Dynamic Clustering for Acoustic Target Tracking in Wireless Sensor Networks

1
Dynamic Clustering for Acoustic Target Tracking
in Wireless Sensor Networks
Faculty EWI, Department of Computer Engineering
Seminar Wireless Sensor Networks Ronny Reyes

2
Introduction

• Analyze location of (moving) target
• Energy-based localization
• System Overview
• Clustering
• Handling of regions

3
Energy-based localization

• Sensors becomes active if an acoustic target is
  sensed
• Different algorithms can be used for Localization
• Initialization process
• Triangulation

December 11, 2009
3
4
Energy-based localization

• Two energy-based location approaches could be
  used
• Voronoi diagram
• Non-linear Optimization-based

5
System Overview
6
Clustering

• Clustering process
• Sensors and Cluster Heads (CH)
• Static and dynamic clustering
• What are the drawbacks of static clustering?

7
Drawbacks of static clustering

• Fault tolerance, if CH dies than the sensors in
  that cluster are useless
• Sensors belongs to explicit CH
• Fixed membership prevents sensor nodes to share
  information with other clusters
• A cluster may not have enough sensor members
• Cannot adapt to high event concentration in a
  particular region

8
Task of CH

• An active CH has the following tasks
• Broadcast a packet that contains the energy of
  the detected signal to sensors
• Receives replies from sensors
• Estimates the location of the target
• Sent the localization results to the sink

9
Relationship between radio transmission range and
the acoustics signal detection range

• TR range determined by transmission power
• ASD range controlled by adjusting the detection
  threshold
• If the ASD range is larger than the radio
  transmission range multiple CHs might become
  active
• If the ASD range is smaller than the radio
  transmission range the localization results
  becomes more accurate
• Setup in experiments TR range twice the ASD range

10
Proposed Dynamic Clustering Algorithm

• Initial distance calibration and tabulation
• CH Volunteering
• Sensor replying
• Reporting tracking results

11
CH Volunteering

• Recruit sensors to form cluster
• Determination of back-off timer values
• D Wmin (Wmax-Wmin).(1-Pr(Id)) U(Wran)
• Two-Phase broadcast mechanism energy packets and
  signature packets

12
Handling of regions

• Case 1 CH1 transmits energy packets earlier than
  any other CHs
• Case 2 CH_I transmits energy packets earlier
  than CH1
• Case 3 CH1 and CH_I packets collide

13
Handling of regions (2)

• Collision region is composed of 4 sub regions
  R1,R2,R3and R4

R4
R3
R1
R2
14
Handling of regions (3)
15
Testing

• Area 180 x 180 m2
• 324 devices, 288 sensors, 36 CH
• 1 sink located at (0,0)
• IEEE 802.11b is adopted
• Acoustic Signal Detection range is 25m
• Radio Transmission range is 50m

16
Parameter Metrics

• Location error
• Latency
• Events detected and reported to sink
• Number of collisions through the simulation
• Total number of control messages
• Full fledge version of algorithm
• Two partial version of algorithm
• Static clustering

17
Backoff Timer

• CH1 may overlap CH_I, I2,3,4 when target is in
  region 4. W_I is the deterministic part of the
  backoff timer values and W_ran de randomized part

18
Simulation results
19
Simulation results (1)
Avg. Error m
Avg. Error m
Delay sec
Delay sec
Magnitude of noise
Maximum speed of target
20
Simulation results (continued-2)
21
Conclusions

• Without the two phase volunteering procedure, the
  first partial version of the proposed algorithm
  incurs significant collision
• Proposed algorithm perform better that other
  clustering even with effects of noise in a square
  and random deployment scenario
• Performance degrades as moving speed increases

22
Questions?
What have we learned.___