1. Node Selection in Distributed Sensor Networks
Olawoye Oyeyele
03/04/2004

2. Outline Motivation Aims and Objectives Modeling Assumptions
Selection Algorithm
Simulation
Conclusions

3. Motivation
Densely deployed wireless sensor networks consume energy through communications
Bandwidth may not be sufficient for data transmission in such networks
Subset of data may provide acceptable detection
Computational burden of processing all available data may be prohibitive.
Network may be short-lived

4. Aims & Objectives
Select a subset of sensors (hence data) for detection
Consistently achieve coverage of entire region
Do not compromise detection performance
Minimize complexity of selection algorithm
Low Energy consumption
Small Latency
Achieve fault-tolerance through redundancy

5. Modeling Assumptions
Target can be modeled as an isotropic radiating source with power law decay
Data captured by nearby sensors are similar, therefore redundant
Spatial dependence can be estimated
Sensor network is a sampling problem in which the spatial process is sampled spatio-temporally
Nodes are capable of power control

6. Typical Environment Nodes fail unpredictably
High Probability of attack from weather and other agents
Nodes operate autonomously

7. Sensor Network Properties
Consists of battery operated sensor nodes
Deployed randomly on the field in large numbers
Resource constrained sensor nodes
power, computation, communication
No networking infrastructure

8. Systematic Sampling
Sample the sensor field as systematically as possible within discs of radius r
Exploit spatial dependence to select sensors

9. Estimating Spatial Dependence
The Semi-variogram is the major tool used in estimating spatial dependence
Is the semi-variance at lag h, h
The lag i.e distance between locations si and sj yi
Value of variable y at location si yj
Value of variable y at location sj
N(h)
Number of pairs of observed data points separated
by lag h

10. The Semi-variogram
Shows the semi-variance plotted vs. lag for different lags.
At a certain lag (distance) called the Range, the data measurements cease to be correlated.
Many theoretical models in use; Power-law, exponential, Gaussian, spherical etc.

11. The Semi-variogram

12. Power-law Variogram A power law variogram implies infinite variance
Solution:
select arbitrary value for ‘Range’ r.
Only requirement is r < rs and r < rt where rs and rt are the sensing and maximum transmission range of sensor node

13. Selection Algorithm
Estimate “Range” or “Correlation Length” and use as a basis for selecting sensors
Simple Packet containing the source node’s remnant energy is sent RM
RM – Remaining Energy
32 bit Number representing the remnant energy on the Node

14. Properties of spatial selection
Ensures that data is collected from all over the area covered by the network
Reduces communication cost
Chosen ‘Range’ used to select sensors
Range may be seen as limit of spatial dependence
Range can be chosen to ensure coverage and connectivity
Achieves high fault-tolerance and may work well in harsh environments

15. State Machine
SLEEP
NEGOTIATE
ACTIVE

16. Pseudo-Code createNodes(number); deploy nodes(AREA);
//transmit up to Range using appropriate energy level
nodes[i].transmit(corRange)
Do { SpatialSelect()
if nodes[i].status = ACTIVE
if nodes[j] == nodes[i].neighbor && nodes[j].status == ACTIVE
nodes[j].receive();
if nodes[j].energy <= nodes[i].energy
nodes[j].status = SLEEP for a predetermined time;
else if nodes[j].energy > nodes[i].energy
nodes[j].status = ACTIVE for predetermined time;
end
end SpatialSelect();
} while (predetermined time) //executed after predetermined time

17. Applicable Scenarios Data Selection Sensor Selection
Refers to the determination of sites where data will be collected
Sensor Selection
The problem of choosing the sensors to be used in detection
Algorithm may be applied at various points in the lifetime of a network
Post-deployment
Change of cluster – head
etc.

18. Simulation Parameters
Operation
Energy Dissipated
Transmitter/Receiver Electronics
50nJ
Transmit Amplifier
100pJ/bit/m2
Parameter
Value
Transmission Rate
100kbps
Number of Nodes
200
Area
50m by 50m
Topology
Uniformly Random

19. Simulation Results Energy consumption
Detector Performance (Receiver Operating Characteristic)
Demonstrate that chosen subset achieves coverage and connectivity

20. Sensor Network

21. Node Configuration : Range = 1

22. Node Configuration: Range 3

23. Node Configuration : Range = 5

24. Selected Sensors vs. Range Values

25. Energy Depletion with Time

26. Energy Consumption

27. ROC performance

28. Discussion
Spatial Selection promises to be a robust selection technique
Consistent coverage and connectivity, fairly stable performance (ROC)
Simplified technique
Increased network lifetime
Flexible – offers two design parameters
Range Value
Sleep Time
Performs better than Randomization or pure Random Selection

29. References Clark Isobel, ‘Practical Geostatistics’ (on the web)
Xu Yingyue, Hairong Qi, ’Decentralized Reactive Clustering in Collaborative Processing Using Different Computing Paradigms’
Sestok C.K., Maya R. Said, Alan V. Oppenheim, ’Randomized Data Selection in Detection with Applications to Distributed Signal Processing’,Proceedings of the IEEE, September, 2003
Dalenius T., Jaroslav Hajek, Stefan Zubrzycki,’On plane sampling and Related Geometrical Problems’,Fourth Berkeley Symposium, 1961.
Ripley Brian D.,’Spatial Statistics’,John Wiley and Sons, 1981.