PowerEfficient Organization of Wireless Sensor Networks

1
Power-Efficient Organization of Wireless Sensor
Networks

• Saa SlijepcevicMiodrag Potkonjak

Computer Science Department University of
California, Los Angeles
2
Objective Short Summary

• Wireless ad hoc sensor networks
• Power-efficient and fault-tolerant coverage

3
Objective Short Summary

• Wireless ad hoc sensor networks
• Power-efficient and fault-tolerant coverage

4
Objective Short Summary

• Organize available sensor nodes into mutually
  exclusive sets
• Only one set is active at a time

5
Objective Short Summary

• Organize available sensor nodes into mutually
  exclusive sets
• Only one set is active at a time

6
Objective Short Summary

• Organize available sensor nodes into mutually
  exclusive sets
• Only one set is active at a time

7
Overview

• Wireless ad hoc sensor networks
• Informal description of the problem
• Problem formulation
• Heuristic solution
• Example
• Results
• Conclusion

8
Wireless Ad Hoc Sensor Networks
9
Wireless Ad Hoc Sensor Networks

• Novel interface to physical environment
• Sensors have always been an interface to physical
  environments, but only as generators of raw data
  for a central processing system
• Self-organized wireless ad-hoc networks of sensor
  nodes, units containing one or more sensors, some
  processing capability, a power source, and a
  wireless communication subsystem
• Simple case batch of nodes dispersed across an
  outdoors areawith the purpose to observe the
  environment
• More complex case - indoor networks of sensor
  nodes connected to existing infrastructure
• Nodes share the knowledge about the local
  environment in order to give a global picture of
  the area

10
Distinctive Properties of Sensor Networks

• Locations of sensor nodes are important
• Low energy consumption is the major requirement
• Unattended
• Interaction with the environment
• Whole network accessed as an intelligent object
  capable of answering queries and executing
  complex tasks

11
Sensor Network Applications

• Environmental monitoring, battlefield, early fire
  detection, disaster recovery, assessment of toxic
  conditions etc.

12
Snapshot of Current Projects in Sensor Networks

• Power efficient sensor nodes
• WINS (UCLA, Sensoria), SmartDust (UC Berkeley),
  Rockwell
• Localization
• UC Berkeley, UCLA (NESL, LECS)
• Protocols and applications
• WIND (MIT), SCADDS (USC)
• Xerox Smart Matter (smart materials)

13
Sensor Network Model

• Nodes are randomly deployed across a rectangular
  monitored area
• After deployment the nodes determine their
  locations
• All sensor nodes have the same available energy,
  known sensing range, and transmission range
• Sensing range is defined by the least detectable
  objects that should be detected

14
Problem Definition - Informal
15
Deterministic Deployment

• The ideal case where the area is covered by the
  minimal number of nodes with the sensing range r
  and the distance between neighboring nodes is r
  R. Williams, 1979

16
Random Deployment

• The number of the nodes that must be deployed is
  higher
• To cover a 500x500 area 19 sensor nodes with the
  sensing range 60 is needed, while if deployed
  randomly an average number of 200 nodes must be
  deployed before complete area is covered
• Having all deployed nodes active at all times
  shortens the lifetime of the network
• Active nodes send and receive messages

17
Model of Coverage

• A field is a region covered by a set of sensor
  nodes, so that all points in the region are
  covered by the same set of sensors

• Level of coverage of a field is defined by number
  of sensor nodes that cover that field

18
Coverage Distribution

• Objective - uniform coverage
• Case 1 sensor nodes are deployed only within the
  area (A)
• Case 2 nodes are also deployed into a region of
  width 30 around the area (OA)
• Case 3 after the first 250 nodes, the rest is
  deployed to parts of the area with lower density
  of nodes (ID)

19
Coverage Distribution

• Simulation parameters 500x500 area, 400 nodes,
  sensing range 60, X and Y coordinates are
  chosen from a uniform distribution

20
Most-Constrained Least-Constraining

• Find sparsely covered fields in the area and
  cover those fields first
• Prevent redundant coverage of sparsely covered
  fields within one set
• Minimize redundant coverage within one set for
  the whole area

21
Formal Definition

• Given
• SNode i Field 1, Field 3, , Field j, ...
• Field j Snode 2, , Snode i,
• One pass of the algorithm
• Step 1 choose the uncovered field that belongs
  to the minimal number of unused and unmarked
  sensors
• Step 2 for all unused und unmarked sensors that
  are covering the chosen field go through all
  other fields those sensors cover
• if a field is already covered, decrease optimum
  function f for that sensor
• if a field is not covered, increase optimum
  function f
• Step 3 Chose the sensor with highest f, cover
  all fields from the set for that sensor, and mark
  all other sensors for the chosen field
• Repeat until each field is covered

22
Example

• S1F1, F5 F1S1, S3, S5, S6
• S2F4, F5, F6 F2S5, S6
• S3F1, F3, F6 F3S3, S4, S6
• S4F3, F4, F6 F4S2, S4, S5
• S5F1, F2, F4, F5 F5S1, S2, S5
• S6F1, F2, F3, F6 F6S2, S3, S4, S6

23
Example

• S1F1, F5 F1S1, S3, S5, S6
• S2F4, F5, F6 F2S5, S6
• S3F1, F3, F6 F3S3, S4, S6
• S4F3, F4, F6 F4S2, S4, S5
• S5F1, F2, F4, F5 F5S1, S2, S5
• S6F1, F2, F3, F6 F6S2, S3, S4, S6
• The field covered by minimal number of sensors is
  F2

• Select between S5 and S6
• S5 chosen because F4 and F5 are less covered than
  F3 and F6

24
Example (contd)

• S1F1, F5
• S2F4, F5, F6
• S3F1, F3, F6
• S4F3, F4, F6
• S6F1, F2, F3, F6
• S6 is marked and cannot be used in this set

SELECTED S5F1, F2, F4, F5

• Field covered by minimal number of sensors is F3
• Choose between S3 and S4
• S3 chosen because F1 is better covered than F4

25
Example (contd)
SELECTED S5F1, F2, F4, F5 S3F1, F3, F6

• First set is selected and contains sensor nodes
  S3 and S5
• Whole area is covered
• New set should be selected

26
Example (contd)

• S1F1, F5 F1S1, S6
• S2F4, F5, F6 F2S6
• S4F3, F4, F6 F3S4, S6
• S6F1, F2, F3, F6 F4S2, S4
• F5S1, S2
• F6S2, S4, S6

27
Example (contd)

• S1F1, F5 F1S1, S6
• S2F4, F5, F6 F2S6
• S4F3, F4, F6 F3S4, S6
• S6F1, F2, F3, F6 F4S2, S4
• F5S1, S2
• F6S2, S4, S6
• The field covered by minimal number of sensors is
  F2
• S6 is selected because its the only sensor that
  covers F2

28
Example (contd)

• S1F1, F5
• S2F4, F5, F6
• S4F3, F4, F6
• Uncovered field with minimal number of sensors is
  F4
• S2 is chosen because it covers other uncovered
  field F5

SELECTED S6F1, F2, F3, F6
29
Example - Obtained Solution

• Set 1 S3F1, F3, F6, S5F1, F2, F4, F5
• Set 2 S2F4, F5, F6, S6F1, F2, F3, F6
• Unused sensor nodes S1F1, F5, S4F3, F4, F6

30
Experimental Results

• Comparison with simulated annealing

31
Experimental Results

• Comparison with simulated annealing

32
Conclusion

• Power efficiency is one of the key requirement
  for wireless ad hoc sensor networks
• Problem abstracted as a set cover problem
• Most-constrained least-constraining heuristic
  essentially optimal and fast
• Next steps distributed version and optimal
  coverage