Download - Constructing Load-Balanced Data Aggregation Trees in Probabilistic Wireless Sensor Networks

Transcript
Page 1: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

CONSTRUCTING LOAD-BALANCED DATA AGGREGATION

TREES IN PROBABILISTIC WIRELESS SENSOR NETWORKS

Jing (Selena) He, Shouling Ji, Yi Pan, Yingshu LiDepartment of Computer Science

Georgia State University

Page 2: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

OUTLINE

Motivation

Solution Overview

Problem Formulation and Analysis

Performance Evaulation

Conclusion2

Page 3: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

3

Motivation• Probabilistic Network Model• Load Balanced Data Aggregation Tree• Challenges

Page 4: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

TRANSITIONAL REGION PHENOMENONMotivation

4

Link Length 0 – 2.6 M 2.6 – 6 M > 6 Mλ > 80% 7 8 0

Total 8 27 15

Page 5: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

5

PROBABILISTIC NETWORK MODELMotivation

Page 6: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

6

LOAD-BALANCED DATA AGGREGATION TREE (LBDAT)

Motivation

Page 7: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

7

CHALLENGESHow to measure the traffic load of each node under

Probabilistic Network Model (PNM)? Potential load Actual load

How to find a Load-Balanced Data Aggregation Tree (LBDAT)?

NP-Complete

Motivation

Page 8: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

OUTLINE

Motivation

Solution Overview

Problem Formulation and Analysis

Performance Evaluation

Conclusion8

Page 9: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

SOLUTION OVERVIEW

9

9Load-Balanced Maximal Independent Set (LBMIS)Connected Maximal Independent Set (CMIS)Load-Balanced Parent Node Assignment (LBPNA)Load-Balanced Data Aggregation Tree (LBDAT)

Page 10: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

OUTLINE

Motivation

Solution Overview

Problem Formulation and Analysis

Performance Evaluation

Conclusion10

Page 11: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

LOAD-BALANCED MAXIMAL INDEPENDENT SET (LBMIS)

11

11

DS Property Constraint

IS Property Constraint

linearization

Relaxing

(quadratic)

ωi1 vi is a dominator0 otherwise

Page 12: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

APPROXIMATION ALGORITHM (RANDOM ROUNDING)

12

12

Due to the relaxation enlarged the optimization space, the solution of LP*

LBMIS corresponds to a lower bound to the objective of INPLBMIS .

Page 13: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

LOAD-BALANCED MAXIMAL INDEPENDENT SET (LBMIS)

13

Page 14: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

CONNECTED MAXIMAL INDEPENDENT SET (CMIS)

14

Page 15: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

Relaxing

LOAD-BALANCED PARENT NODE ASSIGNMENT (LBPNA)

15

Each dominatee can be allocated to only one dominator

Page 16: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

16

LOAD-BALANCED PARENT NODE ASSIGNMENT (LBPNA)

Page 17: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

OUTLINE

Motivation

Solution Overview

Problem Formulation and Analysis

Performance Evaluation

Conclusion17

Page 18: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

18

PERFORMANCE EVALUATION

Our method

Other’s Method

LBDAT prolong network lifetime by 42% on average compared with DAT

Page 19: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

OUTLINE

Motivation

Solution Overview

Problem Formulation and Analysis

Performance Evaluation

Conclusion19

Page 20: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

CONCLUSIONS

LBDAT is NP-Complete, constructed in three steps:Load-Balanced Maximal Independent Set (MDMIS)Connected Maximal Independent Set (CMIS)Load-Balanced Parent Node Allocation (LBPNA)

Approximation algorithms and performance ratio analysis are presented.

20

Page 21: Constructing Load-Balanced Data  Aggregation Trees  in Probabilistic Wireless  Sensor Networks

21

Q & A