Dynamic FineGrained Localization in AdHoc Networks of Sensors

1
Dynamic Fine-Grained Localization in Ad-Hoc
Networks of Sensors

• Andreas Savvides, Chin-Chieh Han and Mani B.
  Strivastava
• University of California, Los Angeles
• ACM SIGMOBILE 01
• Presented by Kisuk Kweon

2
Contents

• Introduction
• Background
• Related work
• Ranging Characterization
• Localization Algorithms
• Experimental Setup and Results
• Centralized vs. Distributed
• Conclusions

3
Introduction

• Sensor Networks and Location Discovery
• A new form of distributed information exchange
• Various Applications environmental and natural
  habitat monitoring, home networking and smart
  battlefields
• Physical location of sensor in space
• GPS is not practical
• Not work Indoors or if blocked from the GPS
  satellites
• Spends the battery life of the node
• Issue of the production cost factor of GPS
• Increase the size of sensor nodes

4
Introduction

• AHLoS (Ad-Hoc Localization awareness)
• Low cost, work indoors, not expensive
  infrastructure
• Limited fraction of the nodes knows their exact
  location
• Nodes dynamically discover their location through
  a two-phase process Ranging and Estimation
  phase
• Ranging phase
• Each node estimate its distance from its
  neighbors
• Estimation phase
• Nodes use ranging information and beacon node
  locations to estimate their positions

5
Background

• Location discovery approaches consist of two
  phases distance estimation, distance combining
• Methods for estimating the distance
• Received Signal Strength Indicator (RSSI)
• Time based methods (ToA, TDoA)
• Angle-of-Arrival (AoA)
• Methods for combining phase
• Hyperbolic tri-lateration
• Triangulation (using the direction of the node)
• Maximum Likelihood (ML) estimation

6

• Hyperbolic tri-lateration
• Triangulation
• ML Multilateration

7
Related Work

• Outdoor
• In 1970s, the automatic vehicle location (AVL)
• Determine the position of police cars
• In 1993, the Global Positioning System (GPS)
• Based on the NAVSTAR satellites (24 satellites)
• LORAN
• Use ground based beacons instead of satellites
• Indoor
• The RADAR system
• Use RF strength from three base stations
• The Cricket location support system
• Use Ultrasound from fixed beacons
• The Bat system
• Node carries an ultrasound transmitter

8
Ranging Characterization

• Received Signal Strength
• RF signal attenuation as a function of distance
• For signal strength measurements use WINS nodes
• 200MHz processor, 128 KB RAM, 1MB Flash
• 15 transmission power levels 0.12 to 36.31 mW
• A pair of RSSI(Received Signal Strength
  Indicator) registers
• Inconsistent Model because of Multipath, fading
  and shadowing effects and the altitude of the
  radio antenna
• A Model is derived by obtaining a least square
  fit for each power level

9
Received Signal Strength
Figure 1
Distance (m)
Figure 2
Figure 1 WINS Sensor Node Figure 2 Radio
Signal Strength Radio Characterization using
WINS Nodes (power level P 7, 13) Figure 3
RSSI Ranging Model Parameters for WINS nodes
Figure 3
10
Ranging Characterization

• ToA using RF and Ultrasound
• The time difference between RF and ultrasound
• For ToA measurements use Medusa nodes
• AVR 8535 processor 8 KB Flash, 512 Bytes SRAM and
  EEPROM
• DR3000 radio module two data rates (2.4 and
  19.2 kbps)
• Six pairs of 40 KHz ultrasonic transducers
• The ultrasound range is about 3 meters
• To estimate the speed to sound, perform a best
  line fit using linear regression
• For this model S 0.4485, k 21.485831

11
Figure 2
Figure 1
Figure 1 Medusa node Figure 2 Distance
measurement using ultrasound and radio
signals Figure 3 Ultrasound Ranging
Characterization
MCU time measurement
Distance (cm)
Figure 3
12
Signal Strength vs. ToA ranging

• ToA is more reliable than received signal
  strength
• Signal strength is greatly affected by amplitude
  variations
• ToA raging only depends on time difference
• AHLoS chose ToA raging

A comparison of RSSI and ultrasound ranging
13
Localization Algorithms

• Some percentage of nodes knows their positions
• Beacon nodes
• Nodes with known positions
• Broadcast their locations to their neighbors
• Unknown nodes
• Nodes with unknown positions
• Use ranging information and beacon node locations
  to estimate their positions
• Once knows its location, becomes a beacon node
• Atomic, Iterative, and Collaborative
  Multilateration

14
Atomic Multilateration

• Requirement 1
• Atomic multilateration can take place if the
  unknown node is within on hop distance from at
  least three beacon nodes. The node may also
  estimate the ultrasound propagation speed if four
  or more beacons are available
• Topology for which atomic multilateration can be
  applied

15
The difference between the measured distance and
estimated Euclidean distance
(Equation 1)
A Maximum Likelihood estimate of the nodes
position can be obtained by taking The minimum
mean square estimate (MMSE) of
equations
By setting equation 1,
squaring and rearranging term
Solve using the matrix solution for MMSE
16
Iterative Multilateration

• Use atomic multilateration
• Repeats until the positions of all the nodes that
  eventually can have three or more beacons are
  estimated

Iterative Multilateration Algorithm as it
executes on a centralized node
17
Collaborative Multilateration

• Used when atomic multilateration requirement is
  not met
• Use of location information over multiple hops
• Ad-hoc network to be G (N,E)
• Beacon nodes are denoted by a set B
• The set of unknown nodes is denoted by U
• Our goal is to solve for xu, yu ? U by minimizing
  

18
Collaborative Multilateration

• Definition 1
• A node is a participating node if it is either a
  beacon or if it is an unknown node with at least
  three participating neighbors
• Definition 2
• A participating node pair is a beacon-unknown or
  unknown-unknown pair of connected nodes where all
  unknowns are participating

19
Collaborative Multilateration

• A sensor field of 100 by 100, sensor range of 10

300 nodes
Percent resolved nodes
Percent beacons
200 nodes
Percent resolved nodes
Percent beacons
20
Experimental Setup and Results

• Testbed
• 9 Medusa nodes and Pentium II 300MHz PC
• All nodes perform ranging and transmit to PC that
  runs the localization algorithm

21
Centralized vs. Distributed
Byte Transmitted
Byte Transmitted
Network Size
Network Size
Traffic in distributed and centralized With 10
beacons
Traffic in distributed and centralized With 20
beacons
22
Centralized vs. Distributed
Energy per node (J)
Energy per node (J)
Network Size
Network Size
Average energy spent at a node during
localization 10 beacons, 20 beacons
23
Conclusions

• The use of ToA ranging is a good for fine-grained
  localization
• Fine-grained localization scheme should operate
  in distributed fashion
• Future Work
• For ranging phase we will use the combination of
  ultrasonic ToA and received signal strength RF