Localization in Wireless Sensor Networks IPSN 07

1
Localization in Wireless Sensor NetworksIPSN 07

• Author
• Masoomeh Rudafshani and Suprakash Datta
• Department of Computer Science and Engineering
• York University
• Presenter
• Dan Wang

2
Outline

• Introduction
• Model and Algorithms
• Evaluation
• Lower bounds

3
Introduction 1/4

• Highlights of MSL MSL
• Range-free
• Work well when any number of sensors are static
  or mobile
• V.S. existing algorithms

4
Introduction 2/4

• Range-free Algorithm
• Radio signal strengths
• Angle of arrival of signals
• Distance measurements
• Requires that each node knows
• Which nodes are within radio range
• Their location estimates
• The (ideal) radio range of sensors
• ltnot necessarily?gt

5
Introduction 4/4

• Contributions
• Both MSL MSL outperform existing algo.
• They exhibit gracefu degradation in performance
  with decreasing seed density
• They converge faster and sampling procedure is
  faster than MCL

6
Model and Algorithms 1/11

• Model irregularity
• Radio range
• Where r ideal radio range
• Deploy
• First hop neighbors
• Second hop neighbors
• In a time step, each node seed can move
• where

7
Model and Algorithms 2/11

• The Monte Carlo Method
• Particle filtering approach

8
Model and Algorithms 3/11

• The Monte Carlo Method (cont.)
• 3 steps
• Initialization p(S0)
• Sampling
• Re-sampling
• The location of a node is estimated as the
  weighted mean of its samples
• Each node updates its samples in every time step

9
Model and Algorithms 4/11

• Algorithm MSL - 1.Initialization
• The first set of samples for each node is chosen
  randomly from the whole sensor field
• The nodes only use the seeds within their
  neighborhood for weighting the samples

10
Model and Algorithms 5/11

• Algorithm MSL - 2.Sampling

11
Model and Algorithms 6/11

• Algorithm MSL - 2.Sampling (cont.)
• How to compute ?
• First-hop, seed q
• Second-hop, seed q
• First-hop, node q
• Second-hop, node q

12
Model and Algorithms 7/11

• Algorithm MSL - 2.Sampling (cont.)

13
Model and Algorithms 8/11

• Algorithm MSL - 3.Resampling
• a sample with a small weight has a lower chance
  of being selected
• a sample with a higher weight has a greater
  chance of being selected

14
Model and Algorithms 9/11

• Algorithm MSL - closeness
• Used to measure the quality of a location
  estimate

15
Model and Algorithms 10/11

• Algorithm MSL
• Every node uses the weights of its neighbors
  (rather than weights of samples of neighbors) to
  weight its samples.
• Different from MSL in step2.Sampling

16
Model and Algorithms 11/11

• Algorithm MSL - 2.Sampling (cont.)
• How to compute ?
• First-hop, seed q (same as in MSL)
• Second-hop, seed q (same as in MSL)
• First-hop, node q
• Second-hop, node q

17
Evaluation 1/9

• System Model and Parameters
• 500 units 500 units square field
• Ideal radio range r 100
• Node density, nd 10
• Seed density, Sd 1

18
Evaluation 2/9

• Convergence of MSL

19
Evaluation 3/9

• Determining the number of samples

20
Evaluation 4/9

• Accuracy Comparison of Different Algorithms

21
Evaluation 5/9

• Accuracy Comparison of Different Algorithms
  (cont.)

22
Evaluation 6/9

• Effect of the Speed of Mobile Sensors

23
Evaluation 7/9

• Effect of Node and Seed Density

24
Evaluation 8/9

• Effects of Irregularity in Radio Range

25
Evaluation 9/9

• Communication Cost
• n of nodes, m of seeds, s of sampling
• MSL nsm
• MSL nm

26
Lower bounds 1/2

• Outline of the Lower Bound Proof
• Let the continuous random variable Z represent
  the maximum
• distance sensor p can be moved without changing
  its
• neighborhood. Then, the expected value EZ is a
  lower
• bound on the localization error of nodes.

27
Lower bounds 2/2

• Outline of the Lower Bound Proof (cont.)

28
(No Transcript)