Better speech and pattern
recognition


Some emotion


Entertainment


Personal Friend



HUMAN


Slow


Intelligent


Easily tired


Imaginative/Creative


Emotional


Desire


Etc.
……

Issues to be considered…


Do not apply AI when


Lack of Data


Simpler techniques are available / sufficient


Further optimization is not
possible


The AI Machine faulty


Are Robots More Intelligent than Humans?


Can Robots Replace Humans?


Human
vs

Machine

AI Applications


Group Assignments ( 1 hour )


Freshen up your industrial attachments…


List down industries application which use the
classical control in their company.


Identify any of the application in the industries
that has applied AI in their system


Propose any one of the AI method for any
suitable/appropriate system in industries and
explain why did you choose the proposed
technique.

AI Applications


Select a specific machine/system and propose
AI technique to make the machine/system
more intelligence.


You need to draw/sketch the machine and show
where to apply the AI technique


Explain why you should apply AI in the
machine/system


What have you learned today?


AI in three(3) techniques; ES, Fuzzy, ANN


Description about each techniques


Differences between AI and Classical control


Applications of AI in industries


Capability to analyze and proposing new technology
to the industries

Achieving CO1 and LO1, LO2