Shuo Guo, Song Min Kim, Ting Zhu, Yu Gu, and Tian He University of Minnesota, Twin Cities.

Post on 28-Dec-2015

216 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

Transcript

Correlated Flooding in Low-Duty-Cycle Wireless

Sensor Networks

Shuo Guo, Song Min Kim, Ting Zhu, Yu Gu, and Tian He

University of Minnesota, Twin Cities

The Need for Long-Term Operation

TargetTracking

BorderControl

EnvironmentalMonitoring

HabitMonitoring

PrecisionAgriculture

InfrastructureIntegrity Analysis

InventoryControl

TrafficControl

AssistedLiving

InteractiveGaming

SpaceMonitor

For hostile environment For unattended operations For low deployment &

maintenance cost

2

How to Prolong WSNs’ Lifetime? Most effective solution: Low-Duty-Cycle

WSN◦ Sensor nodes stay active for a very short interval

and then turn off almost everything and go to sleep

◦ A node periodically switches between active state and dormant state

3

t

…Sender

Active StateDormant State

…Receivert

State-of-arts uses unicasts to do broadcasting/flooding◦ Lai et al., DCOSS’10, Sun et al., Sensys’09, Guo et al.,

MobiCom’09, Inefficient use of energy!

Existing Solutions

Active StateDormant State

B

CD

A

BCD

At

B C D

3 Transmissions in total!

What if working schedules can be flexibly changed?

Existing Solutions

Active StateDormant State

B

CD

A

BCD

At

B C D

3 Transmissions in total!

What if working schedules can be flexibly changed?

Existing Solutions

Active StateDormant State

B

CD

A

BCD

At

B+C+D

1 Transmissions in total!

Build an energy-efficient flooding tree as the basic flooding structure

Children of a common parent tune their working schedules to wake up simultaneously

Children send ACKs to acknowledge the reception

Basic Idea

How to build an energy-efficient flooding tree?◦ Traditional way that only considers link quality is

not sufficient How to avoid ACK-implosion problem?

◦ Should reduce ACKs sent back by the receivers, but still guarantee reliability

New Challenges

S

S

S

R1

S

R2

S

RN

ACK ACK ACK

Collision!

Idea of Correlated Flooding Let highly correlated nodes to wake up at

the same time and receive flooding packets simultaneously ◦ A flooding tree is built with the consideration of

both link quality and link correlation For the same sender, ACK is only sent by

the node with the lowest link quality, to eliminate the ACK implosion problem.

9

More Energy Efficient!

Eliminate ACK-Implosion!

Link Correlation Link Correlation is the phenomenon that the

reception results of a broadcasting packets at different nodes are not independent

10

Given that a broadcasting packet is successfully received by B, what is the probability that it is also received by A?

=95%: no correlation!>95%: positive correlation!< 95%: negative correlation!

Pr(A|B) Pr(A) = 95%

?

Existence of Link Correlation Experiment: 1 sender, 40 receivers, 100

packets Receivers record the reception status using a

bitmap

11

SS

SR1 1110111111

SR2

90%

80%1100111111

SR40

1101111100 70%

Pr(R1|R2) = 100%

Pr(R2|R40) = 86%

Pr(R1|R2R40) = 100%

Link Quality Link Correlation?

Existence of Link Correlation Experiment: 1 sender, 40 receivers, 100 packets Blue: distribution of link quality, # of packets

received Red: distribution of # of packets received, given the

successful reception at other links with lower link quality

12

Impact of Link Correlation on Flooding

Two-node case:

N-node case:

13

With correlated links, q12 ↑, E(m) ↓

↓ with correlation, E(m) ↓

Construction of Flooding Tree Group Division

◦ Senders divide their receivers into a number of groups with high correlation within the group

Sender Selection◦ Each receiver selects only one sender to optimize

the overall performance

14

Construction of Flooding Tree Group Division: nodes with higher

correlation are divided into the same group, using k-mean

15

Construction of Flooding Tree Sender Selection: nodes selects its flooding

parent to make the worst link in each group as high as possible

16

SR3 SR1 SR2

SS1 SS2

80%85% 95%70%

w/o link correlation:R1 chooses S2

w/ link correlation:R1 chooses S1

Idea of Correlated Flooding Let highly correlated nodes to wake up at

the same time and receive flooding packets simultaneously ◦ A flooding tree is built with the consideration of

both link quality and link correlation For the same sender, ACK is only sent by

the node with the lowest link quality, to eliminate the ACK implosion problem.

17

Evaluation Test-bed implementation

◦ 20 MicaZ nodes form a 2-hop network Simulation Setup

◦ Randomly generated network, 200~1000 nodes Baseline

◦ Traditional Energy-Optimal Tree

18

Test-bed Results

Group #

Flood

ing

Dela

y

Group #

Flood

ing

Covera

ge

Flooding Delays Flooding Coverage Ratios

Test-bed Results

Group #

Data

Pack

ets

Sen

t

Group #

AC

Ks

Sen

t

50%10

%

Energy Cost on Data Packets Energy Cost on ACKs

Simulation Results

21

Data

Pack

ets

Sen

t

Network Size Network Size

AC

Ks

Sen

t

Energy Cost on Data Packets Energy Cost on ACKs

70%

25%

Experimentally verify the existence of link correlation, followed by a theoretical study about its impact on broadcasting/flooding

Utilizes the information of both link quality and link correlation to build an energy efficient flooding tree that saves the energy cost on data packets

Receptions at highly correlated nodes are Acknowledged by only one ACK, saving the energy cost on control packets (ACKs)

Summary of Contributions

THE END

A test-bed experiment with 1 sender and 40 receivers

Sender broadcasts 100 packets while receivers recording the reception results

Existence of Link Correlation

111110111111…11100010

Length=100

Link quality: how many 1s are there in a single bitmap?

Link correlation: how many 1s are there in node A’s bitmap while the corresponding bit in node B’s bitmap is 1?

Existence of Link Correlation

A: 11011------80% link qualityB: 11011------80% link qualityCorrelation: Pr(A|B) = 100% !

Existence of Link Correlation

Existence of Link Correlation

top related