A Framework for Energy- Saving Data Gathering Using Two-Phase Clustering in Wireless Sensor Networks Wook Chio, Prateek Shah, and Sajal K. Das Center for.

A Framework for Energy-Saving Data Gathering Using Two-Phase Clustering in Wireless Sensor Networks

Wook Chio, Prateek Shah, and Sajal K. DasCenter for Research in Wireless Mobility and Networking(CReWMaN)

Department of Computer Science and Engineering University of Texas at Arlington

IEEE MobiQuitous 2004

jenchi

Outline

Introduction Related work Two-Phase Clustering

Phase I： Cluster Formation Phase II： Cluster Restructuring

Experimental Study Conclusion

Introduction

A high-density wireless sensor network can be deployed for specific information-gathering Multiple sensors generating and transmitting redundant

sensed data results in unnecessary power consumption

Designing sensor data gather algorithms is to minimize the energy consumption for network longevity The network should be designed to seek local

optimization which makes optimum use of the limited energy

Introduction

Goal of this paper The clustering optimization problem

To eliminate network-wide flooding for control information delivery

To reduces the number of data transmissions with the help of its network hierarchy

Related work

The methods of cluster schemes to elect a cluster head A self-declaration method

A first-declaration-win rule Data packets are exploited using a piggy back techniq

ue ex:： LEACH

A nomination method Each node to nominate the largest indexed node in its

radio range as a cluster head The above schemes don’t consider energy metric as

a main factor in maintaining cluster configuration

Two-Phase Clustering

We proposed a two phase clustering (TPC) scheme in multihop wireless sensor networks It is very likely for closely-located sensors to g

enerate redundant sensed data Hence, data aggregation eliminates unnecessary

data transmissions To provide an energy-saving delay-adaptive d

ata gathering platform

Two-Phase Clustering

Phase I： Cluster Formation To partitioning the network into clusters

Phase II： Cluster Restructuring Cluster members are required to search for a

neighbor closer than the cluster head within the same cluster to set up a data relay link

To obtain further improvement on energy conservation by restructuring the intra-cluster node-connectivity

Two-Phase Clustering—

Phase I： Cluster Formation

1. Every node schedule a CH advertisement with t (0~x) sec. delay

3

107 6

15912

511

8

45

13

Radius R

Sensor node

Two-Phase Clustering—

Phase I： Cluster Formation

Cluster member

Direct link

Cluster Head

Two-Phase Clustering—

Phase I： Cluster Formation

Only one cluster head in a radio range

To choose the nearest CH

advertisement

Two-Phase Clustering—

Phase II： Cluster Restructuring

To restructure the intra-cluster node-connectivity Every cluster member uses the data relay link

instead of the direct link to send its sensed data to the cluster head

Sensed data conveyed along these data relay links is aggregated at each data relay point

Two-Phase Clustering—

Phase II： Cluster Restructuring

Purposes To minimize energy consumption in collecting

sensed data while meeting delay constraints To distribute the cluster head’s workload

The construction of the data relay links is launched by the cluster head’s request to search for a data relay point

Two-Phase Clustering—

Phase II： Cluster Restructuring

Procedure for cluster head Ci

Mci： the member of Ci

Wci： have set up a data relay link : a data relay link from si to sj with n-th forwarding index

Two-Phase Clustering—

Phase II： Cluster Restructuring

Procedure for cluster members (si) of Ci

fix： si with forwarding index xrpsi： a data relay point of si

Two-Phase Clustering—

Phase II： Cluster Restructuring

a

Cluster member

Direct link

Cluster Head

Two-Phase Clustering—

Phase II： Cluster Restructuring

a

1

Cluster member

Direct link

Cluster Head

Data relay link

Two-Phase Clustering—

Phase II： Cluster Restructuring

a

1

2

345

6b

Cluster member

Direct link

Cluster Head

Data relay link

Two-Phase Clustering—

Phase II： Cluster Restructuring

a

1

2

345

6b

1

Cluster member

Direct link

Cluster Head

Data relay link

Two-Phase Clustering—

Phase II： Cluster Restructuring

a

1

2

345

6

1

b

23

4

5

6

Cluster member

Direct link

Cluster Head

Data relay link

Two-Phase Clustering—

Phase II： Cluster Restructuring

The choice of using either the data relay link or the direct link to transmit the sensed data is controlled by the cluster head The cluster head broadcasts an n-relay control

message to all the cluster members Each cluster member calculates fix mod n, if th

e result is zero： to use the direct link otherwise： to use the data relay link

Two-Phase Clustering—

Phase II： Cluster Restructuring

The CSMA/CA is used The TPC allows cluster members to use a

reduced transmission power by means of the data relay link (which is shorter than the direct link)

Two-Phase Clustering—

Delay Adaptive Data Gathering

Two-Phase Clustering—

Localized Cluster head Rotation

The length of the time that sensors serve as a cluster head varies depending on the frequency of the sensed data transmission

TPC limits the cluster head rotation to the local area, thereby saving on energy that would otherwise be consumed by unnecessary cluster head rotations

Two-Phase Clustering—

Localized Cluster head Rotation

a

1

2

345

6

1

b

23

4

5

6

CH broadcast Service Completion message： SCci

Two-Phase Clustering—

Localized Cluster head Rotation

Every cluster member schedules a CH advertisement with t sec. delay

a

1/3

1/2

1/37/15

3/10

1/6

2/7

3/72/3

4/7

1/5

1/20

Two-Phase Clustering—

Localized Cluster head RotationDelay t sec. : t←(Ei-Ec)/Ei Ei=initial energy Ec=current energy

a

1/3

1/2

1/37/15

3/10

1/6

2/7

3/72/3

4/7

1/5

1/20

Two-Phase Clustering—

Localized Cluster head RotationDelay t sec. : t←(Ei-Ec)/Ei Ei=initial energy Ec=current energy

Experimental Study

To primarily study the effects of phase II on the energy-saving and the distribution of cluster head’s workload

To present the results of the comparison of phase II to phase I The execution of phase I can be considered

as the ones of LEACH in a multi-hop setting To conduct experiments using a JAVA

thread-based implementation of TPC

Experimental Study Assumptions

Homogeneous sensors 500 X 500 m2 network space Transmission distance (sensor density)

1 sensor/200m2, 1 sensor/300m2, 1 sensor/400m2, 1 sensor/500m2

Energy consumption measurement： 1 sensor/350m2

Two communication radio ranges R=50m and 100m

Data reporting interval： 3 seconds Sensors are set to generate sensed data 1000 times Using radio electronics energy： 50nJ/bit, radio amplifier en

ergy： 100 pJ/bit, and 512 bit-size sensed data packet A 5-relay control message to have five runs for each network density to collect the expe

riment result

Experimental Study

Snapshot of an experimental result of two phase clustering with R=100m

Experimental Study—

Effect on Transmission Distance

The network density and the size of the radio range affect the transmission distance reduction radio

Experimental Study—

Effect on Energy Saving and Cluster Head’s Workload To show the distribution of the remaining ener

gy level in each sensor after 1000 rounds of sensed data transmission

Relay constraint： 5-relay

Experimental Study—

Effect on Energy Saving and Cluster Head’s Workload

Experimental Study—

Effect on Energy Saving and Cluster Head’s Workload

Conclusion

The goal of the TPC scheme is to provide an energy-saving delay-adaptive data gathering platform to ultimately extend the network’s lifetime to meet the users’ or applications’ specific

requirements such as delay constraints