for Constructing a Reliable MCDS in Probabilistic Wireless Networks - - PowerPoint PPT Presentation

A Genetic Algorithm for Constructing a Reliable MCDS in Probabilistic Wireless Networks

Jing He, Zhipeng Cai, Shouling Ji, and Yi Pan Department of Computer Science Georgia State University, Atlanta, GA, 30303

WASA 2011

1

OUTLINE

 Motivation  Problem Definition  Genetic Algorithm  Overview  Population Initialization  Fitness function  Genetic operations  Simulation Results  Conclusions

2

3

Motivation

• Probabilistic Network Model
• Reliable MCDS
• Chanllenges

TRANSITIONAL REGION PHENOMENON

0-2.6M 2.6-6M >6M

7 > 97% 8 > 95% 6 < 5%

Total: 8 27 15 Motivation

4

RELIABLE MCDS IN PROBABILISTIC WSNS

5

Motivation

CHALLENGES

How to measure the transmission quality of CDS under Probabilistic Network Model (PNM)?

 CDS reliability: the minimum upper limit of the node-

to-node delivery ratio between any pair of dominators in a CDS How to find a minimum-sized CDS?

 NP-Hard

How to find a proper trade-off between the minimum- sized CDS and the CDS reliability while satisfying the user predefined constraint?

6

Motivation

RELIABLE MCDS (RMCDS) PROBLEM

For a WSN represented by graph G = (V, E, P(E)={<e,TSR(e)> | e }) under the Probabilistic Network Model (PNM), and a pre-defined threshold , the RMCDS problem is to find a minimum-sized node set , such that: 1) The induced graph G[D] = (D,E’), where , is connected. 2) and , , such that . 3) CDS Reliability (minimum upper limit of the node-to-node delivery ratio between any pair of dominators in a CDS) .

7 7

Problem Definition

V D  1 ) ( ,    e TSR E ] 1 , (   } ) , ( , , ), , ( | {

'

E v u D v D u v u e e E      V u  D u D v  E v u  ) , (  

8

Genetic Algorithm

• Overview
• Population Initialization
• Fitness function
• Crossover Operations

RMCDS-GA

9 9

Genetic Algorithm

ENCODE SCHEME AND POPULATION INITIALIZATION

10 10

Genetic Algorithm

FITNESS FUNCTION

11

Genetic Algorithm

 Given a solution, its quality should be accurately

evaluated by the fitness value.

D D R D R C f

D D i

CDS

• f

size the is | | reliablity CDS

• f

the is where | | ) (

2 2



GENETIC OPERATIONS

12 12

Genetic Algorithm

SIMULATION

13 13

Simulation Results

CONCLUSIONS

 We identify and highlight the use of lossy links.  In order to measure the quality of a CDS under the PNM

model, we define a new metric CDS Reliability.

 We propose a GA to build a Reliable MCDS under the PNM

model.

 We also conduct simulations to validate our proposed

algorithm.

14 14

Conclusions

15

Q & A