Towards Optimal Sleep Scheduling in Sensor Networks for RareEvent Detection

1
Towards Optimal Sleep Scheduling in Sensor
Networks for Rare-Event Detection

• Q.Cao, T.Abdelzaher, T.He, J.Stankovic
• Presenter Chih-Hung Wu

2
Abstract

• Proposed a protocol for node sleep scheduling
  under energy consumption constraint
• Bounded-delay sensing coverage
• Maximizing network lifetime
• Optimized for rare event detection

3
Current literature

• Random scheduling
• Synchronized scheduling
• Trade-off between energy savings and coverage

4
Rotating sensory coverage

• A better solution
• At any point in time, partially cover the area,
  but ensure any point in the area is eventually
  sensed within a finite period of time
• Minimizing packet delivery latency to a common
  base-station
• Optimized for rare (urgent) events

5
General Framework

• Goal
• Minimize average detection delay by coordinating
  the duty cycles
• Variables
• Duty Cycle Constraint
• ? Ton/(Ton Td)
• Ton warm-up time sensor reading time data
  logging time
• Td Sleep time
• Average detection delay
• The time duration when an event occurs until it
  is detected.

6
Sleep Schedule Optimization

• Detection Delay Optimization
• Local optimality
• Every node ends up with a wakeup point that
  cannot be further improved
• End-to-End Delay Optimization
• Expedite data delivery
• Synchronize duty cycles of nodes into streamlined
  sequence

7
Detection Delay Optimization
8
End-to-End Delay Optimization
9
Evaluation

• Initialization
• 300 nodes
• 100m x 100m area
• 10m of node sensing range
• Ignore packet loss and retransmission
• Limited effect on overall performance

10
Performance

• Absolute average detection delay reduction over
  random scheduling up to 15
• Longer lifetime for the same average detection
  delay
• Lower average detection delay for the same
  lifetime
• Higher detection probability for the same event
  lifetime
• Synchronization increases ADD, but higher
  communication range decreases ADD, thus balanced
  out the negative effect