Dynamic Fine-Grained Localization in Ad

Dynamic Fine-Grained
Localization in Ad-Hoc
Networks of Sensors
Authors: Andreas Avvides, Chih-Chieh
Han and Mani Strivastava
Presenter: Ram Gudavalli
10/28/03
Localization
► Localization
– determining the physical
position of an object w.r.t. some coordinate
system
► Applications to sensor networks
 Location based routing
 Provide location feedback for a sensed
phenomena (such as fire)
 Tracking
Obstacles in GPS-based Localization
► GPS
cannot work indoors
► Power consumption too high
► Cost prohibitive
► Increases node size
Localization Basics
►
Distance (or Angle) Ranging



►
Received Signal Strength
Time of Arrival
Angle of Arrival
Distance (or Angle) Estimation



Hyperbolic Tri-lateration
Triangulation
Maximum Likelihood Estimation
Localization Basics
Ranging Characteristics
► Received
Signal Strength
 RF signal attenuation as a function of distance
 For signal strength measurements, used WINS
nodes
 Inconsistent for most settings (Indoors,
between buildings, parking lot)
 Presented least square fit of two separate
power levels under an idealized setting (Football
field with nodes at ground level)
Received Signal Strength
Received Signal Strength
► Multipath,
Fading, Shadowing problems
► Range varies with altitude of radio antenna
 30m at ground level
 100m at height of 1.5m
► Nodes
must be calibrated to common scale
Ranging Characteristics
► ToA
using RF and Ultrasound
 The time difference between RF and ultrasound
 For ToA measurements use Medusa nodes
 To estimate the speed to sound, perform a best line fit
using linear regression
t = sd + k
s = speed of sound in timer ticks
d = estimated distance between the two nodes
k = constant
For this model s = 0.4485, k= 21.485831
Ranging Characteristics
ToA using RF and Ultrasound
RF/Ultrasound ToA
Medusa node
• 3m ultrasonic range
• Ranging accurate to 2cm
Signal Strength vs. ToA Ranging
► ToA
is much more reliable than received
signal strength
 Signal strength is greatly affected by amplitude
variations
 Time difference of received signals is a more
robust metric
 ToA less susceptible to multipath effects
because shortest-path signal is used
► AHLoS
uses ToA ranging
AHLoS Localization Algorithm
► Beacon
nodes
 Subset of nodes that have a known location
 Broadcast location to their neighbors
► Unknown
nodes
 Nodes with unknown location
 Measure their separation from their neighbors
 Use ranging information and beacon location
information to estimate their position
 Once a position is established, an unknown node
becomes a beacon node
Atomic Multilateration
► Unknown
node must be within one hop of at
least 3 beacon nodes
► Maximum Likelihood estimate of the node's
position can be obtained by taking min mean
square estimate of a system of distance error
equations of the form:
Atomic Multilateration
Iterative Multilateration
► Atomic
multilateration is used a basic
primitive.
► Determine position of unknown nodes with
maximum number of beacons
► When location is estimated, the node
becomes a beacon
► Disadvantage
 accumulation of error when unknown nodes
which become beacons are used in estimation
Iterative Multilateration Accuracy
Collaborative Multilateration
► Position
estimation by considering use of location
information over multiple hops
► Conditions for participation
 A node is a participating node if it is either a beacon or
if it is an unknown with at least three participating
neighbors
 A participating node pair is a beacon-unknown or
unknown-unknown pair of connected nodes where all
unknowns are participating
► Can
be used is assist iterative multilateration
where beacon density is low and requirement for
atomic multilateration not met
Collaborative Multilateration
Most basic case for collaborative multilateration
Definition given is not complete though!!
Collaborative Multilateration and
Beacon density
Node and Beacon Placement
► Localization
success
depends on network
connectivity and
beacon placement
► Probability of a node
having at least 3
beacon neighbors
Beacon requirements
Experimental Setup
Centralized vs. Distributed Schemes
► Centralized
scheme
 Ranging measurements and beacon locations are
collected at central base station
 Computed location values are forwarded back to the
nodes
 Drawbacks
► Route
to the central node must be known
► Time synchronization problem (change in network topology)
► Requires pre-planning
► Energy consumption much higher
► Robustness of system suffers (central stations fail or nodes
close to stations die)
Centralized vs. Distributed Scheme
► AHLoS
uses Distributed scheme
 Distributed setup has 6 to 10 times less
communication overhead than centralized setup
 Network traffic increases in centralized setup as
the number of beacons increase
 In distributed scheme, network traffic decreases
as the percentage of beacons increases
 Centralized implementation gives more accurate
Energy Consumption Comparison
Traffic Comparison
Conclusion
► ToA
ranging is much more accurate than
Received Signal Strength
► Present a multilateration algorithm to
perform dynamic ad-hoc localization
 Iterative multilateration accumulates error
► Distributed
scheme for implementing this
algorithm is preferable
Questions
► What
is the error rate for collaborative
multilateration?
► How well does algorithm work in nodes
without 3 participating neighbors?
► Is it fair to assume a uniformly dense
network?