Fuzzy Handover in Heterogeneous Mobile Networks

Fuzzy Handover in Heterogeneous Mobile Networks

TD-01-010-S

Presenter: Y.F. HuUniversity of Bradford, UK

University of Bradford

COST272 1st MCM, Toulouse, 11-12 October 2001

Contents

q Introductionq Handover Criteriaq Handover Algorithm

• Initiation• Decision

q Simulation Modelq Resultsq Conclusion and Further Work



Introduction

q Integration of a satellite network into a set of terrestrial wireless networks consisting of different mobile networks:• Satellite• Universal Mobile Telecommunications System (UMTS)• General Packet Radio Service (GPRS)

q Objectives:• Propose a handover scheme which takes into account

different QoS parameters and user requirements• Separates the process of handover initiation and

handover decision.



Network Architecture

MTSAT

MTUMTS

MTGPRS

T-IW

U

TE

NOC

FES

GaT

UTRAN

BSS

SGSN

HLR

GGSN

MSC

HLRSGSN

ER

ER

ER

HA

HA

IP based TE

IP based TE

3G3G

2G

FA

FA

FA

UMTS Core Network

GPRS Core Network

N-IWU

NIU

Logical connection between entities



Handover Criteria

q Technology Based QoS Parameters:• Signal Strength• Network Latency• Available Bandwidth• Network Coverage and Reliability

q User Based QoS Parameters:• User’s Preferences/ Priority• Perceived QoS• Charging Models• Service Provision



Why Fuzzy Logic?

q Can be used to solve highly complex problems where it is difficult to create a suitable mathematical model.

q Used to represent processes that are imprecise by nature.

q In fuzzy logic, an event or situation does not have to be either ‘true’ or ‘false’. Instead, these events can be classed as ‘quite true’, fairly true’, ‘very true’ or ‘not true’.



Block Diagram for Handover

Network CoverageQoS Perceived

Bandwidth

LatencyReliability

Battery Status

Cost

Priority

SS’

QoS Application Used

Priority SegmentProhibited Segment

Terminal Type

Bit Error RateSystem Fuzzifier

Averaging SS SS’

FuzzyController

NewAVG INT

Inference

Fuzzy Rules

Defuzzifier Handover

Bandwidth

LatencyReliability

Battery Status

Cost

Priority

SS’

W_Bandwidth

W_LatencyW_Reliability

W_Battery Status

W_Cost

W_Priority

W_SS’

User

FuzzyOrdinalRanking

DetermineWeighting of Criteria M

ultip

le O

bjec

tive

Dec

isio

n M

akin

g

Chosen Segment

Handover Initiation

Handover Decision

Yes

No



Handover Initiation

q Consists of 3 different stagesq 1st stage: Fuzzification

• Data collected from the system is fed into a fuzzifier, to be converted into fuzzy sets.

µ

SS1 SS2 SS3

1Medium StrongWeak

Signal Strength

µ

SS1 SS2 SS3

1Medium StrongWeak

Signal Strength

(a) Fuzzy Set (b) Crisp Set



Handover Initiation

• Membership values are obtained by mapping the values of a particular parameter onto a membership function.



Handover Initiation

q 2nd Stage: Inference Engine• A Set of IF-THEN rules is applied to the system• for eg:

• IF signal strength is strong AND QoS is good AND Bit Error Rate is Medium, AND Network Coverage is Medium, THEN handover =NO

SignalStrength

PerceivedQoS

BER Coverage Dec

Weak YMedium Medium Medium Bad PNMedium Medium Medium Medium PNStrong Good Medium Bad NStrong………

Good…….

Medium………..

Medium……….

IF-THENRuleà

N…



Handover Initiationq 3rd Stage: Defuzzification

• combining the results of the first and second stage• Defuzzification performed using the centroid

calculation method, in which the center of gravity of the membership function is computed.



Handover Decision

q Uses a Fuzzy Multiple Objectives Decision Making Algorithm.

q Objectives: Select a target segment that can fulfil the following objectives:• Low Cost• Good Signal Strength• Optimum Bandwidth• Low Network Latency• High Reliability• Long Battery Life



Handover Decision

q Consists of 2 different stages.q Stage 1: Ordinal Ranking and Weighting of Criteria

• Ranking Procedure: Used to compare the performance of each segments for a particular handover decision criterion.

• for eg: if the following values are assigned to the three segments, GPRS performs best for that particular criteria followed by UMTS and SAT.

=SATUMTSGPRS

C 4925.0,6667.0,0000.11



Handover Decision

• Weighting of Criteria: Developed by obtaining a ratio scale for the criteria based on a paired comparison of each criterion.

• Assume that C={C1,...CP } is a set of p handover criteria. A pairwise comparison of each criterion with respect to the other criteria is performed to obtain matrix, B:

• The final weighting is obtained by finding the unit eigenvector of the matrix, W, which corresponds to the maximum eigenvalue of B, and then multiplying the values with the number of criteria.



Handover Decision

q 2nd Stage: Decision Making• The final decision is obtained by raising the fuzzy set

determined in stage 1 to α.• Therefore, the fuzzy subsets then becomes:

n

nCCCCCD ααααα ∩∩∩∩= ...4321

4321

⇒

=

131

311111

31313333

331333313

13

111331

131

311133

113

13

13

13

1131

131

313331

7

6

5

4

3

2

1

7654321

CCCCCCC

CCCCCCC

B

=

1911.05108.07083.01959.01959.01276.03284.0

W ⇒α

==

1.33753.5756495821.37111.37110.89322.2988

Wn



Simulation Model (Handover Initiation)

q Handover Initiation• Objectives: demonstrates how the value of the

handover factor changes when the inputs of the are varied using a normal distribution.



Results and Analysis

q When values fed into the fuzzifier are weak, handover factor is high and vice versa.

q Results shows that the algorithm behaves correctly under different operating conditions.

00.10.20.30.4

0.50.60.70.80.9

1

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Handover Factor, h

P

Scenario 1 Scanario 2 Scenario 3

P(H

F≤h

)

Handover Initiation



Simulation Model (Handover Decision)

q Handover Decision• Objectives: Demonstrates that by changing certain

inputs into the system, the choice of the selected segment will differ.

Criteria Attributes Applications Real-Time Delay Sensitive (A)

Real-Time, Delay Insensitive (B) Non Real-Time, Interactive (C)

Non Real-Time, Non Interactive (D) Net. Attr. GPRS UMTS SAT

Cost Charging 0.90 0.80 0.70 Bandwidth 0.80 0.90 0.70

Latency 0.60 0.80 0.50 Reliability 0.90 0.90 0.50

QoS

Sig. Str. 0.70 0.69 0.80 Battery 0.90 0.60 0.70 Others

Segment Priority 0.50 0.50 0.50



Results and Analysis

012

3456

Charging Bandwidth Latency Reliability Sig.Strength

Battery Priority

A

B

C

D

0

1

2

3

4

5

6

Charging Bandwidth Latency Reliability Sig. Strength Battery Priority

A

B

C

D

Case (a)

Case (b)

Conditions Result

Real-Time Delay Sensitive UMTS, SAT

Real-Time, Delay Insensitive UMTS, SAT

Non Real-Time, Interactive GPRS, UMTS, SAT

Case (a)

Cost

Non Real-Time, Non Interactive GPRS, UMTS, SAT

Real-Time Delay Sensitive UMTS, SAT

Real-Time, Delay Insensitive SAT, UMTS

Non Real-Time, Interactive GPRS, UMTS, SAT

Case (b)

Quality

Non Real-Time, Non Interactive SAT, GPRS, UMTS

q Results show that a user’s requirements play an important role in determining the weighting of the criteria. The weightings arebased on user’s preferences and applications.

q In addition, it is also found that the chosen segment is not dependent only on one criterion but also takes into account other factors as well.



Conclusion

q A handover algorithm based on the concept of fuzzy logic has been presented.

q The algorithm allows the system to decide not only on the time to initiate a handover but also on the best segment to transfer to.

q When tested under different environments, the algorithm produces correct handover responses.

q Thus, fuzzy logic offers a feasible solution in solving handover problems in a heterogeneous environment.

Documents

Fuzzy Handover in Heterogeneous Mobile Networks