GPS less Low Cost Outdoor Localization For Very Small Devices

1
GPS less Low Cost Outdoor Localization For Very
Small Devices
Nirupama Bulusu John Heidemann Deborah Estrin
Presented By Rishabh Tandon University of
Southern California
2
Aim

• To Implement a coarse grained cost effective
  localization mechanism for small cheap low power
  devices operating in a large network
• To leverage the RF communication capability of
  these devices and use connectivity metric as a
  basis for localization
• Support rapid and ad hoc deployment
• Target outdoor environments

3
Motivation

• Numerous applications for wireless sensor
  networks Most of these benefit from
  localization.
• - Intelligent low cost routing
• - Animal Tagging
• - Package Tracking
• Conventional Methods too good for wireless
  networks consisting of small low power devices
• - GPS Not feasible due to power and
  computational

4
Design Objectives

• RF based Devices
• Computation at receiver and not reference points
• Ad hoc No preplanning required for deployment
• Responsiveness Localize in fairly low response
  time
• Adaptive Fidelity Accuracy should be adaptive
  to the granularity of available reference points

5
Classification of approaches
Localization
Fine Grained
Coarse Grained
Range Finding
Signal Pattern Matching Based
Directionality Based
Timing Based
Signal Strength Based
6
Fine Grained Range finding approaches

• Timing Based use time of flight measurements to
  estimate location
• GPS
• Uses four satellites to localize the position of
  an object.
• Cannot be used indoors
• Not suitable for use on small devices
• LPS
• Subdivides the interior of a building into
  multiple cells with each cell connected up to 16
  antennas
• Suitable for indoor use but requires a lot of
  infrastructural support

7
Fine Grained Range finding approaches (contd.)

• Timing Based
• Active Bat
• Uses explicit time of arrival measurements based
  on two distinct modalities of communication,
  ultrasound and radio
• Ultrasound may not be suitable for outdoor
  environments due interference with other
  ultrasound sources
• Acoustic Rangefinder
• Strives to support ad hoc deployment by using a
  large number of sensors that exhibit better
  interference rejection in multi-path obstructive
  environment .
• Still in a highly nascent stage though

8
Fine grained range finding approaches (contd.)

• Signal Strength Based estimate location of an
  object on the basis of signal strength at that
  point.
• RADAR Compute distance from measured signal
  strength using a model based on wall attenuation
  factor
• Better results in indoor model than GPS
• However the approach is centralized as
  computation is done on base stations
• Requires infrastructural setup if RF mapping
  approach is used instead of the WAF model

9
Fine grained signal pattern matching approaches

• Combines multi-path patterns with other signal
  characteristics to generate a signature unique to
  a location and stores these signatures in a
  database
• Localization is then reduced to a mapping problem
• Like other methods this also requires
  infrastructural setup, customization and base
  station processing
• However the method has been successfully
  implemented for tracking cell phones in
  baltimore, Maryland

10
Fine grained approaches based on directionality

• VOR/VORTAC stations are used in aviation for
  localization
• Each VOR station emits a signal that is
  electrically phased so that it is different in
  each 360 degree radial
• Based on the signal that the aircraft is
  receiving, it can determine its direction with
  respect to the VOR station and estimate its
  location

11
Fine grained approaches based on directionality
(contd.)

• Asis Nasipuri and Kai Li suggest a localization
  approach based on directionality
• They estimate the direction of an object relative
  to four fixed locations and use trigonometric
  calculations to estimate exact location
• Their approach suffers from multi-path
  reflections.

12
Fine grained approaches based on directionality
(contd.)

• Small aperture direction finding is yet another
  approach
• Multiple cell sites estimate the direction of the
  cell phone
• Triangulation is then used to estimate the
  position of the cell phone
• Again since the approach requires the base
  station to do the localization calculation, it
  may not scale well ...

13
Coarse grained approaches

• Active Badge
• A mesh of IR sensors is created throughout the
  building (one sensor in each room for e.g.)
• The badge sends an IR signal to the sensor in the
  room
• The sensor sends this information to a base
  station, which displays badges position
• IR based solution performs well indoors.
  Similar concept with RF is not possible due to
  multi-path effects.
• Again infrastructural support is an issue in this
  scheme

14
In a nutshell

• The approaches discussed are not adequate because
  they fit in one of the following categories
• They are either centralized
• They require support infrastructure (not ad hoc)
• They are not computationally feasible on small
  devices
• They dont scale
• They use a mechanism that doesnt extends to
  outdoor localization

15
Connectivity based localization The way to go

• Estimate distance from base stations on the basis
  of connectivity
• Define connectivity metric as
• CMi Number of beacon packets received in time t
  from station i
• ------------------------------------------
  --------------------------------------------------
  ------------------------ x 100
• Number of beacon packets sent in
  time t from station i
• Location of the receiver with respect to station
  i f (CMi)
• Location of a receiver is given by the centriod
  of the location estimates obtained from various
  stations

16
Connectivity based localization (Contd.)

• Not all sensors are considered for localization
  estimation however. Only those sensors for which
  CMi CMThresh are considered.
• Xest, Yest (Xi1 Xi2 .. Xik)/K, (Yi1 Yi2
  .. Yik)/K
• Localization error sqrt(sqr(Xest Xa)
  sqr(Yest Ya))
• Goal is to minimize the localization error
• This can be done by increasing the range overlap
  of reference points that populate the grid

17
Validation

• The validity of the theoretical model was
  achieved by measuring connectivity at 1m
  intervals over a 10m quadrant.
• The indoor model measurements of propagation
  varied widely depending on walls etc. thus
  multi-path effects play an important role in
  indoor localization
• For outdoor model, 87 co-relation between
  experimental and theoretical model was observed.

18
Experimental Results

• Experimental test bed consisted of a receiver
  along with four other Radiometrix RPC 418 modules
  serving as reference nodes
• Localization error is minimum near the centroid
  and increases on the edges
• The theoretical and experimental curves for
  cumulative localization error follow each other
  closely

19
Granularity and overlap of reference points

• Granularity of localization improves as more
  reference points are added.
• As the ratio R/d is increased from 1 to 4,
  maximum error decreases from 0.5d to 0.25d

20
Discussion and Directions

• A mechanism for collision avoidance must be
  devised as multiple stations would be
  transmitting beacons at the same time
• The number of packets to be transmitted (S) for
  connectivity estimates, must be tuned for power
  conservation, so should be T based on R/d value
  and collision avoidance scheme
• To be useful in real world scenario, support for
  non-uniformly positioned reference points must be
  added, apparently 15 of grid points experience
  worst localization due to non-uniform reference
  points
• Issues pertaining to reference point failure must
  be addressed
• The model must be adapted to noisy radio channels

21
Design Objectives (Revisited)

• RF based Devices
• Computation at receiver and not reference points
• Ad hoc No preplanning required for deployment
• Responsiveness Localize in fairly low response
  time
• For the Experimental System,
• T (Time interval between two beacons) 2sec
• S (Number of packets sent) 20
• which implies t (receiver sampling time)
  41.9 sec
• Adaptive Fidelity Accuracy should be adaptive
  to the granularity of available reference points

22
In retrospect

• Serious criticism for centralized approaches,
  Approaches requiring infrastructural setup
• While the paper presents a good approach, it is
  far from ready for deployment
• Till then centralized approaches requiring
  infrastructural setup may be used.
• Centralized approach may not be always bad. It
  depends on the granularity.
• Cell phone operators today support more than 2
  million subscribers per city not all
  subscribers are active at the same time same
  can be expected of localization service in such a
  scenario.
• Also in the above case, infrastructural set up is
  already present. It may be easy to tap into it.

23
Thank You