Sif Project

DATA AGGREGATION IN WSNs

Sensors And Information Fusion 6933

Instructor: Dr.Bill Buckles

Ning ‘Martin’ Xu Kalyan Pathapati Subbu

Shijun Tang

TOPICS

• Introduction• Problem Definition• Clustering• Aggregation• Methodology• Experiments• Performance Analysis & Results• Conclusions

WIRELESS SENSOR NETWORKS

What?• Provide new paradigm

for sensing and disseminating information

• Collection of micro mechanical devices

• Each device capable of wireless communication and signal processing

Features• small size• Robustness• large area coverage• enhanced monitoring

precision

Applications:• Health applications• Environmental and

Structural monitoringDesign Aspects:• Deployment, mobility• Topology, Density• Self configuration• Security

• Environmental monitoring( a group of temperature sensors)

• Similar or even identical readings, minimal difference

DATA AGGREGATION

• Data aggregation is the process of combining similar data from multiple sources to eliminate such redundant transmission and provide fused information to the base station

• Compression can be performed combining multiple data packets into one to reduce overhead of control information (as opposed to data)in the transmission

• Results of arithmetic operations on the data set, such as the average, maximum and minimum, can be sent instead of the original data

• Data In Feature Out form of fusion

PACKAGE

• Hardware– MicaZ mote

• 7.38 MHz Atmel processor with a 128 KB program memory,

• 4 KB RAM and 512 KB non-volatile storage.

• Chipcon SmartRF CC2420, with 2.4GHz frequency

– MTS310 sensor board– Mib520 programming

board

• Software– TinyOS: OS for wireless

sensor networks.– nesC: programming

language for TinyOS.

PROBLEM DEFINITION

Problems:• Energy consumption

– Operating on small batteries : intangible cost to lose data due to battery depletion

• Computational Costs• Storage constraintsSolution for Energy Efficient

operation:• Topology control

– CLUSTERING• Efficient data collection

– AGGREGATION

CLUSTERING

• Grouping of sensors• Distance or proximity • Signal Strength• Logical organizing

• Topology control approach• Load balancing, network

scalability

• Types of clustering• Static: local topology control• Dynamic: changing network

parameters• Single hop and multi hop• Homogeneous and

heterogeneous

HEED- Hybrid Energy Efficient Distributed clustering

• Assumptions:• Sensor quasi-

stationary• Links are symmetric• Energy consumption

non-uniform for all nodes

• Nodes-location unaware

• Processing and communication capability-similar

Algorithm:• Cluster head selection• Factors:

• Primary- residual energy • Secondary-

communication cost • Number of rounds of

iterations • Tentative CHs formed • Final CH until CHprob=1

• Different power levels used for intra and inter-cluster communication

AGGREGATIONWhat?• Process of combining similar data from multiple sources

– Eliminate redundant transmission – Provide fused information to the base station

How?• Sum, Average, Maximum and Minimum

Scenario• Environmental monitoring:

– Group of temperature sensors within the vicinity of one another– Moreover, readings from a single sensor – minimal difference

during a certain period of time in the day.– Primary interest reducing the redundancy coming from

different sensor sources– Average might be sufficient for a small region

METHODOLOGY Exp 1: Clustering and Aggregation iHEEDX• Cluster the nodes according to HEED• Nodes sense temp, light and send to

respective CHs• CH performs aggregation and sends to

Base station

Exp 2: No Clustering and No Aggregation Collection Tree

• Individual nodes sense temp and light

• All nodes directly send to Base station

ENERGY CONSUMPTION• Transmission

– Inter Cluster power level

• Collection Tree all nodes use this power level

• iHEEDX : Only CHs use this power level

– Intra Cluster power level

• All non CH nodes use this power level

• Aggregation

– CH performs Averaging operation on the readings received

– Energy calculated for number of instructions executed by processor

• CREP : Credit point system

CREP System

• The smallest energy can be expressed and well-represented as a multiple of 1 uJ.

• The points in CREP are therefore assigned: The battery capacity is– Battery capacity 23,760,106 points

– other components 70,380 points

– transmission 860 points/packet

– reception 90,000 points

– radio idle state 171 points.

EXPERIMENTAL SETUP

• Nodes placed into three groups – Group 1: 2 and 5, B250

– Group 2: 4 and 7, B245

– Group 3: 1, 3 and 6, B251

• Different places chosen

–Variation in sensed values

Metrics Collected

• Temp, Light

• Overhead_Agg

• Overhead_NoAgg

• Packets_Recvd at BS

• Packets_Recvd_Org at Indv nodes

• Packets_Count sent by Indv nodes

EXPERIMENTAL SETUP (cont’d)


Group1


Group3


Group2


Base Station

• Effect of Data aggregation on sensed data

• Cluster size effect on Energy Consumption and Aggregation

• Overhead comparison for Aggregated and Non-Aggregated scenarios

PERFORMANCE ANALYSIS

EFFECT OF DATA AGGREGATION

• Readings of individual nodes 2 and 5

• Averaged readings from CHs

• Similar data, reduced redundancy, ENERGY SAVED!

0 20 40 60 80 10019.5

20

20.5

21

21.5

Time (seconds)

Tem

pera

ture

(C

)

No.2

No.5Agg

IMPACT OF CLUSTER SIZE

• Cluster size ranging from 1 to 7 nodes

• More the number aggregation performed, ENERGY SAVED!

1 2 3 4 5 6 7540

560

580

600

620

640

660

680

Cluster Size

Ene

rgy

Con

sum

ptio

n (p

oint

s)

Without Aggregation

With Aggregation

OVERHEAD INCURRED

0 100 200 300 400 500 600 7000

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5x 10

5

Number of packets transmitted

Ene

rgy

Con

sum

ptio

n(po

ints

)

With Aggregation

Without Aggregation

• Fewer transmissions, lesser transmission power for intra cluster communication, ENERGY SAVED!

• Joint advantages of clustering and data aggregation

• Experiment in real testbed

• Empirical results confirm energy conservation

Hurdles:• Steep Learning curve - TinyOS

• Hardware issues

Future Work:

• Data aggregation with no prior knowledge

• Outdoor experimentation

CONCLUSIONS

Thank you! Questions?

Sif Project

Technology