Energy Efficient Robust Sensing Coverage in Large Sensor Networks

1
Energy Efficient Robust Sensing Coverage in Large
Sensor Networks
2

• Paper focuses on
• Localized design
• Energy Efficient
• Robust Coverage
• Long lived
• Distributed Protocol
• Probing mechanism
• Control Working node Density
• N/w lifetime ? Sensor population

3
Background

• Sensor Networks
• Large number of sensors
• Vast hostile environment
• Issue Density Control
• Low vs. High (pros and cons)
• Irregular density to constant density..how?

4
Background (contd.)

• Energy efficiency In Ad hoc networks
• Keep only nodes involved in data forwarding
• Problem w.r.t sensor networks.
• Density Control challenges
• Fully distributed localized
• Scalable
• Simple ( computational constraints)
• Minimal state information
• Auto configuration

5
Design Choices

• 1. Adaptive Power Control
• Extend lifetime by reducing transmission range
• Degraded
• Sensing capability
• Forwarding quality

6
Design Choices (contd.)

• 2.Turn off nodes via local topology discovery
• Find neighbors Topology discovery
• Neighbors Sleeping / working?
• Problems
• Topology discovery
• How can sleeping node do this.
• Working node cannot do this because.
• Neighborhood topology working ? sleeping node
• Location dependant nature of wireless
  communication

7
Design Choices (contd.)

• 3. Topology discovery using GPS
• GPS receivers provide Geographical info.
• Nodes form Square grid.turn on/off
• Problems
• Cost of GPS receivers
• Energy cost of the same
• 4. Misc.
• Use dir. Info (using antenna)
• Not practical in sensor networks

8
The idea

• An Analogy Growth of trees in a forest
• An active node.tree
• A sleeping node.seed
• Advantages
• No location info. Needed
• No topology info. Needed

9

• The Density Control Algorithm
• Overview
• Designing issues
• probing range
• wake up rate

10
The Algorithm

• Three modes
• Sleeping, Wakeup
• Working
• Sensing and data communication
• A sleeping node wakes up after
• An exponentially distributed period of time
• Specified by wakeup rate ?

11
The Algorithm (contd.)

• Upon wake-up
• Send PRB within radius r.
• If wake-up node hears PRB_RPY within some time Tw
  
• Update ? if necessary
• sleep again for time St seconds
• Else start working continually

12
The Algorithm (contd.)

• Advantages
• Fully distributed and localized
• No per-neighbor state info.
• Reduced computation and memory overhead
• Linear system lifetime
• N/w lifetime ? Sensor population

13
The Algorithm (contd.)

• Critical design issues.
• rensures sensing coverage
• ?ensures continual density control.
• Challenges
• Sensor population is dynamically changing
• Positioning of a working node affects density
• Sensor population is loc. Dependant
• Therefore, localized ?.

14
Designing the Probing range

• Working node density ?
• No. of sensors in unit area.
• GOAL choose r to achieve ?.
• Intuitively ? 1 / S(r)
• S(r) Average occupying area.
• Find smallest Occupation Area
• Find Largest Occupation Area
• Use these to Choose a probing range.

15
Designing the Probing range (Contd.)

• Smallest Occupation Area.
• Corresponds to the densest case

16
Designing the Probing range (Contd.)

• Distance between any two nodes (D)
• r ? where ? ltlt 1
• What happens if D lt r ?
• Cannot be active simultaneously.
• Length of each edge is r / ? 3.
• Area of the hexagon ? 3 r2 / 2.
• Therefore, Maximum working node density 2/ (? 3
  r2 )

17
Designing the Probing range (Contd.)

• Largest Occupation Area

18
Designing the Probing range (Contd.)

• The distance between two adjacent nodes
• ? 3 (r?)
• What if the D gt ? 3 r.
• We can insert one more node in the middle.
• Area of the hexagon 3? 3 r2 / 2.
• Therefore, density 2/ (3? 3 r2 )

19
Designing the Probing range (Contd.)

• 3. Choosing the probing range r.
• Max. Min. case will occur.rarely.
• Experimental results show ? ? 1/ r2 .
• ? Avg. of Min. and Max densities.
• 0.77 / r2.
• c / r2.

20
Designing the Probing range (Contd.)

• 4. Achieving the Desired Redundancy
• Each are monitored by at least one node.
• Sensing range rm Coverage Area S.
• Nm ? S/? rm2 ...there must be some overlapping
• For at least K nodes to monitor an area.
• Np K Nm we know Np and Nm .for fixed rm we
  have,
• rm c / ? Krequired probing range inversely
  proportional to K

21
Designing the wake up rate - ?

• Determines how fast replacements occur.
• Depends on our tolerance to intermittence.
• The working node sets the aggregate ?d.
• Working node measures current ?d in its locality
  
• Convey this upon wake up of the sleeping node.

22
Designing the wake up rate - ? (Contd.)

• Issue
• How does a working node obtain a good estimate of
  ?a.
• How do sleeping nodes adjust themselves while
  receiving such information.

23
Designing the wake up rate - ? (Contd.)

• 1. Measurement of Aggregate wake up rate
• What would be a Straightforward method.
• Problems?
• Maintain per neighbor state
• Topology discovery
• Problems?
• Use Statistical Sampling
• Exponential distribution of sleeping
  time..Poisson process
• Ta 1/ ?a .

24
Designing the wake up rate - ? (Contd.)

• So calculate Ta to find ?a .
• Periodic averaging chosen
• Have a counter
• Ta (tk t0) / kand calculate ?a
• K is a pre specified value.
• Send ?a and ?d in the reply messages.
• Reset the counter.start next measurement.
• Found k gt 32 to suffice.

25
Designing the wake up rate - ? (Contd.)

• Auto-adjustment of Sleeping Nodes Wakeup rate
• Probing nodes receive ?d and ?a values.
• ? ? . (?d / ?a )

26
Implementation issues

• 1.Channel Errors
• Loss of PRB , PRB_RPY messages
• Solution Send multiple messages.
• 2. Measurements from multiple nodes
• Solution
• Do not update
• Maintain a flag indicating the same

27
Implementation issues (Contd.)

• 3. Probing and transmission
• Probing, transmission and sensing ranges could be
  different
• So what will probing range be
• Issues?

28
Results /Conclusion

• Working node no. uniform w.r.t time.
• ? increases gradually w.r.t time for sleeping
  nodes.
• Aggregate ? very close to desired ?.
• Capable of converting an irregular deployment to
  regular working node topology.
• Relationship b/w probing range and working node
  no.
• Efficient Coverage of the network.
• vs. GAF longer system lifetime

29