Initializing Newly Deployed Ad Hoc and Sensor Networks

1
Initializing Newly Deployed Ad Hoc and Sensor
Networks

• Fabian Kuhn, Thomas Moscibroda, and Roger
  Wattenhofer
• (MobiCom04)
• 2005.10.25
• Presented by Choi, Sik

2
Contents

• Introduction
• Model
• Algorithm
• Clustering algorithm
• Analysis
• Simulation
• Conclusion

3
Introduction

• Ad hoc and sensor network
• Many applications
• Disaster relief, community mesh networks,
  monitoring and surveilance, or data gathering
• Initialization problem
• Self-organizing a structured multi-hop radio
  network
• But, no established pattern, no information of
  neighbors.
• Propose clustering algorithm
• Cluster heads (dominators) coordination point
  for MAC scheme
• First algorithm addressing the requirements of
  newly deployed networks

4
Model

• Quasi Unit Disk Graph (QUDG)
• Two nodes are connected their Euclidean distance
  is less than or equal to d.
• Wireless multi-hop networks
• Nodes do not feature a reliable collision
  detection algorithm
• Nodes can wake up asynchronously
• No global clock
• Nodes have no knowledge about the other nodes
  distribution or wake-up pattern

5
Algorithm (1/3)

• 3 channels
• channel for competing to become dominators
• , channels for remaining small
  dominators
• Start and waiting phase
• Find existing dominators
• Decide the node knows of the existence of a
  dominator in its neighbors

6
Algorithm (2/3)

• Completion phase 1 ( )

7
Algorithm (3/3)

• Completion phase - 2

8
Analysis (1/3)

• Di be the circle centered at the center of Ci
  having radius Rdd/2
• Every node in a circle Ci can hear all other
  nodes in Ci
• Nodes outside Di are not able to cause a
  collision in Ci

9
Analysis (2/3)
10
Analysis (3/3)
11
Applications

• Energy saving protocol
• Each period consist of a sleeping phase and
  listen phase
• Dominators use their message to synchronize
  associated nodes
• All nodes in the same cluster listen and sleep at
  the same time

12
Simulation (1/3)

• Unit Disk Graph (UDG) model
• d 1
• s the number of sleeping nodes at time t
• In time slot t1, each sleeping node wakes up
  with probability (np)/s
• P1 synchronous wake-up behavior (all nodes
  wake up immediately)

13
Simulation (2/3)
14
Simulation (3/3)
15
Conclusion

• Efficient structuring algorithm in wireless ad
  hoc or sensor network after its deployment
• Propose clustering algorithm
• Analysis
• Further work
• Mobility
• Node-failure