Range-Free Sensor Localization Simulations with ROCRSSI-based Algorithm

1
Range-Free Sensor Localization Simulations with
ROCRSSI-based Algorithm

• Matt Magpayo
• Matthew.magpayo_at_tufts.edu

2
Presentation Outline

• Introduction to WSN
• The Localization Problem
• ROCRSSI and Signed-ROCRSSI
• Implementation and Simulation Results
• Future Work

3
Why Wireless Sensors Networks?

• WSNs involve the use of numerous small, wireless
  sensors that are inexpensive and easily deployed.
• Various applications
• Habitat monitoring
• (forest fire detection, water pollutants)
• Military surveillance
• (enemy tracking, sniper detection)
• Medical care
• (smart hospitals, patient monitoring)
• Introduces Design Challenges
• Limited storage capacity, limited energy supply,
  limited communication bandwidth
• All designs must take each into consideration.

4
WSN Research Areas

• Tracking
• Detection and tracking in a sensor network
• Routing
• Routing protocols of the sensor network.
• Localization
• Location information of sensor nodes.

5
Localization

• Solution 1 Marking the location of each node as
  deployed
• Impractical for large number of nodes, limited
  mobility
• Solution 2 GPS capabilities on all nodes
• Expensive and more energy consumption
• Solution 3 Anchor Nodes
• Have a small subset of nodes have GPS. Sensors
  use them to find relative location.
• Using Ranged-Based and Ranged-Free schemes

6
Range-Based Localization

• Distance estimation
• Time of Arrival (TOA)
• measure signal propagation time to obtain range
  information
• Angel of Arrival (AOA)
• estimate and map relative angles between anchors
• Received Signal Strength Indicator (RSSI)
• use theoretical or empirical model to translate
  signal strength into distance (RADAR, SpotOn)
• Distance estimation done by
• Most methods require complex hardware.

7
Ranged-Free Localization

• Never tries to estimate the absolute
  point-to-point distance.
• Some available solutions
• Centroid Algorithm
• After receiving location information of several
  anchors node, use centroid formula to estimate
  its location
• DV-HOP
• Anchor node flood their location and hop count
  throughout the network. Nodes calculate their
  position based on the received anchor location,
  hop count and average-distance per hop.
• Ring Overlapping based on Comparison of Received
  Signal Strength Indicator (ROCRSSI)
• Reduces location of sensor to a ring of finite
  definite thickness by comparing RSSI values.

8
Summary of ROCRSSI

• Ring Overlapping based on Comparison of Received
  Signal Strength Indicator
• Basic Procedure
• Reduces location of sensor to a rings of finite
  definite thickness.
• Adds rings to grid. (increments counter in these
  positions).
• Takes region of grid with highest values.
• Center of gravity of region sensor location.
• All the sensor needs
• a list of its neighboring anchors and relative
  RSSI, and, for each anchor in that list, a list
  of their neighboring anchors and relative RSSI.
• Does not require sensor nodes to send out control
  messages

9
Improving ROCRSSI (Signed-ROCRSSI)

• Improvement
• Adding of rings to the grid where sensor cannot
  be (negative rings)

• Original Algorithm

• Allowing Negative Rings

10
Implementation and Simulation

• TinyOs and TOSSIM
• NesC programming
• Lacked signal strength simulation
• OMNet Mobility Framework
• C programming
• Open source network simulator
• Layer by layer implementation

11
Simulation Timeline

1. All anchors send a broadcast message with its
   location.
2. Other anchors upon receiving broadcast messages,
   store the locations and RSSI of the message in a
   list of their neighboring anchors.
3. After a predetermined interval of time, each
   anchor then broadcast its location, and its list
   of neighbors and RSSIs.
4. This broadcast is heard from sensor nodes,
   received, and used to compute its location.

12
Preliminary Simulations
Sensor Real Location Estimated Location
0 350,250 331,257
1 450,200 457,195
2 375,150 374,152
3 375,275 331,257
4 180,250 170,220
5 300,200 299,192
6 550,200 516,165

• Real loc 350 , 250
• Estimated loc 374 , 258

13
First extensive simulation

• Ten simulations
• 15 anchor nodes and 45 sensor nodes randomly
  placed in a 2000x2000 playground
• Error Percentage (distance error/sensor radio
  distance)
• Poor results increase in error

14
Grid Scan Algorithm and Negative Rings

• Increase of error must be attributed to the grid
  scan portion of the algorithm.
• Highest block sum approach
• High negative values near or around the area of
  intersection can throw off the grid scan, causing
  the algorithm to search elsewhere

15
Alleviating shifting

• No degrees of exclusion
• Once ALL rings were added to the grid.
• Negative values are taken as zero.
• Ten simulations of random placement were
  performed again and the results recorded.
• However an improvement from the first set of
  simulations, no overall improvement.
• Not a lot of negative rings produced.

16
Further simulations

• Anchors/Sensors
• Overall increase in accuracy with more anchors.
• Spike in Centriod at 60.
• This could be attributed to the shifting of a
  centroid that an additional anchor provides,
  ruining an otherwise accurate estimation.

17
Average Number of Neighboring Anchors

• Overall increase in accuracy with more
  neighboring anchors
• ROCRSSI and
• S-ROCRSSI significantly better than Centroid

18
Varying Anchor Placement

• Simulations on how anchor topology effects the
  estimation accuracy
• Overall decrease in accuracy
• S-ROCRSSI outperforms by 20
• Negative Rings Produced 88 of the time

19
Result Summary

• Lack of improvement to estimation accuracy in
  many cases.
• lack of cases where the information negative
  rings gave actually came into use
• Usually the negative rings only reinforced
  information that the original ROCRSSI algorithm
  already knew.
• Substantial Difference in Unattractive Topologies
• Where the negative rings actually made a
  substantial difference was when anchors were not
  placed along the perimeter.
• This caused a large amount of negative rings to
  be produced, giving the S-ROCRSSI algorithm more
  information and a better location estimate.
• Sensor nodes situated outside the perimeter of
  the anchor nodes, will obtain a more accurate
  location estimation using the S-ROCRSSI algorithm.

20
Conclusion / Possible Future Work

• Despite the lack of improvement in some cases,
  the project did still demonstrate the
  effectiveness of the ROCRSSI algorithm.
• A 16 estimation error percentage is better than
  most range-based approaches out there.
• This project also helped uncover an improving
  algorithm for sensor nodes located outside the
  anchor perimeter
• Test algorithms with actual motes in real world
  conditions.
• The sensor node could alternate which location
  algorithm it uses by somehow estimating its
  general location in respect to the perimeter of
  the network anchor nodes.

21
Questions?