Survey: Target Tracking in Wireless Sensor Networks

1
Survey Target Tracking in Wireless Sensor
Networks

• By Tim St. John

2
Introduction

• Applications
• Military
• Golf courses
• Endangered Species Research
• Research
• Algorithm Optimization
• Sensor Deployments

3
Overview

• Power Conservation related to Quality of Service
• Classification Algorithms
• Deployment Strategies in a Real-Time Target
  Tracking System.

4
Power Conservation

• 2004 Research in Power Conservation
• University of California
• Sleeping Protocols
• RIS
• PEAS
• GAF
• Mesh

5
New sleeping Protocol

• PEAS
• (Probing Environment and Adaptive Sleeping)
• Stay awake until energy is depleted
• PECAS Derived From PEAS
• (Probing Environment and Collaboration Adaptive
  Sleeping)
• Remains working for more than 1 sec.

6
Experiment

• GloMoSim
• 400x400
• Target moves at a rate of 10 m/s
• 800 nodes
• Transition Radius range from 20 m to56 m.
• Work times range from 1, 4 and 10 sec

7
Experimental Results
8
Classification Algorithms

• 2003 Pennsylvania State University and the
  University of Wisconsin
• Different forms of CSP in Sensor Networks
• Single Node Multiple Modality
• Multiple Node Single Modality
• Multiple Node Multiple Modality
• Decision Fusion
• Data fusion

9
Test Conditions

• Conducted Small scale Simulation in the DARPA
  SenselT Program.
• Tests conducted on 3 types of Vehicles
• Humvee
• Amphibious Assault Vehicle
• Dragon Wagon
• Compare the performance of 2 types of Algorithms.
• Data Fusion Classifiers
• Decision Fusion Classifiers

10
Test Results

• Test 1 Confusion Matrix for Single Node
  Classifiers
• For both cases Average PD .72
• Test 2 Confusion Matrix for Data Fusion
  Classifiers
• Average PD of .81
• Test 3 Confusion Matrix for Decision Fusion
  Classifiers
• Average PD of .78

11
Real Time Target Tracking Deployment Strategies

• 2007 Paper by University of Virginia, Minnesota,
  Illinois
• VigilNet System
• Tested the system in 2 tests
• 200 XSM motes 10 meters apart.
• Large Scale Simulations 10000 nodes within a
  100,000 m2 rectangle.

12
Target Detection Phases
13
Node Deployment
14
Detection Strategies

• Higher Density of sensor deployment for slow
  moving targets.
• Detection Strategies concentrated on the
  perimeter.
• Power Savings
• Reduced Communications
• Honeycomb deployment Simulation Results.

15
Conclusion

• Power Conservation related to Quality of Service
• Classification Algorithms
• Deployment Strategies in a Real-Time Target
  Tracking System.

16
Questions?

• tdstjohn_at_gmail.com